From 58c77dd3d83a773622b84d355112121ebfa41b3e Mon Sep 17 00:00:00 2001
From: Sarah Cohen <sarah.h.cohen@asu.edu>
Date: Thu, 28 Dec 2023 17:06:06 -0500
Subject: [PATCH] Re-rendered to get links right

---
 docs/advanced-openrefine.html                 |  88 +--
 docs/advanced-pdf.html                        |  72 ++-
 docs/advanced-regex-r.html                    | 106 ++--
 docs/advanced-regex.html                      |  88 +--
 docs/advanced-scrape1.html                    |  70 ++-
 docs/advanced-scrape2.html                    |  68 +-
 docs/advanced.html                            |  40 +-
 docs/appendix-filetypes.html                  |  36 +-
 docs/appendix-math.html                       |  38 +-
 docs/appendix-ppp.html                        |  36 +-
 docs/appendix-program.html                    |  36 +-
 docs/assets/images/r-groupby-squished.png     | Bin 19820 -> 0 bytes
 docs/index.html                               |   2 +-
 docs/quickstart.html                          |  36 +-
 docs/r-data-import.html                       |  56 +-
 docs/r-install.html                           |  36 +-
 docs/r-quarto.html                            |  38 +-
 docs/r-recipes.html                           |  92 +--
 docs/r-start.html                             |  38 +-
 docs/r-verb-mutate.html                       |  16 +-
 docs/r-verbs.html                             |   8 +-
 .../figure-html/hot-100-top-10-list-1.png     | Bin 0 -> 173820 bytes
 docs/search.json                              | 585 ++++++++----------
 docs/sitemap.xml                              |  86 +--
 docs/start-build-own.html                     |  36 +-
 docs/start-data-def.html                      |  48 +-
 docs/start-hunt.html                          |  36 +-
 docs/start-tidy.html                          |  36 +-
 docs/start.html                               |  36 +-
 docs/vis-demo.html                            |  70 ++-
 docs/vis-maps-begin.html                      |  64 +-
 docs/vis-maps-firstmap.html                   |  74 ++-
 docs/vis-practice.html                        |  64 +-
 docs/vis-reporting.html                       |  56 +-
 docs/vis-tables.html                          |  92 +--
 docs/vis.html                                 |  40 +-
 start-data-def.qmd                            |  11 +-
 37 files changed, 1235 insertions(+), 1099 deletions(-)
 delete mode 100644 docs/assets/images/r-groupby-squished.png
 create mode 100644 docs/r-verbs_files/figure-html/hot-100-top-10-list-1.png

diff --git a/docs/advanced-openrefine.html b/docs/advanced-openrefine.html
index 8f20446..0328399 100644
--- a/docs/advanced-openrefine.html
+++ b/docs/advanced-openrefine.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 28&nbsp; Open refine walkthrough</title>
+<title>Data Reporting - 29&nbsp; Open refine walkthrough</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -110,7 +110,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-openrefine.html"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-openrefine.html"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -267,35 +267,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -312,37 +318,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -359,25 +365,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -421,33 +427,33 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#installing-openrefine" id="toc-installing-openrefine" class="nav-link active" data-scroll-target="#installing-openrefine"><span class="header-section-number">28.1</span> Installing OpenRefine</a>
+  <li><a href="#installing-openrefine" id="toc-installing-openrefine" class="nav-link active" data-scroll-target="#installing-openrefine"><span class="header-section-number">29.1</span> Installing OpenRefine</a>
   <ul class="collapse">
   <li><a href="#mac-users-unidentified-developer" id="toc-mac-users-unidentified-developer" class="nav-link" data-scroll-target="#mac-users-unidentified-developer">Mac users: Unidentified developer</a></li>
   <li><a href="#windows-users" id="toc-windows-users" class="nav-link" data-scroll-target="#windows-users">Windows users</a></li>
   <li><a href="#limitations" id="toc-limitations" class="nav-link" data-scroll-target="#limitations">Limitations</a></li>
   <li><a href="#following-the-walkthrough" id="toc-following-the-walkthrough" class="nav-link" data-scroll-target="#following-the-walkthrough">Following the walkthrough</a></li>
   </ul></li>
-  <li><a href="#starting-openrefine" id="toc-starting-openrefine" class="nav-link" data-scroll-target="#starting-openrefine"><span class="header-section-number">28.2</span> Starting OpenRefine</a></li>
-  <li><a href="#importing-data" id="toc-importing-data" class="nav-link" data-scroll-target="#importing-data"><span class="header-section-number">28.3</span> Importing data</a>
+  <li><a href="#starting-openrefine" id="toc-starting-openrefine" class="nav-link" data-scroll-target="#starting-openrefine"><span class="header-section-number">29.2</span> Starting OpenRefine</a></li>
+  <li><a href="#importing-data" id="toc-importing-data" class="nav-link" data-scroll-target="#importing-data"><span class="header-section-number">29.3</span> Importing data</a>
   <ul class="collapse">
   <li><a href="#returning-to-your-project" id="toc-returning-to-your-project" class="nav-link" data-scroll-target="#returning-to-your-project">Returning to your project</a></li>
   <li><a href="#rename-or-copy-columns" id="toc-rename-or-copy-columns" class="nav-link" data-scroll-target="#rename-or-copy-columns">Rename or copy columns</a></li>
   </ul></li>
-  <li><a href="#facets" id="toc-facets" class="nav-link" data-scroll-target="#facets"><span class="header-section-number">28.4</span> Facets</a></li>
-  <li><a href="#clustering" id="toc-clustering" class="nav-link" data-scroll-target="#clustering"><span class="header-section-number">28.5</span> Clustering</a>
+  <li><a href="#facets" id="toc-facets" class="nav-link" data-scroll-target="#facets"><span class="header-section-number">29.4</span> Facets</a></li>
+  <li><a href="#clustering" id="toc-clustering" class="nav-link" data-scroll-target="#clustering"><span class="header-section-number">29.5</span> Clustering</a>
   <ul class="collapse">
   <li><a href="#clustering-details" id="toc-clustering-details" class="nav-link" data-scroll-target="#clustering-details">Clustering details</a></li>
   <li><a href="#facet-for-long-entries" id="toc-facet-for-long-entries" class="nav-link" data-scroll-target="#facet-for-long-entries">Facet for long entries</a></li>
   </ul></li>
-  <li><a href="#filtering" id="toc-filtering" class="nav-link" data-scroll-target="#filtering"><span class="header-section-number">28.6</span> Filtering</a>
+  <li><a href="#filtering" id="toc-filtering" class="nav-link" data-scroll-target="#filtering"><span class="header-section-number">29.6</span> Filtering</a>
   <ul class="collapse">
   <li><a href="#simple-filter" id="toc-simple-filter" class="nav-link" data-scroll-target="#simple-filter">Simple filter</a></li>
   <li><a href="#complex-filter" id="toc-complex-filter" class="nav-link" data-scroll-target="#complex-filter">Complex filter</a></li>
   <li><a href="#split-by-regex" id="toc-split-by-regex" class="nav-link" data-scroll-target="#split-by-regex">Split by regex</a></li>
   <li><a href="#split-using-openrefine-programming-language" id="toc-split-using-openrefine-programming-language" class="nav-link" data-scroll-target="#split-using-openrefine-programming-language">Split using OpenRefine programming language</a></li>
   </ul></li>
-  <li><a href="#finishing-up" id="toc-finishing-up" class="nav-link" data-scroll-target="#finishing-up"><span class="header-section-number">28.7</span> Finishing up</a></li>
+  <li><a href="#finishing-up" id="toc-finishing-up" class="nav-link" data-scroll-target="#finishing-up"><span class="header-section-number">29.7</span> Finishing up</a></li>
   </ul>
 </nav>
     </div>
@@ -456,7 +462,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-openrefine" class="quarto-section-identifier"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></h1>
+<h1 class="title"><span id="advanced-openrefine" class="quarto-section-identifier"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></h1>
 </div>
 
 
@@ -476,8 +482,8 @@ <h1 class="title"><span id="advanced-openrefine" class="quarto-section-identifie
 <div class="alert alert-info">
 <p>Right-click to <a href="https://cronkitedata.s3.amazonaws.com/csv/ppp_az_openrefine_input.csv">download the data</a> for this walkthrough and save it into your projects folder</p>
 </div>
-<section id="installing-openrefine" class="level2" data-number="28.1">
-<h2 data-number="28.1" class="anchored" data-anchor-id="installing-openrefine"><span class="header-section-number">28.1</span> Installing OpenRefine</h2>
+<section id="installing-openrefine" class="level2" data-number="29.1">
+<h2 data-number="29.1" class="anchored" data-anchor-id="installing-openrefine"><span class="header-section-number">29.1</span> Installing OpenRefine</h2>
 <p>Download the latest version from the <a href="https://openrefine.org/download.html">OpenRefine website</a> and follow the instructions there to install.</p>
 <section id="mac-users-unidentified-developer" class="level3 unnumbered">
 <h3 class="unnumbered anchored" data-anchor-id="mac-users-unidentified-developer">Mac users: Unidentified developer</h3>
@@ -496,8 +502,8 @@ <h3 class="unnumbered anchored" data-anchor-id="following-the-walkthrough">Follo
 <p>Steps that you should take are shown in bullet points. The rest is explanation.</p>
 </section>
 </section>
-<section id="starting-openrefine" class="level2" data-number="28.2">
-<h2 data-number="28.2" class="anchored" data-anchor-id="starting-openrefine"><span class="header-section-number">28.2</span> Starting OpenRefine</h2>
+<section id="starting-openrefine" class="level2" data-number="29.2">
+<h2 data-number="29.2" class="anchored" data-anchor-id="starting-openrefine"><span class="header-section-number">29.2</span> Starting OpenRefine</h2>
 <p>OpenRefine is a Java-based program that runs inside your browser, but lives on your computer instead of the internet. It runs much better in Chrome than in Safari.</p>
 <ul>
 <li><p>Find the OpenRefine app on your computer, and double-click. When you first start it up, it will open with your default browser.</p></li>
@@ -511,8 +517,8 @@ <h2 data-number="28.2" class="anchored" data-anchor-id="starting-openrefine"><sp
 </figure>
 </div>
 </section>
-<section id="importing-data" class="level2" data-number="28.3">
-<h2 data-number="28.3" class="anchored" data-anchor-id="importing-data"><span class="header-section-number">28.3</span> Importing data</h2>
+<section id="importing-data" class="level2" data-number="29.3">
+<h2 data-number="29.3" class="anchored" data-anchor-id="importing-data"><span class="header-section-number">29.3</span> Importing data</h2>
 <p>OpenRefine works best on comma- or tab-delimited data. To create a dataset like that from R, use the function <code>write_csv("path-to-csv-file.csv")</code> , which can be used like any other verb after a continuation pipe.</p>
 <ul>
 <li><p>To follow along with this tutorial, <a href="https://cronkitedata.s3.amazonaws.com/csv/ppp_az_openrefine_input.csv">download the sample data file</a> and save it into your R project folder .</p></li>
@@ -546,8 +552,8 @@ <h3 class="unnumbered anchored" data-anchor-id="rename-or-copy-columns">Rename o
 <p>To make a copy, use the dropdown menu <em>Edit column-&gt;Add column based on this column</em> and give it a name.</p>
 </section>
 </section>
-<section id="facets" class="level2" data-number="28.4">
-<h2 data-number="28.4" class="anchored" data-anchor-id="facets"><span class="header-section-number">28.4</span> Facets</h2>
+<section id="facets" class="level2" data-number="29.4">
+<h2 data-number="29.4" class="anchored" data-anchor-id="facets"><span class="header-section-number">29.4</span> Facets</h2>
 <p>For many text columns, you can use facets to edit your data. Under your newly named <code>project_state_refined</code>, choose <em>Facet-&gt;Text facet</em> to get a menu on the left. It will then show you the values it found in that column, and the number of times each one appears, much like at <code>group_by</code> query.</p>
 <p>OpenRefine only acts on rows that are selected, which makes using facets and filtering so powerful. You never have to worry that you’re changing something that you don’t see.</p>
 <div class="quarto-figure quarto-figure-center">
@@ -566,8 +572,8 @@ <h2 data-number="28.4" class="anchored" data-anchor-id="facets"><span class="hea
 <li><p>With the same selection on, change the county name to “OUT OF STATE” using the same method.</p></li>
 </ul>
 </section>
-<section id="clustering" class="level2" data-number="28.5">
-<h2 data-number="28.5" class="anchored" data-anchor-id="clustering"><span class="header-section-number">28.5</span> Clustering</h2>
+<section id="clustering" class="level2" data-number="29.5">
+<h2 data-number="29.5" class="anchored" data-anchor-id="clustering"><span class="header-section-number">29.5</span> Clustering</h2>
 <p>One reason a lot of reporters use OpenRefine is to take advantage of its fuzzy-matching to clean items like names and addresses. It provides about a dozen different algorithms that provide suggestions on ways to change unstandardized values in a column that often make a lot of sense.</p>
 <p>This walkthrough will work on the city name.</p>
 <div class="alert alert-warning">
@@ -628,8 +634,8 @@ <h3 class="unnumbered anchored" data-anchor-id="facet-for-long-entries">Facet fo
 <p>The list isn’t perfect, but it’s much better than what we started with.</p>
 </section>
 </section>
-<section id="filtering" class="level2" data-number="28.6">
-<h2 data-number="28.6" class="anchored" data-anchor-id="filtering"><span class="header-section-number">28.6</span> Filtering</h2>
+<section id="filtering" class="level2" data-number="29.6">
+<h2 data-number="29.6" class="anchored" data-anchor-id="filtering"><span class="header-section-number">29.6</span> Filtering</h2>
 <p>Even after fixing all of the cities we can through clustering and faceting, there are usually some bad entries left over.</p>
 <section id="simple-filter" class="level3">
 <h3 class="anchored" data-anchor-id="simple-filter">Simple filter</h3>
@@ -677,8 +683,8 @@ <h3 class="unnumbered anchored" data-anchor-id="split-using-openrefine-programmi
 </ul>
 </section>
 </section>
-<section id="finishing-up" class="level2" data-number="28.7">
-<h2 data-number="28.7" class="anchored" data-anchor-id="finishing-up"><span class="header-section-number">28.7</span> Finishing up</h2>
+<section id="finishing-up" class="level2" data-number="29.7">
+<h2 data-number="29.7" class="anchored" data-anchor-id="finishing-up"><span class="header-section-number">29.7</span> Finishing up</h2>
 <ul>
 <li>Try using the clustering and other transformation options to clean up the borrower names. Only change the ones you are sure should be combined.</li>
 </ul>
@@ -932,12 +938,12 @@ <h2 data-number="28.7" class="anchored" data-anchor-id="finishing-up"><span clas
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-pdf.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./advanced-scrape1.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/advanced-pdf.html b/docs/advanced-pdf.html
index 26d159d..249877a 100644
--- a/docs/advanced-pdf.html
+++ b/docs/advanced-pdf.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 27&nbsp; Working with PDF files</title>
+<title>Data Reporting - 28&nbsp; Working with PDF files</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -78,7 +78,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-pdf.html"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-pdf.html"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -235,35 +235,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -280,37 +286,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -327,25 +333,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -389,14 +395,14 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#the-pdf-problem" id="toc-the-pdf-problem" class="nav-link active" data-scroll-target="#the-pdf-problem"><span class="header-section-number">27.1</span> The PDF problem</a>
+  <li><a href="#the-pdf-problem" id="toc-the-pdf-problem" class="nav-link active" data-scroll-target="#the-pdf-problem"><span class="header-section-number">28.1</span> The PDF problem</a>
   <ul class="collapse">
   <li><a href="#understanding-your-pdf" id="toc-understanding-your-pdf" class="nav-link" data-scroll-target="#understanding-your-pdf">Understanding your pdf</a></li>
   <li><a href="#developing-a-strategy" id="toc-developing-a-strategy" class="nav-link" data-scroll-target="#developing-a-strategy">Developing a strategy</a></li>
   <li><a href="#example-document" id="toc-example-document" class="nav-link" data-scroll-target="#example-document">Example document</a></li>
   <li><a href="#work-from-the-goal-backwards" id="toc-work-from-the-goal-backwards" class="nav-link" data-scroll-target="#work-from-the-goal-backwards">Work from the goal backwards</a></li>
   </ul></li>
-  <li><a href="#pdf-conversion-tools" id="toc-pdf-conversion-tools" class="nav-link" data-scroll-target="#pdf-conversion-tools"><span class="header-section-number">27.2</span> PDF conversion tools</a>
+  <li><a href="#pdf-conversion-tools" id="toc-pdf-conversion-tools" class="nav-link" data-scroll-target="#pdf-conversion-tools"><span class="header-section-number">28.2</span> PDF conversion tools</a>
   <ul class="collapse">
   <li><a href="#tabula" id="toc-tabula" class="nav-link" data-scroll-target="#tabula">Tabula</a></li>
   <li><a href="#cometdocs" id="toc-cometdocs" class="nav-link" data-scroll-target="#cometdocs">Cometdocs</a></li>
@@ -405,8 +411,8 @@ <h2 id="toc-title">In this chapter:</h2>
   <li><a href="#able2extract" id="toc-able2extract" class="nav-link" data-scroll-target="#able2extract">Able2Extract ($)</a></li>
   <li><a href="#google-pinpoints-table-extractor" id="toc-google-pinpoints-table-extractor" class="nav-link" data-scroll-target="#google-pinpoints-table-extractor">Google Pinpoint’s Table Extractor</a></li>
   </ul></li>
-  <li><a href="#finishing-up" id="toc-finishing-up" class="nav-link" data-scroll-target="#finishing-up"><span class="header-section-number">27.3</span> Finishing up</a></li>
-  <li><a href="#use-a-program-instead" id="toc-use-a-program-instead" class="nav-link" data-scroll-target="#use-a-program-instead"><span class="header-section-number">27.4</span> Use a program instead</a></li>
+  <li><a href="#finishing-up" id="toc-finishing-up" class="nav-link" data-scroll-target="#finishing-up"><span class="header-section-number">28.3</span> Finishing up</a></li>
+  <li><a href="#use-a-program-instead" id="toc-use-a-program-instead" class="nav-link" data-scroll-target="#use-a-program-instead"><span class="header-section-number">28.4</span> Use a program instead</a></li>
   </ul>
 </nav>
     </div>
@@ -415,7 +421,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-pdf" class="quarto-section-identifier"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></h1>
+<h1 class="title"><span id="advanced-pdf" class="quarto-section-identifier"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></h1>
 </div>
 
 
@@ -431,8 +437,8 @@ <h1 class="title"><span id="advanced-pdf" class="quarto-section-identifier"><spa
 </header>
 
 <p><a href="https://cronkitedata.s3.amazonaws.com/slides/adv-pdf/adv-pdf/adv-pdf/adv-pdf-slides.html">Slide show from this chapter</a></p>
-<section id="the-pdf-problem" class="level2" data-number="27.1">
-<h2 data-number="27.1" class="anchored" data-anchor-id="the-pdf-problem"><span class="header-section-number">27.1</span> The PDF problem</h2>
+<section id="the-pdf-problem" class="level2" data-number="28.1">
+<h2 data-number="28.1" class="anchored" data-anchor-id="the-pdf-problem"><span class="header-section-number">28.1</span> The PDF problem</h2>
 <p>Many government agencies distribute data in PDF, or Adobe Portable Document Format. As the full name implies, these are documents, not data. It is geared at getting text and images properly placed on a page, not creating tidy data as we have been dealing with in previous chapters.</p>
 <p>Some software – Excel, Google Sheets, Microsoft Access and others – create PDF’s as one of their output formats. In fact, there is an entire ecosystem of pdf report makers, such as Crystal Reports. But converting them back into a usable dataset is often a challenge because we want columns and rows, not positions of words on a page.</p>
 <p>Some people call the process of returning a PDF to a data format as “scraping”. Others call it “report mining”. People with a lot of money can use automation tools, notable Altria’s Monarch product, to do this at scale with few errors, but it costs at least $2,000 / year / person for software like this, so few reporters have access.</p>
@@ -460,7 +466,7 @@ <h3 class="anchored" data-anchor-id="understanding-your-pdf">Understanding your
 </div>
 </div>
 </div>
-<p></p><figcaption class="figure-caption">Figure&nbsp;27.1: PDF types</figcaption><p></p>
+<p></p><figcaption class="figure-caption">Figure&nbsp;28.1: PDF types</figcaption><p></p>
 </figure>
 </div>
 <p>When you have an image, you will need a first step, or more expensive software, to deal with it. Most reporters use Google Pinpoint or DocumentCloud to do the “optical character recognition”, or OCR, necessary to work with them. A full version of Adobe Acrobat (not the reader) will also work. Some of the more expensive products listed below have versions that include OCR. But just know that if you have an image, your ability to parse these could be very difficult.</p>
@@ -496,8 +502,8 @@ <h3 class="anchored" data-anchor-id="work-from-the-goal-backwards">Work from the
 <p>In this case, we have to find a way to apply the name of the organization and its most recently updated date to all of the rows of the lobbyist names. The pattern is that there is no date on the lobbyist list, but there is always a date on the organization row. (Sometimes indentations and formatting can help, but not always.)</p>
 </section>
 </section>
-<section id="pdf-conversion-tools" class="level2" data-number="27.2">
-<h2 data-number="27.2" class="anchored" data-anchor-id="pdf-conversion-tools"><span class="header-section-number">27.2</span> PDF conversion tools</h2>
+<section id="pdf-conversion-tools" class="level2" data-number="28.2">
+<h2 data-number="28.2" class="anchored" data-anchor-id="pdf-conversion-tools"><span class="header-section-number">28.2</span> PDF conversion tools</h2>
 <p>You can’t just open a pdf or import it into an R program or Excel. Instead, you’ll need a tool to make the conversions. Here are some of the common ones used in newsrooms.</p>
 <section id="tabula" class="level3">
 <h3 class="anchored" data-anchor-id="tabula">Tabula</h3>
@@ -540,7 +546,7 @@ <h3 class="anchored" data-anchor-id="adobe-acrobat">Adobe Acrobat ($)</h3>
 </div>
 </div>
 </div>
-<p></p><figcaption class="figure-caption">Figure&nbsp;27.2: Using Acrobat</figcaption><p></p>
+<p></p><figcaption class="figure-caption">Figure&nbsp;28.2: Using Acrobat</figcaption><p></p>
 </figure>
 </div>
 <p>It tried to replicate the way the data looks in the page, which smushed together the individual lobbyists even though they’re clearly on separate lines.</p>
@@ -557,8 +563,8 @@ <h3 class="anchored" data-anchor-id="google-pinpoints-table-extractor">Google Pi
 <p><img src="assets/images/advanced-pdf-azdpsform.png" class="img-fluid" style="width:75.0%"></p>
 </section>
 </section>
-<section id="finishing-up" class="level2" data-number="27.3">
-<h2 data-number="27.3" class="anchored" data-anchor-id="finishing-up"><span class="header-section-number">27.3</span> Finishing up</h2>
+<section id="finishing-up" class="level2" data-number="28.3">
+<h2 data-number="28.3" class="anchored" data-anchor-id="finishing-up"><span class="header-section-number">28.3</span> Finishing up</h2>
 <p>Using the csv file produced by Able2Extract, here’s one method for finishing up, using formulas in Google Sheets to figure out how to put together the pages. Here are the steps I took:</p>
 <ol type="1">
 <li>Remove all empty rows, so that copying an pasting is easier.</li>
@@ -572,8 +578,8 @@ <h2 data-number="27.3" class="anchored" data-anchor-id="finishing-up"><span clas
 <div class="quarto-video"><video id="video_shortcode_videojs_video2" class="video-js vjs-default-skin vjs-fluid" controls="" preload="auto" data-setup="{}" title=""><source src="https://cronkitedata.s3.amazonaws.com/images/pdf-lobbyist-fix.mp4"></video></div>
 <p><a href="https://docs.google.com/spreadsheets/d/1XyxpeIKcmcU3wl6if61HP_bcQUywWkQDHYId7hKi4FE/edit?usp=sharing">This Google Sheet</a> shows how some of the data was transferred from the converters, and the final spreadsheet that is in “tidy” format.</p>
 </section>
-<section id="use-a-program-instead" class="level2" data-number="27.4">
-<h2 data-number="27.4" class="anchored" data-anchor-id="use-a-program-instead"><span class="header-section-number">27.4</span> Use a program instead</h2>
+<section id="use-a-program-instead" class="level2" data-number="28.4">
+<h2 data-number="28.4" class="anchored" data-anchor-id="use-a-program-instead"><span class="header-section-number">28.4</span> Use a program instead</h2>
 <p>We won’t go through all of the methods of using a program to deconstruct PDFs, but you’ll sometimes find that it’s easier to program it than to go through all of the manual steps. <a href="https://cronkitedata.s3.amazonaws.com/rdata/parse_lobbying_report.html">This Quarto document shows how you might process our example</a> . There are a number of functions you’ve never seen in here before, since we’re working with vectors and lists to begin with rather than data frames.</p>
 
 
@@ -828,7 +834,7 @@ <h2 data-number="27.4" class="anchored" data-anchor-id="use-a-program-instead"><
   </div>
   <div class="nav-page nav-page-next">
       <a href="./advanced-openrefine.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/advanced-regex-r.html b/docs/advanced-regex-r.html
index 10fad71..cb0b6cf 100644
--- a/docs/advanced-regex-r.html
+++ b/docs/advanced-regex-r.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 19&nbsp; Inexact matching and regular expressions in R</title>
+<title>Data Reporting - 20&nbsp; Inexact matching and regular expressions in R</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -120,7 +120,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./advanced-regex-r.html"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./advanced-regex-r.html"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -277,35 +277,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -322,37 +328,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -369,25 +375,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -431,18 +437,18 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#resouces-for-using-regular-expressions-in-r" id="toc-resouces-for-using-regular-expressions-in-r" class="nav-link active" data-scroll-target="#resouces-for-using-regular-expressions-in-r"><span class="header-section-number">19.1</span> Resouces for using regular expressions in R</a></li>
-  <li><a href="#sample-data" id="toc-sample-data" class="nav-link" data-scroll-target="#sample-data"><span class="header-section-number">19.2</span> Sample data</a>
+  <li><a href="#resouces-for-using-regular-expressions-in-r" id="toc-resouces-for-using-regular-expressions-in-r" class="nav-link active" data-scroll-target="#resouces-for-using-regular-expressions-in-r"><span class="header-section-number">20.1</span> Resouces for using regular expressions in R</a></li>
+  <li><a href="#sample-data" id="toc-sample-data" class="nav-link" data-scroll-target="#sample-data"><span class="header-section-number">20.2</span> Sample data</a>
   <ul class="collapse">
   <li><a href="#string-functions-for-regular-expressions" id="toc-string-functions-for-regular-expressions" class="nav-link" data-scroll-target="#string-functions-for-regular-expressions">String functions for regular expressions</a></li>
   <li><a href="#convert-everything-to-lowercase" id="toc-convert-everything-to-lowercase" class="nav-link" data-scroll-target="#convert-everything-to-lowercase">Convert everything to lowercase</a></li>
   </ul></li>
-  <li><a href="#using-str_detect" id="toc-using-str_detect" class="nav-link" data-scroll-target="#using-str_detect"><span class="header-section-number">19.3</span> Using <code>str_detect</code></a>
+  <li><a href="#using-str_detect" id="toc-using-str_detect" class="nav-link" data-scroll-target="#using-str_detect"><span class="header-section-number">20.3</span> Using <code>str_detect</code></a>
   <ul class="collapse">
   <li><a href="#example-1-filter-for-biden" id="toc-example-1-filter-for-biden" class="nav-link" data-scroll-target="#example-1-filter-for-biden">Example 1: Filter for “biden”</a></li>
   <li><a href="#keywords" id="toc-keywords" class="nav-link" data-scroll-target="#keywords">Keywords</a></li>
   </ul></li>
-  <li><a href="#types-of-regular-expression-patterns" id="toc-types-of-regular-expression-patterns" class="nav-link" data-scroll-target="#types-of-regular-expression-patterns"><span class="header-section-number">19.4</span> Types of regular expression patterns</a>
+  <li><a href="#types-of-regular-expression-patterns" id="toc-types-of-regular-expression-patterns" class="nav-link" data-scroll-target="#types-of-regular-expression-patterns"><span class="header-section-number">20.4</span> Types of regular expression patterns</a>
   <ul class="collapse">
   <li><a href="#literal-strings" id="toc-literal-strings" class="nav-link" data-scroll-target="#literal-strings">Literal strings</a></li>
   <li><a href="#wild-cards" id="toc-wild-cards" class="nav-link" data-scroll-target="#wild-cards">Wild cards</a></li>
@@ -450,15 +456,15 @@ <h2 id="toc-title">In this chapter:</h2>
   <li><a href="#escaping" id="toc-escaping" class="nav-link" data-scroll-target="#escaping">Escaping</a></li>
   <li><a href="#match-groups" id="toc-match-groups" class="nav-link" data-scroll-target="#match-groups">Match groups</a></li>
   </ul></li>
-  <li><a href="#sample-data-1" id="toc-sample-data-1" class="nav-link" data-scroll-target="#sample-data-1"><span class="header-section-number">19.5</span> Sample data</a></li>
-  <li><a href="#testing-and-examining-patterns-with-regex101" id="toc-testing-and-examining-patterns-with-regex101" class="nav-link" data-scroll-target="#testing-and-examining-patterns-with-regex101"><span class="header-section-number">19.6</span> Testing and examining patterns with Regex101</a>
+  <li><a href="#sample-data-1" id="toc-sample-data-1" class="nav-link" data-scroll-target="#sample-data-1"><span class="header-section-number">20.5</span> Sample data</a></li>
+  <li><a href="#testing-and-examining-patterns-with-regex101" id="toc-testing-and-examining-patterns-with-regex101" class="nav-link" data-scroll-target="#testing-and-examining-patterns-with-regex101"><span class="header-section-number">20.6</span> Testing and examining patterns with Regex101</a>
   <ul class="collapse">
   <li><a href="#looking-for-specific-words-or-characters" id="toc-looking-for-specific-words-or-characters" class="nav-link" data-scroll-target="#looking-for-specific-words-or-characters">Looking for specific words or characters</a></li>
   </ul></li>
-  <li><a href="#practice-1-extract-date-parts" id="toc-practice-1-extract-date-parts" class="nav-link" data-scroll-target="#practice-1-extract-date-parts"><span class="header-section-number">19.7</span> Practice #1: Extract date parts</a></li>
-  <li><a href="#practice-2-extract-pieces-of-phone-numbers" id="toc-practice-2-extract-pieces-of-phone-numbers" class="nav-link" data-scroll-target="#practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">19.8</span> Practice #2: Extract pieces of phone numbers</a></li>
-  <li><a href="#practice-3-extract-address-pieces" id="toc-practice-3-extract-address-pieces" class="nav-link" data-scroll-target="#practice-3-extract-address-pieces"><span class="header-section-number">19.9</span> Practice #3: Extract address pieces</a></li>
-  <li><a href="#on-your-own" id="toc-on-your-own" class="nav-link" data-scroll-target="#on-your-own"><span class="header-section-number">19.10</span> On your own</a></li>
+  <li><a href="#practice-1-extract-date-parts" id="toc-practice-1-extract-date-parts" class="nav-link" data-scroll-target="#practice-1-extract-date-parts"><span class="header-section-number">20.7</span> Practice #1: Extract date parts</a></li>
+  <li><a href="#practice-2-extract-pieces-of-phone-numbers" id="toc-practice-2-extract-pieces-of-phone-numbers" class="nav-link" data-scroll-target="#practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">20.8</span> Practice #2: Extract pieces of phone numbers</a></li>
+  <li><a href="#practice-3-extract-address-pieces" id="toc-practice-3-extract-address-pieces" class="nav-link" data-scroll-target="#practice-3-extract-address-pieces"><span class="header-section-number">20.9</span> Practice #3: Extract address pieces</a></li>
+  <li><a href="#on-your-own" id="toc-on-your-own" class="nav-link" data-scroll-target="#on-your-own"><span class="header-section-number">20.10</span> On your own</a></li>
   </ul>
 </nav>
     </div>
@@ -467,7 +473,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-regex-r" class="quarto-section-identifier"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></h1>
+<h1 class="title"><span id="advanced-regex-r" class="quarto-section-identifier"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></h1>
 </div>
 
 
@@ -519,8 +525,8 @@ <h1 class="title"><span id="advanced-regex-r" class="quarto-section-identifier">
 <p>Now that you’ve looked at the general idea of regular expressions, this chapter is a brief overview of how they are implemented in the tidyverse.</p>
 <p>The key functions of inexact matching that you’ll use all take the same general form. Instead of a direct operator, such as <code>mycolumn &lt; 100</code> , the are implemented in functions.</p>
 <p>We’ve already looked at the function <code>str_like()</code>, which allows for upper or lower case with wild cards. This chapter, however, use the regular expressions that were described in the last section.</p>
-<section id="resouces-for-using-regular-expressions-in-r" class="level2" data-number="19.1">
-<h2 data-number="19.1" class="anchored" data-anchor-id="resouces-for-using-regular-expressions-in-r"><span class="header-section-number">19.1</span> Resouces for using regular expressions in R</h2>
+<section id="resouces-for-using-regular-expressions-in-r" class="level2" data-number="20.1">
+<h2 data-number="20.1" class="anchored" data-anchor-id="resouces-for-using-regular-expressions-in-r"><span class="header-section-number">20.1</span> Resouces for using regular expressions in R</h2>
 <ul>
 <li><p>The <a href="https://github.com/rstudio/cheatsheets/blob/main/strings.pdf">tidyverse cheat sheet</a> on the package <code>stringr</code> (which is party of tidyverse) has the outlines of each function that uses these regular expression. It’s a little hard to read, but has all of the shortcuts on the second page. Look at the part that says, “type this”, which translates to “to mean this” as in Regex101.</p></li>
 <li><p>Nathan Collins, a student in the 2022 MAIJ cohort, [did an explainer on using <code>str_detect()</code> on the ACLED])(https://npcdata.github.io/projects/collins_explainer.html) data that many of you are using for your projects</p></li>
@@ -543,8 +549,8 @@ <h2 data-number="19.1" class="anchored" data-anchor-id="resouces-for-using-regul
 </div>
 </div>
 </section>
-<section id="sample-data" class="level2 page-columns page-full" data-number="19.2">
-<h2 data-number="19.2" class="anchored" data-anchor-id="sample-data"><span class="header-section-number">19.2</span> Sample data</h2>
+<section id="sample-data" class="level2 page-columns page-full" data-number="20.2">
+<h2 data-number="20.2" class="anchored" data-anchor-id="sample-data"><span class="header-section-number">20.2</span> Sample data</h2>
 <p>For this example, we’ll use a small set of Tweets from Donald Trump, representing the last few months he was on on the platform. It runs from Nov.&nbsp;1, 2020 to Jan.&nbsp;8, 2021, when Twitter banned him. Retweets were removed, but there are a number of tweets that are just links to other content.</p>
 <p>Here is a random sample of the 1,112 tweets:</p>
 <div class="cell page-columns page-full">
@@ -579,8 +585,8 @@ <h3 class="anchored" data-anchor-id="convert-everything-to-lowercase">Convert ev
 </div>
 </section>
 </section>
-<section id="using-str_detect" class="level2 page-columns page-full" data-number="19.3">
-<h2 data-number="19.3" class="anchored" data-anchor-id="using-str_detect"><span class="header-section-number">19.3</span> Using <code>str_detect</code></h2>
+<section id="using-str_detect" class="level2 page-columns page-full" data-number="20.3">
+<h2 data-number="20.3" class="anchored" data-anchor-id="using-str_detect"><span class="header-section-number">20.3</span> Using <code>str_detect</code></h2>
 <p>The function <code>str_detect()</code> results in either <code>TRUE</code> or <code>FALSE</code>. It’s very useful in two situations:</p>
 <ul>
 <li>As a filter for rows in your data.</li>
@@ -653,8 +659,8 @@ <h3 class="anchored" data-anchor-id="keywords">Keywords</h3>
 </ul>
 </section>
 </section>
-<section id="types-of-regular-expression-patterns" class="level2" data-number="19.4">
-<h2 data-number="19.4" class="anchored" data-anchor-id="types-of-regular-expression-patterns"><span class="header-section-number">19.4</span> Types of regular expression patterns</h2>
+<section id="types-of-regular-expression-patterns" class="level2" data-number="20.4">
+<h2 data-number="20.4" class="anchored" data-anchor-id="types-of-regular-expression-patterns"><span class="header-section-number">20.4</span> Types of regular expression patterns</h2>
 <section id="literal-strings" class="level3">
 <h3 class="anchored" data-anchor-id="literal-strings">Literal strings</h3>
 <p>These are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.</p>
@@ -700,8 +706,8 @@ <h3 class="anchored" data-anchor-id="match-groups">Match groups</h3>
 <p>Use parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with <code>lists</code>.</p>
 </section>
 </section>
-<section id="sample-data-1" class="level2" data-number="19.5">
-<h2 data-number="19.5" class="anchored" data-anchor-id="sample-data-1"><span class="header-section-number">19.5</span> Sample data</h2>
+<section id="sample-data-1" class="level2" data-number="20.5">
+<h2 data-number="20.5" class="anchored" data-anchor-id="sample-data-1"><span class="header-section-number">20.5</span> Sample data</h2>
 <p>Here are three small text files that you can copy and paste from your browser into the <a href="https://regex101.com/">regex101.com</a> site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.</p>
 <ol type="1">
 <li>A <a href="https://cronkitedata.s3.amazonaws.com/csv/regex_phones.txt">list of phone numbers</a> in different formats</li>
@@ -710,8 +716,8 @@ <h2 data-number="19.5" class="anchored" data-anchor-id="sample-data-1"><span cla
 <li>A <a href="https://cronkitedata.s3.amazonaws.com/csv/regex_h2bvisas.txt">small chunk of the H2B visa applications</a> from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format.</li>
 </ol>
 </section>
-<section id="testing-and-examining-patterns-with-regex101" class="level2" data-number="19.6">
-<h2 data-number="19.6" class="anchored" data-anchor-id="testing-and-examining-patterns-with-regex101"><span class="header-section-number">19.6</span> Testing and examining patterns with Regex101</h2>
+<section id="testing-and-examining-patterns-with-regex101" class="level2" data-number="20.6">
+<h2 data-number="20.6" class="anchored" data-anchor-id="testing-and-examining-patterns-with-regex101"><span class="header-section-number">20.6</span> Testing and examining patterns with Regex101</h2>
 <p><a href="https://regex101.com">Regex 101</a> is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\` in R.</p>
 <section id="looking-for-specific-words-or-characters" class="level3">
 <h3 class="anchored" data-anchor-id="looking-for-specific-words-or-characters">Looking for specific words or characters</h3>
@@ -719,8 +725,8 @@ <h3 class="anchored" data-anchor-id="looking-for-specific-words-or-characters">L
 <p><img src="assets/images/advanced-regex1.png" class="img-fluid" style="width:100.0%"></p>
 </section>
 </section>
-<section id="practice-1-extract-date-parts" class="level2" data-number="19.7">
-<h2 data-number="19.7" class="anchored" data-anchor-id="practice-1-extract-date-parts"><span class="header-section-number">19.7</span> Practice #1: Extract date parts</h2>
+<section id="practice-1-extract-date-parts" class="level2" data-number="20.7">
+<h2 data-number="20.7" class="anchored" data-anchor-id="practice-1-extract-date-parts"><span class="header-section-number">20.7</span> Practice #1: Extract date parts</h2>
 <p>In Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.<a href="#fn2" class="footnote-ref" id="fnref2" role="doc-noteref"><sup>2</sup></a></p>
 <p>We want to turn dates that look like this:</p>
 <pre><code>  1/24/2018</code></pre>
@@ -748,8 +754,8 @@ <h2 data-number="19.7" class="anchored" data-anchor-id="practice-1-extract-date-
 <p>Now each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces.</p>
 <p><img src="assets/images/advanced-regex3.png" class="img-fluid" style="width:100.0%"></p>
 </section>
-<section id="practice-2-extract-pieces-of-phone-numbers" class="level2" data-number="19.8">
-<h2 data-number="19.8" class="anchored" data-anchor-id="practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">19.8</span> Practice #2: Extract pieces of phone numbers</h2>
+<section id="practice-2-extract-pieces-of-phone-numbers" class="level2" data-number="20.8">
+<h2 data-number="20.8" class="anchored" data-anchor-id="practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">20.8</span> Practice #2: Extract pieces of phone numbers</h2>
 <p>Here are some phone numbers in different formats:</p>
 <div class="sourceCode" id="cb18"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb18-1"><a href="#cb18-1" aria-hidden="true" tabindex="-1"></a><span class="in">    623-374-1167</span></span>
 <span id="cb18-2"><a href="#cb18-2" aria-hidden="true" tabindex="-1"></a><span class="in">    760.352.5212</span></span>
@@ -765,8 +771,8 @@ <h2 data-number="19.8" class="anchored" data-anchor-id="practice-2-extract-piece
 <p>(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)</p>
 <p>Anything within parentheses will be “captured” in a block.</p>
 </section>
-<section id="practice-3-extract-address-pieces" class="level2" data-number="19.9">
-<h2 data-number="19.9" class="anchored" data-anchor-id="practice-3-extract-address-pieces"><span class="header-section-number">19.9</span> Practice #3: Extract address pieces</h2>
+<section id="practice-3-extract-address-pieces" class="level2" data-number="20.9">
+<h2 data-number="20.9" class="anchored" data-anchor-id="practice-3-extract-address-pieces"><span class="header-section-number">20.9</span> Practice #3: Extract address pieces</h2>
 <p>Here are a few lines of the data from <a href="https://cronkitedata.s3.amazonaws.com/docs/regex_for_rest_of_us.pdf">Prof.&nbsp;McDonald’s tutorial</a>, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)</p>
 <div class="sourceCode" id="cb20"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb20-1"><a href="#cb20-1" aria-hidden="true" tabindex="-1"></a><span class="in">    "10111 N LAMAR BLVD</span></span>
 <span id="cb20-2"><a href="#cb20-2" aria-hidden="true" tabindex="-1"></a><span class="in">    AUSTIN, TX 78753</span></span>
@@ -814,8 +820,8 @@ <h2 data-number="19.9" class="anchored" data-anchor-id="practice-3-extract-addre
 <span id="cb20-44"><a href="#cb20-44" aria-hidden="true" tabindex="-1"></a><span class="in">    ROLLINGWOOD, TX 78746</span></span>
 <span id="cb20-45"><a href="#cb20-45" aria-hidden="true" tabindex="-1"></a><span class="in">    (30.271592738000436, -97.79583786499967)"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
-<section id="on-your-own" class="level2" data-number="19.10">
-<h2 data-number="19.10" class="anchored" data-anchor-id="on-your-own"><span class="header-section-number">19.10</span> On your own</h2>
+<section id="on-your-own" class="level2" data-number="20.10">
+<h2 data-number="20.10" class="anchored" data-anchor-id="on-your-own"><span class="header-section-number">20.10</span> On your own</h2>
 <p><a href="https://cronkitedata.s3.amazonaws.com/csv/regex_h2bvisas.txt">This is a small list of H2A visa applications</a>, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.</p>
 <p>You might try:</p>
 <ul>
@@ -1072,12 +1078,12 @@ <h2 data-number="19.10" class="anchored" data-anchor-id="on-your-own"><span clas
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-regex.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./r-recipes.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/advanced-regex.html b/docs/advanced-regex.html
index 2a96138..716415e 100644
--- a/docs/advanced-regex.html
+++ b/docs/advanced-regex.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 18&nbsp; Regular Expressions for pattern matching</title>
+<title>Data Reporting - 19&nbsp; Regular Expressions for pattern matching</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -113,7 +113,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./advanced-regex.html"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./advanced-regex.html"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -270,35 +270,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -315,37 +321,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -362,25 +368,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -424,7 +430,7 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#types-of-regular-expression-patterns" id="toc-types-of-regular-expression-patterns" class="nav-link active" data-scroll-target="#types-of-regular-expression-patterns"><span class="header-section-number">18.1</span> Types of regular expression patterns</a>
+  <li><a href="#types-of-regular-expression-patterns" id="toc-types-of-regular-expression-patterns" class="nav-link active" data-scroll-target="#types-of-regular-expression-patterns"><span class="header-section-number">19.1</span> Types of regular expression patterns</a>
   <ul class="collapse">
   <li><a href="#literal-strings" id="toc-literal-strings" class="nav-link" data-scroll-target="#literal-strings">Literal strings</a></li>
   <li><a href="#wild-cards" id="toc-wild-cards" class="nav-link" data-scroll-target="#wild-cards">Wild cards</a></li>
@@ -432,15 +438,15 @@ <h2 id="toc-title">In this chapter:</h2>
   <li><a href="#escaping" id="toc-escaping" class="nav-link" data-scroll-target="#escaping">Escaping</a></li>
   <li><a href="#match-groups" id="toc-match-groups" class="nav-link" data-scroll-target="#match-groups">Match groups</a></li>
   </ul></li>
-  <li><a href="#sample-data" id="toc-sample-data" class="nav-link" data-scroll-target="#sample-data"><span class="header-section-number">18.2</span> Sample data</a></li>
-  <li><a href="#testing-and-examining-patterns-with-regex101" id="toc-testing-and-examining-patterns-with-regex101" class="nav-link" data-scroll-target="#testing-and-examining-patterns-with-regex101"><span class="header-section-number">18.3</span> Testing and examining patterns with Regex101</a>
+  <li><a href="#sample-data" id="toc-sample-data" class="nav-link" data-scroll-target="#sample-data"><span class="header-section-number">19.2</span> Sample data</a></li>
+  <li><a href="#testing-and-examining-patterns-with-regex101" id="toc-testing-and-examining-patterns-with-regex101" class="nav-link" data-scroll-target="#testing-and-examining-patterns-with-regex101"><span class="header-section-number">19.3</span> Testing and examining patterns with Regex101</a>
   <ul class="collapse">
   <li><a href="#looking-for-specific-words-or-characters" id="toc-looking-for-specific-words-or-characters" class="nav-link" data-scroll-target="#looking-for-specific-words-or-characters">Looking for specific words or characters</a></li>
   </ul></li>
-  <li><a href="#practice-1-extract-date-parts" id="toc-practice-1-extract-date-parts" class="nav-link" data-scroll-target="#practice-1-extract-date-parts"><span class="header-section-number">18.4</span> Practice #1: Extract date parts</a></li>
-  <li><a href="#practice-2-extract-pieces-of-phone-numbers" id="toc-practice-2-extract-pieces-of-phone-numbers" class="nav-link" data-scroll-target="#practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">18.5</span> Practice #2: Extract pieces of phone numbers</a></li>
-  <li><a href="#practice-3-extract-address-pieces" id="toc-practice-3-extract-address-pieces" class="nav-link" data-scroll-target="#practice-3-extract-address-pieces"><span class="header-section-number">18.6</span> Practice #3: Extract address pieces</a></li>
-  <li><a href="#on-your-own" id="toc-on-your-own" class="nav-link" data-scroll-target="#on-your-own"><span class="header-section-number">18.7</span> On your own</a></li>
+  <li><a href="#practice-1-extract-date-parts" id="toc-practice-1-extract-date-parts" class="nav-link" data-scroll-target="#practice-1-extract-date-parts"><span class="header-section-number">19.4</span> Practice #1: Extract date parts</a></li>
+  <li><a href="#practice-2-extract-pieces-of-phone-numbers" id="toc-practice-2-extract-pieces-of-phone-numbers" class="nav-link" data-scroll-target="#practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">19.5</span> Practice #2: Extract pieces of phone numbers</a></li>
+  <li><a href="#practice-3-extract-address-pieces" id="toc-practice-3-extract-address-pieces" class="nav-link" data-scroll-target="#practice-3-extract-address-pieces"><span class="header-section-number">19.6</span> Practice #3: Extract address pieces</a></li>
+  <li><a href="#on-your-own" id="toc-on-your-own" class="nav-link" data-scroll-target="#on-your-own"><span class="header-section-number">19.7</span> On your own</a></li>
   </ul>
 </nav>
     </div>
@@ -449,7 +455,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-regex" class="quarto-section-identifier"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></h1>
+<h1 class="title"><span id="advanced-regex" class="quarto-section-identifier"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></h1>
 </div>
 
 
@@ -484,8 +490,8 @@ <h1 class="title"><span id="advanced-regex" class="quarto-section-identifier"><s
 <ul>
 <li>The <a href="https://stringr.tidyverse.org/">vignette on the <code>stringr</code> package</a> in the Tidyverse is reasonably helpful.</li>
 </ul>
-<section id="types-of-regular-expression-patterns" class="level2" data-number="18.1">
-<h2 data-number="18.1" class="anchored" data-anchor-id="types-of-regular-expression-patterns"><span class="header-section-number">18.1</span> Types of regular expression patterns</h2>
+<section id="types-of-regular-expression-patterns" class="level2" data-number="19.1">
+<h2 data-number="19.1" class="anchored" data-anchor-id="types-of-regular-expression-patterns"><span class="header-section-number">19.1</span> Types of regular expression patterns</h2>
 <section id="literal-strings" class="level3">
 <h3 class="anchored" data-anchor-id="literal-strings">Literal strings</h3>
 <p>These are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.</p>
@@ -531,8 +537,8 @@ <h3 class="anchored" data-anchor-id="match-groups">Match groups</h3>
 <p>Use parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with <code>lists</code>.</p>
 </section>
 </section>
-<section id="sample-data" class="level2" data-number="18.2">
-<h2 data-number="18.2" class="anchored" data-anchor-id="sample-data"><span class="header-section-number">18.2</span> Sample data</h2>
+<section id="sample-data" class="level2" data-number="19.2">
+<h2 data-number="19.2" class="anchored" data-anchor-id="sample-data"><span class="header-section-number">19.2</span> Sample data</h2>
 <p>Here are three small text files that you can copy and paste from your browser into the <a href="https://regex101.com/">regex101.com</a> site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.</p>
 <ol type="1">
 <li>A <a href="https://cronkitedata.s3.amazonaws.com/csv/regex_phones.txt">list of phone numbers</a> in different formats</li>
@@ -541,8 +547,8 @@ <h2 data-number="18.2" class="anchored" data-anchor-id="sample-data"><span class
 <li>A <a href="https://cronkitedata.s3.amazonaws.com/csv/regex_h2bvisas.txt">small chunk of the H2B visa applications</a> from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format.</li>
 </ol>
 </section>
-<section id="testing-and-examining-patterns-with-regex101" class="level2" data-number="18.3">
-<h2 data-number="18.3" class="anchored" data-anchor-id="testing-and-examining-patterns-with-regex101"><span class="header-section-number">18.3</span> Testing and examining patterns with Regex101</h2>
+<section id="testing-and-examining-patterns-with-regex101" class="level2" data-number="19.3">
+<h2 data-number="19.3" class="anchored" data-anchor-id="testing-and-examining-patterns-with-regex101"><span class="header-section-number">19.3</span> Testing and examining patterns with Regex101</h2>
 <p><a href="https://regex101.com">Regex 101</a> is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\` in R.</p>
 <section id="looking-for-specific-words-or-characters" class="level3">
 <h3 class="anchored" data-anchor-id="looking-for-specific-words-or-characters">Looking for specific words or characters</h3>
@@ -550,8 +556,8 @@ <h3 class="anchored" data-anchor-id="looking-for-specific-words-or-characters">L
 <p><img src="assets/images/advanced-regex1.png" class="img-fluid" style="width:100.0%"></p>
 </section>
 </section>
-<section id="practice-1-extract-date-parts" class="level2" data-number="18.4">
-<h2 data-number="18.4" class="anchored" data-anchor-id="practice-1-extract-date-parts"><span class="header-section-number">18.4</span> Practice #1: Extract date parts</h2>
+<section id="practice-1-extract-date-parts" class="level2" data-number="19.4">
+<h2 data-number="19.4" class="anchored" data-anchor-id="practice-1-extract-date-parts"><span class="header-section-number">19.4</span> Practice #1: Extract date parts</h2>
 <p>In Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.<a href="#fn1" class="footnote-ref" id="fnref1" role="doc-noteref"><sup>1</sup></a></p>
 <p>We want to turn dates that look like this:</p>
 <pre><code>  1/24/2018</code></pre>
@@ -579,8 +585,8 @@ <h2 data-number="18.4" class="anchored" data-anchor-id="practice-1-extract-date-
 <p>Now each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces.</p>
 <p><img src="assets/images/advanced-regex3.png" class="img-fluid" style="width:100.0%"></p>
 </section>
-<section id="practice-2-extract-pieces-of-phone-numbers" class="level2" data-number="18.5">
-<h2 data-number="18.5" class="anchored" data-anchor-id="practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">18.5</span> Practice #2: Extract pieces of phone numbers</h2>
+<section id="practice-2-extract-pieces-of-phone-numbers" class="level2" data-number="19.5">
+<h2 data-number="19.5" class="anchored" data-anchor-id="practice-2-extract-pieces-of-phone-numbers"><span class="header-section-number">19.5</span> Practice #2: Extract pieces of phone numbers</h2>
 <p>Here are some phone numbers in different formats:</p>
 <div class="sourceCode" id="cb14"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb14-1"><a href="#cb14-1" aria-hidden="true" tabindex="-1"></a><span class="in">    623-374-1167</span></span>
 <span id="cb14-2"><a href="#cb14-2" aria-hidden="true" tabindex="-1"></a><span class="in">    760.352.5212</span></span>
@@ -596,8 +602,8 @@ <h2 data-number="18.5" class="anchored" data-anchor-id="practice-2-extract-piece
 <p>(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)</p>
 <p>Anything within parentheses will be “captured” in a block.</p>
 </section>
-<section id="practice-3-extract-address-pieces" class="level2" data-number="18.6">
-<h2 data-number="18.6" class="anchored" data-anchor-id="practice-3-extract-address-pieces"><span class="header-section-number">18.6</span> Practice #3: Extract address pieces</h2>
+<section id="practice-3-extract-address-pieces" class="level2" data-number="19.6">
+<h2 data-number="19.6" class="anchored" data-anchor-id="practice-3-extract-address-pieces"><span class="header-section-number">19.6</span> Practice #3: Extract address pieces</h2>
 <p>Here are a few lines of the data from <a href="https://cronkitedata.s3.amazonaws.com/docs/regex_for_rest_of_us.pdf">Prof.&nbsp;McDonald’s tutorial</a>, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)</p>
 <div class="sourceCode" id="cb16"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb16-1"><a href="#cb16-1" aria-hidden="true" tabindex="-1"></a><span class="in">    "10111 N LAMAR BLVD</span></span>
 <span id="cb16-2"><a href="#cb16-2" aria-hidden="true" tabindex="-1"></a><span class="in">    AUSTIN, TX 78753</span></span>
@@ -645,8 +651,8 @@ <h2 data-number="18.6" class="anchored" data-anchor-id="practice-3-extract-addre
 <span id="cb16-44"><a href="#cb16-44" aria-hidden="true" tabindex="-1"></a><span class="in">    ROLLINGWOOD, TX 78746</span></span>
 <span id="cb16-45"><a href="#cb16-45" aria-hidden="true" tabindex="-1"></a><span class="in">    (30.271592738000436, -97.79583786499967)"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
-<section id="on-your-own" class="level2" data-number="18.7">
-<h2 data-number="18.7" class="anchored" data-anchor-id="on-your-own"><span class="header-section-number">18.7</span> On your own</h2>
+<section id="on-your-own" class="level2" data-number="19.7">
+<h2 data-number="19.7" class="anchored" data-anchor-id="on-your-own"><span class="header-section-number">19.7</span> On your own</h2>
 <p><a href="https://cronkitedata.s3.amazonaws.com/csv/regex_h2bvisas.txt">This is a small list of H2A visa applications</a>, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.</p>
 <p>You might try:</p>
 <ul>
@@ -902,12 +908,12 @@ <h2 data-number="18.7" class="anchored" data-anchor-id="on-your-own"><span class
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./r-verb-join.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./advanced-regex-r.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/advanced-scrape1.html b/docs/advanced-scrape1.html
index b66ed08..59ea10b 100644
--- a/docs/advanced-scrape1.html
+++ b/docs/advanced-scrape1.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 29&nbsp; Scraping without programming</title>
+<title>Data Reporting - 30&nbsp; Scraping without programming</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -110,7 +110,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-scrape1.html"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-scrape1.html"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -267,35 +267,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -312,37 +318,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -359,25 +365,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -421,14 +427,14 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#where-reporters-get-data" id="toc-where-reporters-get-data" class="nav-link active" data-scroll-target="#where-reporters-get-data"><span class="header-section-number">29.1</span> Where reporters get data</a></li>
-  <li><a href="#a-json-miracle-walkthrough" id="toc-a-json-miracle-walkthrough" class="nav-link" data-scroll-target="#a-json-miracle-walkthrough"><span class="header-section-number">29.2</span> A json miracle walkthrough</a>
+  <li><a href="#where-reporters-get-data" id="toc-where-reporters-get-data" class="nav-link active" data-scroll-target="#where-reporters-get-data"><span class="header-section-number">30.1</span> Where reporters get data</a></li>
+  <li><a href="#a-json-miracle-walkthrough" id="toc-a-json-miracle-walkthrough" class="nav-link" data-scroll-target="#a-json-miracle-walkthrough"><span class="header-section-number">30.2</span> A json miracle walkthrough</a>
   <ul class="collapse">
   <li><a href="#a-harder-example" id="toc-a-harder-example" class="nav-link" data-scroll-target="#a-harder-example">A harder example</a></li>
   <li><a href="#an-even-harder-example" id="toc-an-even-harder-example" class="nav-link" data-scroll-target="#an-even-harder-example">An even harder example</a></li>
   </ul></li>
-  <li><a href="#no-json-no-problem-maybe" id="toc-no-json-no-problem-maybe" class="nav-link" data-scroll-target="#no-json-no-problem-maybe"><span class="header-section-number">29.3</span> No json? No problem (maybe)</a></li>
-  <li><a href="#recap" id="toc-recap" class="nav-link" data-scroll-target="#recap"><span class="header-section-number">29.4</span> Recap</a></li>
+  <li><a href="#no-json-no-problem-maybe" id="toc-no-json-no-problem-maybe" class="nav-link" data-scroll-target="#no-json-no-problem-maybe"><span class="header-section-number">30.3</span> No json? No problem (maybe)</a></li>
+  <li><a href="#recap" id="toc-recap" class="nav-link" data-scroll-target="#recap"><span class="header-section-number">30.4</span> Recap</a></li>
   </ul>
 </nav>
     </div>
@@ -437,7 +443,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-scrape1" class="quarto-section-identifier"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></h1>
+<h1 class="title"><span id="advanced-scrape1" class="quarto-section-identifier"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></h1>
 </div>
 
 
@@ -456,8 +462,8 @@ <h1 class="title"><span id="advanced-scrape1" class="quarto-section-identifier">
 <p>This section contains the vocabulary you need and strategies to avoid scraping altogether by finding data already on your computer and using Google Sheets for simple lists.</p>
 <p>The next chapter reviews how t o pick apart a page using the R library <code>rvest</code>, which is used to parse web pages.</p>
 <p>The last chapter shows you how to programmatically walk through a lot of pages to recompile a more detailed or larger dataset.</p>
-<section id="where-reporters-get-data" class="level2" data-number="29.1">
-<h2 data-number="29.1" class="anchored" data-anchor-id="where-reporters-get-data"><span class="header-section-number">29.1</span> Where reporters get data</h2>
+<section id="where-reporters-get-data" class="level2" data-number="30.1">
+<h2 data-number="30.1" class="anchored" data-anchor-id="where-reporters-get-data"><span class="header-section-number">30.1</span> Where reporters get data</h2>
 <p>Reporters can get data from people, using FOIA, asking nicely or by finding a whistleblower to leak it. But we often also get data from publicly published sources, usually on the web.</p>
 <p>There are three ways to get data from the web:</p>
 <ul>
@@ -466,8 +472,8 @@ <h2 data-number="29.1" class="anchored" data-anchor-id="where-reporters-get-data
 <li><p>Scrape it. This set of chapters goes over how to scrape content that is delivered in HTML form – a web page. There would be other methods to scrape PDFs, which can be easy or hard.</p></li>
 </ul>
 </section>
-<section id="a-json-miracle-walkthrough" class="level2" data-number="29.2">
-<h2 data-number="29.2" class="anchored" data-anchor-id="a-json-miracle-walkthrough"><span class="header-section-number">29.2</span> A json miracle walkthrough</h2>
+<section id="a-json-miracle-walkthrough" class="level2" data-number="30.2">
+<h2 data-number="30.2" class="anchored" data-anchor-id="a-json-miracle-walkthrough"><span class="header-section-number">30.2</span> A json miracle walkthrough</h2>
 <p>This walkthrough shows you how to find some json in your browser. Use the Chrome browser for this - Firefox and Safari also have similar features, but they look different.</p>
 <p><a href="https://cronkitedata.s3.amazonaws.com/docs/simple-page.html">Here is a simple page</a> that will show you what the json looks like and how to extract it. This is what a human sees:</p>
 <p><img src="assets/images/advanced-scrape1-human.png" class="img-fluid" style="width:100.0%"></p>
@@ -564,8 +570,8 @@ <h3 class="anchored" data-anchor-id="an-even-harder-example">An even harder exam
 <p>This time, it’s hard to find the item in the list of files in the browser. Instead, right-click in the “Preview” area, and copy the object. You can paste that into the box in the JSON to csv converter instead of entering a URL.</p>
 </section>
 </section>
-<section id="no-json-no-problem-maybe" class="level2" data-number="29.3">
-<h2 data-number="29.3" class="anchored" data-anchor-id="no-json-no-problem-maybe"><span class="header-section-number">29.3</span> No json? No problem (maybe)</h2>
+<section id="no-json-no-problem-maybe" class="level2" data-number="30.3">
+<h2 data-number="30.3" class="anchored" data-anchor-id="no-json-no-problem-maybe"><span class="header-section-number">30.3</span> No json? No problem (maybe)</h2>
 <p>You may not be able to find a json file – either it’s too hard to find, or it’s not useful, or it doesn’t exist. For a simple page, there’s no problem getting the data in Google Sheets. (This is one area where Excel lags behind Google Sheets.) We’ll go into HTML tags in more depth in the next chapter, but if your data is held in a table or structured list, you can import it directly into Google sheets.</p>
 <p>Note that this trick is really only useful if your page doesn’t change a lot, or if you just want a one-time snapshot. It doesn’t automatically update, and I don’t know how to capture changes – it would involve a Google scripting program, which I don’t know how to do. I’ve never learned because I usually only gather data for my own use, and it’s easier to program it than to finagle Google Sheets.</p>
 <p>The trick to using Google Sheets is to find a table tag (<code>&lt;table&gt;</code>) or a list tag (<code>&lt;ul&gt;</code> or <code>&lt;ol&gt;</code>) that contains the data you want.</p>
@@ -586,8 +592,8 @@ <h2 data-number="29.3" class="anchored" data-anchor-id="no-json-no-problem-maybe
 <p><img src="assets/images/advanced-scrape1-googlesheet1.png" class="img-fluid figure-img" style="width:100.0%"></p>
 <figcaption class="figure-caption">googlesheet</figcaption>
 </figure>
-</div><p>When you hit “enter” the whole table will populate on your Google Sheet. Unfortunately, you can’t get the link to the city from this method, which means you don’t have a good way to extract a city name. We’ll come back to this when we go to scraping in R. (There is a way to get this in Google Sheets, but it’s not very reliable – it will choke as soon as it encounters a missing URL.)</p><p>But if you just need the text of a table or list in a spreadsheet, this is an easy way to get it. (To get a list from an “ol” or “ul” (ordered and unordered lists) tag, use “list” instead of “table”.)</p><section id="recap" class="level2" data-number="29.4">
-<h2 data-number="29.4" class="anchored" data-anchor-id="recap"><span class="header-section-number">29.4</span> Recap</h2>
+</div><p>When you hit “enter” the whole table will populate on your Google Sheet. Unfortunately, you can’t get the link to the city from this method, which means you don’t have a good way to extract a city name. We’ll come back to this when we go to scraping in R. (There is a way to get this in Google Sheets, but it’s not very reliable – it will choke as soon as it encounters a missing URL.)</p><p>But if you just need the text of a table or list in a spreadsheet, this is an easy way to get it. (To get a list from an “ol” or “ul” (ordered and unordered lists) tag, use “list” instead of “table”.)</p><section id="recap" class="level2" data-number="30.4">
+<h2 data-number="30.4" class="anchored" data-anchor-id="recap"><span class="header-section-number">30.4</span> Recap</h2>
 <p>Sometimes – especially on modern websites that create interactive elements on the fly – the data you need is already sitting on your computer. In fact, it’s quite hard to scrape those in other ways because the HTML is created when it’s loaded into your browser.</p>
 <p>But when it’s not, there may be another simple way to get the content.</p>
 <p>The problem is that getting the content without programming can leave you unsatisfied because you can only get the text, not any of the underlying information. The next chapter shows you one method of getting more information from a web page using CSS selectors.</p>
@@ -604,12 +610,12 @@ <h2 data-number="29.4" class="anchored" data-anchor-id="recap"><span class="head
 </section><nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-openrefine.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./advanced-scrape2.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav><footer class="footer">
diff --git a/docs/advanced-scrape2.html b/docs/advanced-scrape2.html
index b99e786..46cea86 100644
--- a/docs/advanced-scrape2.html
+++ b/docs/advanced-scrape2.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 30&nbsp; Introduction to scraping in R</title>
+<title>Data Reporting - 31&nbsp; Introduction to scraping in R</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -113,7 +113,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-scrape2.html"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./advanced.html">Special skills</a></li><li class="breadcrumb-item"><a href="./advanced-scrape2.html"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -270,35 +270,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -315,37 +321,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -362,25 +368,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -424,21 +430,21 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#understanding-a-web-page-and-its-structure" id="toc-understanding-a-web-page-and-its-structure" class="nav-link active" data-scroll-target="#understanding-a-web-page-and-its-structure"><span class="header-section-number">30.1</span> Understanding a web page and its structure</a>
+  <li><a href="#understanding-a-web-page-and-its-structure" id="toc-understanding-a-web-page-and-its-structure" class="nav-link active" data-scroll-target="#understanding-a-web-page-and-its-structure"><span class="header-section-number">31.1</span> Understanding a web page and its structure</a>
   <ul class="collapse">
   <li><a href="#a-special-type-table" id="toc-a-special-type-table" class="nav-link" data-scroll-target="#a-special-type-table">A special type: table</a></li>
   <li><a href="#a-harder-example-ballotpedia" id="toc-a-harder-example-ballotpedia" class="nav-link" data-scroll-target="#a-harder-example-ballotpedia">A harder example: Ballotpedia</a></li>
   <li><a href="#getting-the-link" id="toc-getting-the-link" class="nav-link" data-scroll-target="#getting-the-link">Getting the link</a></li>
   </ul></li>
-  <li><a href="#your-turn" id="toc-your-turn" class="nav-link" data-scroll-target="#your-turn"><span class="header-section-number">30.2</span> Your turn</a></li>
-  <li><a href="#cheat-sheet" id="toc-cheat-sheet" class="nav-link" data-scroll-target="#cheat-sheet"><span class="header-section-number">30.3</span> Cheat sheet</a>
+  <li><a href="#your-turn" id="toc-your-turn" class="nav-link" data-scroll-target="#your-turn"><span class="header-section-number">31.2</span> Your turn</a></li>
+  <li><a href="#cheat-sheet" id="toc-cheat-sheet" class="nav-link" data-scroll-target="#cheat-sheet"><span class="header-section-number">31.3</span> Cheat sheet</a>
   <ul class="collapse">
   <li><a href="#html-tags-elements" id="toc-html-tags-elements" class="nav-link" data-scroll-target="#html-tags-elements">HTML tags / elements</a></li>
   <li><a href="#common-attributes-for-tags" id="toc-common-attributes-for-tags" class="nav-link" data-scroll-target="#common-attributes-for-tags">Common attributes for tags</a></li>
   <li><a href="#rvest-functions" id="toc-rvest-functions" class="nav-link" data-scroll-target="#rvest-functions">Rvest functions</a></li>
   <li><a href="#examples-of-common-css-selectors" id="toc-examples-of-common-css-selectors" class="nav-link" data-scroll-target="#examples-of-common-css-selectors">Examples of common CSS selectors</a></li>
   </ul></li>
-  <li><a href="#going-further" id="toc-going-further" class="nav-link" data-scroll-target="#going-further"><span class="header-section-number">30.4</span> Going further</a></li>
+  <li><a href="#going-further" id="toc-going-further" class="nav-link" data-scroll-target="#going-further"><span class="header-section-number">31.4</span> Going further</a></li>
   </ul>
 </nav>
     </div>
@@ -447,7 +453,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="advanced-scrape2" class="quarto-section-identifier"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></h1>
+<h1 class="title"><span id="advanced-scrape2" class="quarto-section-identifier"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></h1>
 </div>
 
 
@@ -470,8 +476,8 @@ <h1 class="title"><span id="advanced-scrape2" class="quarto-section-identifier">
 </ul>
 <p>If there is one reason to sign up for DataCamp it may be for the value of their scraping lessons.</p>
 <p>This chapter assumes you are, by now, relatively comfortable with R code chunks. In particular, it’s assumed you know how to assign the output of a piece of code to a new variable, and that you understand the pipe structure with the key verbs of the tidyverse.</p>
-<section id="understanding-a-web-page-and-its-structure" class="level2" data-number="30.1">
-<h2 data-number="30.1" class="anchored" data-anchor-id="understanding-a-web-page-and-its-structure"><span class="header-section-number">30.1</span> Understanding a web page and its structure</h2>
+<section id="understanding-a-web-page-and-its-structure" class="level2" data-number="31.1">
+<h2 data-number="31.1" class="anchored" data-anchor-id="understanding-a-web-page-and-its-structure"><span class="header-section-number">31.1</span> Understanding a web page and its structure</h2>
 <p>Web pages are written in HTML, even if they don’t have “html” at the end of the file name.</p>
 <p>HTML is like an upside-down tree. It has a trunk, which is an <code>&lt;html&gt;</code> tag, then two main branches: <code>&lt;head&gt;</code> and <code>&lt;body&gt;</code>. The content of the page branches out from the <code>body</code> tag:</p>
 <div class="quarto-figure quarto-figure-center">
@@ -683,8 +689,8 @@ <h3 class="anchored" data-anchor-id="getting-the-link">Getting the link</h3>
 <p>It doesn’t work perfectly, but it generally gets us a bit closer to a city name.</p>
 </section>
 </section>
-<section id="your-turn" class="level2" data-number="30.2">
-<h2 data-number="30.2" class="anchored" data-anchor-id="your-turn"><span class="header-section-number">30.2</span> Your turn</h2>
+<section id="your-turn" class="level2" data-number="31.2">
+<h2 data-number="31.2" class="anchored" data-anchor-id="your-turn"><span class="header-section-number">31.2</span> Your turn</h2>
 <p>A lot of people use IMDB pages as practice for scraping because its HTML is a little primitive. Try extracting the name, year, rating, and rank of each item in this list <a href="https://www.imdb.com/chart/toptv/?ref_=nv_tvv_250" class="uri">https://www.imdb.com/chart/toptv/?ref_=nv_tvv_250</a></p>
 <p>Hints:</p>
 <ul>
@@ -696,8 +702,8 @@ <h2 data-number="30.2" class="anchored" data-anchor-id="your-turn"><span class="
 <li><p>To put them all together, use the add_column() verb</p></li>
 </ul>
 </section>
-<section id="cheat-sheet" class="level2" data-number="30.3">
-<h2 data-number="30.3" class="anchored" data-anchor-id="cheat-sheet"><span class="header-section-number">30.3</span> Cheat sheet</h2>
+<section id="cheat-sheet" class="level2" data-number="31.3">
+<h2 data-number="31.3" class="anchored" data-anchor-id="cheat-sheet"><span class="header-section-number">31.3</span> Cheat sheet</h2>
 <section id="html-tags-elements" class="level3 unnumbered">
 <h3 class="unnumbered anchored" data-anchor-id="html-tags-elements">HTML tags / elements</h3>
 <p>Each tag has an opening and closing it. This is what a “p” open and close tag looks like, with the text inside it shown on the page:</p>
@@ -934,8 +940,8 @@ <h3 class="unnumbered anchored" data-anchor-id="examples-of-common-css-selectors
 <p>You have to go through the whole path to an element if you need it, which is why you have to look in the inspector section of your browser or use the CSS Selector Gadget (a chrome extension that I’ve never been able to work properly!)</p>
 </section>
 </section>
-<section id="going-further" class="level2" data-number="30.4">
-<h2 data-number="30.4" class="anchored" data-anchor-id="going-further"><span class="header-section-number">30.4</span> Going further</h2>
+<section id="going-further" class="level2" data-number="31.4">
+<h2 data-number="31.4" class="anchored" data-anchor-id="going-further"><span class="header-section-number">31.4</span> Going further</h2>
 <p>The CSS selectors shown in this chapter are a little limiting – you’ll find times when the information you’re seeking isn’t defined using those selectors. That’s when the XPATH selectors we saw in the last chapter are used. It’s beyond the scope of this tutorial, but ask for some help or try to find XPATH examples if you can’t figure out how to get to a part of the page with your css selections – it’s pretty common for this to happen.</p>
 
 
@@ -1187,7 +1193,7 @@ <h2 data-number="30.4" class="anchored" data-anchor-id="going-further"><span cla
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-scrape1.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
diff --git a/docs/advanced.html b/docs/advanced.html
index e90bb4f..8888b0a 100644
--- a/docs/advanced.html
+++ b/docs/advanced.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -655,12 +661,12 @@ <h1 class="unnumbered">Introduction</h1>
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-maps-firstmap.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./advanced-pdf.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/appendix-filetypes.html b/docs/appendix-filetypes.html
index c95103a..cbe72d0 100644
--- a/docs/appendix-filetypes.html
+++ b/docs/appendix-filetypes.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/appendix-math.html b/docs/appendix-math.html
index aedea81..4bbf98d 100644
--- a/docs/appendix-math.html
+++ b/docs/appendix-math.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -1065,7 +1071,7 @@ <h3 class="anchored" data-anchor-id="the-median">The median</h3>
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-scrape2.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
diff --git a/docs/appendix-ppp.html b/docs/appendix-ppp.html
index 581848e..4e2a013 100644
--- a/docs/appendix-ppp.html
+++ b/docs/appendix-ppp.html
@@ -232,35 +232,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -277,37 +283,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -324,25 +330,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/appendix-program.html b/docs/appendix-program.html
index d05804e..da16c67 100644
--- a/docs/appendix-program.html
+++ b/docs/appendix-program.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/assets/images/r-groupby-squished.png b/docs/assets/images/r-groupby-squished.png
deleted file mode 100644
index af6e49bfdba1cc6d1128f378f41fb06f4a7780b5..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 19820
zcmb?@Wmr`G-zOyuCEeXABHi8PAl)G4NH<7J3DOPHFodLlbV+v$(%sz+d-y;5>~nkX
ztG%v$G0XsG=A84(Pkkp8tRjn!LW}|f1A{IvC-n{n1|A)F?FB*vezLgf69G@KPVZzT
zV9JI`cYt4{%(dk$l$Bwaf!81y*boQ|!t)`(gBW-KZ_k2-fd`&ppZ}Hx_uv18N6&))
z@7FM1&j*G)qE*4bh{4E9iL1NA9;72BlXm^+QIz1{@(ZjKlaY#9fu)`^LrWgy>|0OE
z{I#CO$s=y2)V)p`-mChB6+r?xu*CbpN`Gy!erB$|vAKD!@%zzSdP+vd;jZu1xw?y%
z3vVx9DxX76we5m#I2x>E8zwu#e?R%ZlQ8b9G_PEkwi~cSwY02C@9TZ?;!p@yN@YnG
zbW1dB{v@9%;+>~gYdc<MAP5=H3oWT0R~XL{57^&w)aQ@UNqlX@rkKbYqf=$kmnR=*
zHkkZ|!*yRfMZo2|*Hr0GG$Y@vI)tisgyT)6O*P$#<L*!yofW97>g;?H@oaZFl`)7A
zF4*tCp2U7(ibc1c0@uQf#umaYqa1xfKN44_B2UO?pLKpPr`6zk!2C+DCSE3j7^;aM
zuM>}&AvyM8zV1i$hq3m8u9%7o`nQQGY$N4b2iZ~AH2s5kYb^<_mxj6R9u{=!9RxAs
zqu(`(i7mP+dH+un{8xjF)3AR!WUaS2jWQ?yAZnRyEohN{zPq4@h)%qofQXBTa373L
zqN(F9s{@(HkEphpn|cCEwcH&F&e#r0rHqya3|dbJePwVL%DFnq^DTA-?$LRyB0ibh
z5dH5KhsD(X@jgiZYGAr@5c!yj+j+(J+Og!^S)u)7aVdRqS;Kedqs4_06!lJ0CM8av
z;SzXAHfuDGN3>QHn}OHjF($T9JBv^fbILgQJGZ45$^U6QKPvEBEaN=i+x8pBk@d_0
z%8Ew}9P^9rS9wOQJ{-nvFh?DuO&(679bkG#Jox0Fls5OYy<dg;NXS<a%8${GpmO<K
z@$XBCM*h1uo;zjI6alm8N#xKoC9{vb-mb|x#ahI?rKGZ-+0d_twx|IYF~|rH>2!Cg
zwRMDqe3_2cJ;+jxSVDb&{vU(vxd93h5ilDF>hZ#D)3v9POiE8Up>A7ci#s~Xv6#EX
z-+CpL6YT4Pk=#QkY~o}*7Zp_<3Vm?bKFs|5Yk&avD35?yXOzmJLeFLtn;;Q@M>ONu
zjon)PaU?^yIhVr+sf+rwoj<inXA#eEnJEu>0N-;_83I!M;q=$Ihn37~1663X?@cho
z%eySc&u~;Id+v<ohB$AJ{yOT2cEt$h2-TqWYx;BN-?Rb?e3Z(u3XwuY4xN@_gu~kO
zUw4<~L)8X)7>6ouEIH*TGF|Id;Vaf=G~dhEUcXUSHxY}@6m+vq!HbmpYD7%^cC|8v
zba>3_K|sN0JMn?GNiDd4A1@~TokRJcP*I(`p6x%|pAarV3I=M@-^7196#K&bop!;B
zT`{2vwYsW*8Wh|UL9#W`ImmcFl*T`C1ZP0s5eO5Hcp8ipkLVtT(}eM~KN0Kng$GJy
z80Tv9k9h_(BMX*Tov&v%t<D(bx8}$*0>!77>3OC-kv@5?IaO$=1Z=nq+k{=T_uN)f
zN?_9Q*kPWwuUma5_}ERYJRWIHC*9vn>=T9=lvAnKdMsa`^<U)mnOmAMe~V#4PmmtY
z;QVWPF*3s?V>a_z>ZS-G4H|9Y1@{kfD1GA@`bPK4ut^F`PRyk*Pt0qJ{gBFIqnWMa
zZ(jr31-rP)<(a@X-r_CjaW+)DTFdN$&Nd?S^e&Gti`~B2OJ}BuCBfgbH$J$Z+H!>G
zAJf1Xgo_S)VM*9pSOR(X1f@Qh{drb}hG*~w^6T@9SO;o{iTqdhKYnn&{y0_QF;$)u
z6o}+SvW&dADn;~1yuq~R<qK8PPv_fgjfdx+<&0jpHz416^B%G^N!Kg)se~3D*=+?$
zOAO+_&xi^eAKdkoV4xcMwcl*!rvs;*Vdt(i9@V(@QcLq-<`HhpQ?iix>sZ~G$;!2M
zRQk8a_ytD{5lP56^u5GZC8`za-&HYr=8eBJolG@WUW^XSkC(IbRF#hZ9O6<(7V-S^
z_lP3|Esyw(vzvP8#C<z_r7Qmx|5X#blddnYQfVrtiu$%xq_^+uRL#FDoDo#gs?k4g
zvxXP?rI(gi`;qCW`Km=JLx|s}czimU*8BX#($Hj?@XDdkx&6swG(Pw{^@ovvE-S_$
zbwRj~8S7Ga93Cz<Q|&1P+@+%&#f((ID2b(4{x%R!G?C{W;%YnD>VJo4A<2h}<o>yq
zZ^&v=jA!<P<jrybhWYN~cR`k740w8lz0yWkv+wG1UBBXk-CxcAJ-@|DC4teN;<TSW
zkyQeV_1R;sJA%b*w?j!RSB9l;T^<J-=cH}B(D2iFZ&o9<&KX4<orLFHh>i(89`+2H
zEm<4Sr1rgx3yhD8cwqvjKx#*o-kbn4(!L`5Yxetp1Wx0x@tLoIDX7E?Ovp?_S=E3g
zf}Y23Ih>YOO#5TN$sClito$BdLl=KA2@%7vnM2-Ec%oD<sl<v+-Jj7ZHx@RZOC6if
zbMWtesRLzkRlC|EO_<hjoN3+;g;17?O7?Pd!eRDHQprCMFTYZW_pgk{?xK86Hx;4v
z_f_3~hRHp<rmEJ}oDzbR#eLAYE?0JM!)fXY<pAT0T4A^9D3;@dsVeew*ZlYX{a<^_
zt}tAw&Mu=2r&i~_@B1sOQ;yKuJdR1)D9C9U{uymzxo-6bRe8u;NE@>&Y5vbY4WvD!
zH2*yJE+<^CbojjKEKWhuK+<*0ox8iT5$87;JBJ*!cqqrF`bl+|i-A29HpssSpH%VV
z8kg!1zgJQsT-6?Dn>XBn$_|VOIf=plvsTA*!13X`^~Q6@-#GT&_sTcafL3?+(Vjmq
z!C^kKKy7fZvQX=dn+4&Yhhf}cX1E=4B*_W{wSQj~BRpufxSUaN^I_)n#n15{gqD`>
zUvbQ{Sbs|Y3O!45e!iyCe%xT%LKXtVMCdxLTK&6OwSTUl@lh~vuDLQ1W~GR>IJL<}
z9on_|JJ&DHJ|=`><I@}?*irow1U~OV7XlLhzc&d?5T$X|mbhnaz*#?P|1EM8r>8<G
z@dxu!f{ay!BcLmj$1!Dl%Ko*-`!iGeT}V&lD>v_4`gtf0<ed+egvJ<2rhJor3HJTZ
z+HC5hqNC6!P_Voy?_iAL5utXBA$lL3y#5TK+5?e8uY7KI7*e>cG*_F$5K>dX(xa7i
z|GX{Sm#{J$Sk*eD7RTe2)caPbmVHr9#cW4-VW#f>1$_UxC!PsR*)7p&RD9NG&%=7z
zmh088q;L8WqsRm`$vxJweBrx{ua8&EY3ifPTW%FZ9?mB^J%uC%5};R$o^M^}9c0V#
z4u4J?p5AO_YVj16Y~?GZP=B*+fSk{{EqPZR&ISEXe*+k<%??!Pft`E4$7{`IgKs{<
zQ({l@dP(T<PIm<}0-md>O-&9!8S%dc<0Q=9ZyEXAKv{JvrDVbhu^P{o210Qd`X8>A
zC|)@4PRdUi`95qn(5uYW*sy3BxFigwaLU%qeq`Z4>BMoWeB*Jt&T$pn{-xS_hQkEk
z>`#TsquUNNDbZsJ?(ccwBk6}10*rksvbox%2!>%M_yPEdp1yEt)Q^+3umia6eo1`4
zIDsAbQ|tOi(zN5-IlI;bSa`(V)vm8WIsvqwE`Dp7ue1l9w!ve#tK|`KTEwezB=d`n
z(jVU6JQ>{KQ{AUnW}H+JQFtAT=j8L-%yM~LuY_zYdL9j>p19u~)NWLbh$z07N5mlC
z_f<cXA+e!7YC2PSKYQHVg|FAUSKdkvg^!xEu8Pn~%xPN=EGqh~sW<d=d$C6_|3mGc
zGZf7q6V?kRmDi5#Abq*Hs|j!1Z07X1W^kk=+RFDPzdHdDxS8l>yo{A}j`&zn_zgxY
zO#H+$zpZ4XHyTwNiF%*198<2XD}Eu8cN)0PNZ(!_>|HT}9<ab@e`<~QN^9{g=oWpS
z9=;3|sTK4y7ne|fv#t!~)gl)D<m~+RE#|7{#V(!TL3NzhX&*yOxg<ae%r+9Vrapwq
zCJr||-tKKos=Za7d<lVJMC&1S3KwlT?PqZWj`iHgmye?lg?t`(ygPc0A$%U4*(Von
z=Ks&ZObEA#B$g1IFy|8evwMU!pr0GZH#*8<YEVI8kRzH`s1w<6rCu`Rr6E2WoiAel
zJ1!_JUAE`dp--pxa-07t(2SqX()hBJrp1yGO!n{?rm7cQ=WL*EM~gn^zVmfYDi6_*
z4SQvKvIjFU0%cswYQG|XGg1vQgLRUT#Pg<0^=jQ+$GA~Oh%D1qrk|SkD+h9pujK(m
zxd9CSIQW-*Vv&5SpmO#$Eo`h|r9OFvbXs5r)N1j>uIeQJH3|IOXNF!k?#k>n0!&43
zNS_%!J&PQ>k^25-?#euy<O3T|41-6HQ?M+6lP!o9hB9GEzvgfm*8~`diYf)P+-%lD
za?4vE8x&mns#VI0$XpVUIgJzHa@sKuQ(}xQs423!Gdwm^7V~xi?xCn|(ay!*A+ZCQ
zDoOFWT5ww_Epc*}lApEjK;I~-8jeMx@M0g^XWFA`{2#^#AYXDCzk0r6j5mMgT3rNZ
z%$#SMF<Xqdw1!D9<W+~_qX|arJ!0L`maz<1h#0otomplHnP06$CRM^r)5u1~C%)3V
z9dlDY3?eW*rkAoRZ>~YSq2YN<hJ9twz@^G(*-5bhj|8zmPNo94tES8#1Uh(8P3c&Y
zHE%@O%vPU7gu6=yqn)52-H5Sng~8=;l8GFP`d-V=13eZ)f!L44NiH8KKh9K;Ms3%8
zk(eoOqPc!KXLt2-QvQTH{C6h!y}}PH2+{tBf8bd$7aG5`o;FrZI?&2oWHPI$YcfNe
zV0C~0Z6b26S;W-AMz02<d63q0m#EK!p{<KYchO$BN995Mi*cI+4dIdhK>IL`iMGM^
zRI9R-Aao0<<W0)?kUU~Sn%i8Ngcsz$$P3I#v#}oKWs%X3PI@S=R3;fg*KA_y;hMvn
z-|W~-yTea5#oQJ>EZ5B<<i}(wyz}JE$ufN(J<7wQ=_lmx)n`C79J!9bl#!vAAAG;O
zX5UAbL@OY%rsj{s<wH6me3druIPl8E(be`JH|9r$DG8Gj`Ae@v;8^xx3i@^b{=yxC
z>lD`gVQ2eDG?pYy&Cvx1;D$BX@g(VP0K%(fGZ21ebIeJayhcGvt^PRGoP4-jm5t>e
zEcS&*yEF_nLW$r>4J5!R{>J&R*<`TkcM5@$;gdZ_khJ43?d@mvgWuZge!ILJ;j>$#
zjZjl!W5&&|AFs9l=W3*Y42u}1(@n95!>X%2_4M0fU}K2emxMhsY2);9uRq1R3aMhv
zf!Wi$YlLQ_WUSBA-G<w-#JaUEhC;{^h78AJy#?G-YTeT1Q^v@Np1_1_A>kl~b*xI5
z1vXhGHWyd+T)nEJU>Jmcd}rcEacyn^6$+vnt~K~4ECnxN0rCiaXAz;AONk=Sw;$Sw
z_X5IL#DTW~t)FD`U+KiVw{rOBxs3L@IWnUi!@bqcXfrJm5M$x_sR;oi9{h6vJyYFI
zEaT%O|M!T-I{6f8P=nTV#%B7FFxiXgtGBm0h|N#0>pkxanfc{72gHf<F51>Jan3#?
zx%)G?2lS-Y*vp$$!}<3uHsQ&&<@`+z2;sQ>VKhlf@;X0=4D6cvw9IX__D<NoN^@A4
zO)!Dnj(0gG>tSEKWH_f-G|k$df!0=Ar8OmK-mmB=CiUW|`J9aS@LB5KyE-@b=jl|=
z5wyx(*)9Y{c&?z6|E8u7(|*tJLW?+=t^F!1U)fiogxs+bHPL~yG%e=fAApPy?g9kM
zt=2yI4a++o(TmLWX`R8t+aBX4-)>rpOT=D6wAHlb4(z%=De1HHBjnt6>Ye6E>(V6}
zN$m&!a6K*~ABWbY{HP>?{D>sF9|jp~q2-TfW;5M_B5o%r0JR{=SBJ8V;ORq*iI@^!
zo1^Qhk*+Eiy`N@8Ge*vr`FpW9#{vz-g9){XDR6QQT2m;~Er;XWruS(Eh|AS4#R)!R
zAD>S;JSGuRnH%Rq2h6KtOwATTo$}PShM#2&9hcu4dx1K!5@36%{|-TZNdDVQLpskl
z<`p6oV{)p5ZIiu!VRA|$SX?Chv1f8$>rRDlB1x~DkfA2quP1rie?4pyM^X0vLsU_f
z`L?B~B0+7mrDg5s4^NK|=A&8S83gX%Spv3!=4~URnnqY)t`YP6TiBrFhTC0gQ&32&
zs%C2KwiWbSSb8w@*`3!F(alX#vYX>@N!^g%=ksi_zW%Mf!|PJE)LExkyM!C;50CUI
zC|ljlXINwNqbjkZxy8!gjT9{ALg^QT*bmn(Fx<n47JmM8KJm6a!z{vbC`P0_D*l8_
zuaEbkq3`vAVDGpLo}OIwNhRL6=PH`cLxKFb>8%iCwz{a=K~Sw~iAZJSn!SPm{-#~A
z=<ZG6g_(lnm*Dh&&Lv||JS!Y%XTPo__ny^hHeZ<pGF_G)P9bV&F_=uHyg8hdkr|!{
zO;P3ep+=JO2U-fKT@n^GwrpQhxnk)ee9nA7thzec)|jPKXEuUoM|3C`3I-DoCpP8Q
z5L*vOrp6Dj{Np(3RlRS|V={z2Z<`4fydl5!6YLZu+00n~Zqh7TEb!I`Tu`7pkESUi
zOqWR*r&cPrRU)f)g;?s4E@Sh2IBev^Eh=g_c8;js>0}NxQ(6ppXZ4-{O!(((wpxLT
zp{hwjUc+h!=4x9#6S{-c{_I}^B5NH>%)Lhn85U5k2ETxb^HtKB0jz%Z^@>%$4vhaw
zcU{Zi{RW+8d7wKvG=KtiA{tSokr9->fFjR`KFGyh?4N>tvKPz$t_k)T{dfKg{qG{c
zfldZ7DUWl-D3@uDtH!tXXR6GH)A%2IA73)QWSwd7s@Fs8i^<fnAIV||Fzh=3=I$w{
z^W_MANOc)<qtfd7HXe;ig!;EcS%?G55@I*?^D7^i$rZ|CWHoH!`>h0Dg)H<H85>eG
z^DZ5RQJP8#A_p5Q{pN4=#T!9=^h)uEz)I!}++fVH<*2E)!fNqM3^{Uf6fhRNOI%p0
z{)u8PNMSW7!HD6fG{%e@J-zS$N1**Lo%VtlR*iVq<w9@e-@NA@lsq0yLG#t*uRRQq
z5v>qD&KIog6l7jnZP+&K<a4+mFMl#1sz({Q+<Lm5ZZ+E+OsSmxSlkD|n!PtPKDD+B
zj>4LyD{=J7v6h+MQ(F$(*&(;TLEpJM1k&nEbV68aQKXBl8^2r;Ad%XO;nm4KG47>R
zc~K1qmFU^3^COH;E)f`cQBnBb|EO5J9~L~)uMgj6`p+K59Oie1KcbY(?k^cJ3T)mP
z4nmf;tq+&PeXTW?x5phAo?dN*(qHi(=e=2T*lRQFBXF~Mj=?l9%7c~A2R_35d`YX5
z;u<q)go_$RVq5p?N)ez9-ajf!et5Orw4=U0zgcn`5l$I+-P%Oilvmn%8|WhL_fra>
zCRvoAWRG{!KTD{4e9y;Zy(6`Zynj}l^a6-)Hwnq9Kk=9)u;=M@>pgWxFnV0;)5DQ8
zo58`rVf|J{8~}0+Wg<y+Og2Ss)^QyA%bU)}74dv(L`1=J4r|Sx$G?y^`dP})X?E~R
zi{B@{E@~}gY)yz*`e~)L#86tlnVO)SW0ui))N<E}zUuCC)Ue0ax}!#wT;6xfg7EYo
z?mlUQ@KFkqEaXwiev$Dtk<Ebl-kHQMP}g-^ZFO;bEZ2OsGn9BeYJa&$c%d(<?X*2@
zHsQj{rct}X*k9;kQZ1u2f@iij$p!#9mlpcZF!tCq4x6qk!m3}`VvrC3fX3HrQKJ1N
zI#oIL>^=ZseSwqv3?R{bR69{JDx?K3x?T6>rCI<T^1`A0mYVIV{nNVF{q>1_A1)V~
z$V<_Cd;XjCIP=Lu)r=5##QOfODi8*MwW^!$PX``#Y+6$h^`-(uFLXh0NqJX()oX<B
z1+aTC)nWnWwr)nJrQ;zsm)aspx@KY`G#o0K9#WV0E5gz$cH8o$(6Jc**&ydbW87rE
zlSx9kmvPLVaN&m(rwalTf%<FOmrd_NNU3I7Y^0D4JqozdNWYZZ7z0Y$rS~|Vmi`Jw
zXhFfw2fZ;bF0a{Y$j|Sh;FeX{MJO-?Re5I2a8h{fGMdhY1s{dTU+<)%=dp{T+!;qC
z<v()Sdi0#0j&2p!>U+yJE&h2`+vM~4F1`r-kb3$-78tIDL7=*qO&>Pp57)k_I?*CY
zTWpVJ6Y&ac6jc8f>T}VK*@++->MLp3{r)(Mtd{~1dFC~N-E9%b{*NAVBHmk-Hyi+)
zFK^uvF?>vI9bju548eLE3Ng<J%A8eDDe7ltxIewSKO5n+UFT>h*pxR(KEpZ5S4!p-
z7k#>SvH);tcm@?&*9Sl-&{>&p1SEvB_SgaE_J~SZm+wMG_P<^bU5TGS8Zc}(F?_Ej
zH8X~j=*|G9C+o53JfZw3V)0T1GiuegYKVu(zKD%lbVDafh4FZ1gC&V1i0y3un_bIJ
zrtsZ<RTX=-Av*-z9|T>DqBv_t>eNER;9rkMySDOpRs+Iuv^iwST-0%0?@xLx&11?-
zhHL3RT>``TY(&6>17+gsjdh$#MHu^X`qV8OMlEap$L;n5T;c@9H+dLp{UI7<r%(?J
z=5e<Ez1Pd!6q%jbQF-C#QJ)SiwCgtZ6z@K+l6vm{z+xM2S?n3(0oq}<+ucs`(|U~T
zebwux<Ifo6Z{AA<i-?_UJh&hHoX(MLcm6s<A%%9o)yagC@^Cr#dbbrIFty>AM9b11
zPBAzQRg3)n4Z{L^*DYz5r@xJa7HH4%ygZ_1fbqANVOKCkb>BT8OBS@`wEE0^95fG}
zH>q7rxNvMdz^@4E+AC{hbLzyBm7)lvh5D;}!2prPk#LXHVV$5+WRci4y$LWbp&-U{
zRAqG!rl2sFTf@x%%p!r>u4Y4TiM=9d{AxtUGCiUG?y!;5HjzG#Rw>X#Xrfr_^^-$x
z_%o01LFifmgq<rmI*Rwb5h9i^1^3x~DN6bRs!vsuGnv^;lAT%x^#QD<>RqR*n@Q`^
z)`KlRl>&DY#eN_BIoSvE=e|j!_=iF^$H;}4C6EmcE2CbjBNb(_#8-~NDnPeAEBV}o
z%#gM`w=Yd#fNS#k_+`)zOX8zUWUJQ^Y*g19^%X=Vtj9vDjU;xBZV5xj-z(=k6S*Am
z1zdp;2wB|gEszn|qwWU8eZM6$k?W6i-R+v*f`l@Vgj%e>p?F3fxU*$N0J;0B@AZy6
z(73-)bWy%PjQzXGc>&tyD27!sGy3eNI^<b|$>lpAQV>!Uw5M4mnvJbN=Iz)z`R%*_
zs_~VY_<j8T0;#{8oC#xze9_K+InC|MW4Kd%HmxHk7gkODMWUW<U;D@qmwA72YaSOK
zPM3f(d0t%p5oiwfKG-E<9JYQYzi@Gwe=7~mn?WJ?oB74n!N_w@(!)us5w|=&-fhF@
zP#923yKk9B=W($U4~NhZP$i#~xTbYVYSV|u4B4N@u?r~i((N4yQuyolO{E^&$K$L>
zGRY@{I?|L|WG-YyKF0fIdSCYOoduS7>TmZ_^sDQloT>Q--%@9SsT$WJ1-%GYKp=-(
zfG6kP#S@HU<V+En2&@K;s-@@U7MYMJ>y?>C@^>daYcxZB)Dnd#^3}48$OCD}wDWSG
zFmNSL-~BR7KtLe3KrQ5D-@NK3vAd3+i2uo<5Wwkuwi6PWp&QRt2boXh<1o(V3#y8P
zRD53I+EDz2!92_6ctHTUQ8$SZ;bJYQdCd;q;7GIShjNWhU3g3A1gFhp_MFZ;ja*Fg
zE_qe8(vw#WM!!6#{30=%*k`nq0yKVx{~rp4uRfyS5@#5C{8tTTuG(`W#pNx#9lvBc
zFdMS<FShKuA8~OWE}=s-iQyFUei+Ijce5`Gacq!uqpDx@rj$wFbMoZ{I6vk;06vgd
zx^&hD5%g^DhiEEEebEaNZo5VOw*#a<URmCV&{@Aq*9oz1JgD(0Qr$nMYEO7PFA%*^
zV?aO_?e#XkKUsQutgWrzXphLau4uU43XCs#pAJZ|kF!ijfWb##=XNB5sN;a*0NXNp
z<ok5TSdY&0hf<&?7~j!CP(t|Pa6!-Ks&PsB>PO1UKJVT@sCzcK@TcqCyFa(4Z=S{z
zSqR>W-X8~$y3IROb`x94edDpAo?CMFa1y4#czZ^^&8eb#K5*_h<k*aPBnfCpc@Asc
zs|ZpQ`Mw&jit}+q-}&?+Q=I70c^YVxjJF_ZyeQ1^YS8&n0!@RoMw6(2*)3Piy)SlI
zvvZk+6x>)YFjimL-EN6K?UY|F`TF|a190aaOO*E(ZgzXx$oHUQ{=3g(p)dCg#WKUP
z(r(ie!6RCe;BuHKyte&YZ{c_&;nkkJ>cvlcH)<6Q{b{0ZaIK4F=WlGt@pc5SZMS0=
zCx!DK><U!UJBPY&*;ifm|BSuHU5-bhu<!-g444;-F?GBRg;v42E`dt(rvuOIYQ9|a
za&f;{GCs<SPG;N@p|kL(QHW3(fYTgWccs0xS0o));16WB1EbhFQAFVp(J*-{p{iky
zW>ALfKcDtWH-K&Jj6zv6kI9AE?gb0!+uA?xQ61=!E2ng=JS?4fZ;hbT1s)jR<Hh+g
zwwQ}aG?})FZar8CG_zE{Ee4!k2WpYUBL}TkUzC3S!P#?K9F{FqH|9aa^#M~}Hb9~Y
zhatl~=jo*ESkLp|r+3=y@s`Ejz47m-`;jMi*R;%7>-vqv$N3@FTTl7nA*m0eS^iNy
zM%cCx+zLj2QU=HwquJ6gw?K_9Fk_(+K8k&Q89!gO@$)v03c^)}jK;hd^8`SHkZU;O
zn|E=E&XZwK@e#qJrVm{x4}~$qz`Wa!qp{H>cid$42~3K@&J{0jIt??TV$gSCw0VS*
zMqznsUN}swwGcX=g4m*mH`#Lt-|yIIE|1LI-u@Zh;rPT%D!n&dp2;W6?;C~jYM}QJ
z&^=Y>^DsU-SIIFbf8)8}r`l(MG7&gj=p=S&IEv7=Gcr?Xid2*g=c&)61alqvZ>o-&
zrd9?Yy)@5V{GM}9x1y=qQq1YY*ZV6>LIF}iZTs#f?1nsVLu~+0l%owsb^3e(jc4>J
zpUMI!oj)q7zl7jnI7XJ7NH~ld9c$PRqbl##w5x7bf{`7~%fM-fyG1C?$gz)ZH7(ei
zZ#tek6!9e~;S1l}or1?Sx6Bq;=0L}Io+yb+GRALca&Pq32S>bg(r!$Ki3liuJw9XT
zyyrnx0o<2NKISmhbX3dxGF2?cm(3yqxpd(Q4@Xar<*hz|piM2EXERr$lP?=yMzdZx
zgN?CYw-)(We|w5_nH(~joy*6x6JN!w8;BDH)n}0m=Mo=&lDM3;dhMW+>tpeS-LjVD
zV6K*6>!=MXble7iTHE@dETP6-dz;Xz73T}$iFPU@uq<XB_F0O`!qPOXuNy%_2_*Uh
zO)}gaVomd|!H}d9e7Zkt_Ve>Y5I!H3^y!oFRrPP>Iud^oI}nzd|I(u;%{S0D@+wSq
zP0y~Gn<YI=pLRtvW~*3`ZGcp;l`<2ZKCWXvgC0y=@+C>V<$Vhh<R);M_I8cpiSUJ8
z&5VUlX0rX;ByearH6p`zMytjp^CW|u#wNRu)Uu+|t%ug@-cPCtJb32dWt`gg;B#*i
zd)kN;mav%Nx_y9Ut>VL;RJDZ37D{O(E>wHqem}Ao36t>VU&3{GG%|GXb6bqlEY;Ph
zA-qfjxZ$KL!aSqNnC9Dd9YwP5q-;c2s$UYV6&(W~0e9E<H8pQP9)DlVOO=Gw#muMs
zEyfHrW{$iloI-(}XH!8W{J7p}OUA+K3|VFQc;E;yMZRjIlhH1GniCOXQlbE9wNk#)
zRC^a96tN~VN{zG-@KQi7z=_;xBSf-PO@_LqZ=`hYF#TQ*O}5|u0!MS~GsUW^QMAAf
z*4w^_-5WlY-qEa&{Hqldry&8%@#XrRCLWrWC<F;)1>(Fp$v=hYiaGc<sKO*~>nmu7
zXcf-L8G}p0p`EA2%Y(bG6dLSdHq;c<G&SYNEFG%dE^ky*ocn&~IR>%pe(;g6+i9M4
zbmoj7%Rx6Tj7Kcadj^uR#}i(0k@dHztNANxWX8WC!h-=pg!Z@4eLl^qK$fk0XLDK`
zrxi~{Vq091u&-+FY1-!DSGp5xa;++v=yXuMl1IvQ4YCA}r)s4YI6Ra2Ykn5!kI%Z=
zGWhgh2FKMTE#H&#Sa$h1+0+)ABC-z)Pe899msEt;3Y1u-Y#VKhMm$f2L+Vk#oE6-;
zW>2@PLoLzz{rh}tzIDy+#JH^mq-nwRH^O23=r5i4tb;5NejCM(_ruP!*Z)!~cG;T>
z9bDT-suA|Q9GvB1d*hfPw(t4IG%F*{I)CMbr9O8=4*DwXe5Lt-5EH=`4<DD+)ZiK6
zW-`#UU6&*wVhq<(mxF;OZc0Q{wYT9*>u5;r{opr?0bb@C3#RnJ0&EDtvhl?kH3gAg
zVA8}3HpEN&mn5)mJdo$Da3a6vInp&x(8v)<96ME0;qQ{|;tKJBe(wx6J%{I;!KZHi
zeyT<`ayAU^en}>vu19pXq1xLbdA53|mvZFcVr&fo1Xi&NDzGy&`!?)769G1?wNVEi
ze|vRGtQ*PaAaPeiF$A<bxX(HvvETb#e|?up{!_1mVCUMyHihXUf6YMjztbX<-U!4A
z5?Bj}W{aw9&X>$DS`<@yerp*m+)>gdWZNyZn14?1OETX2UNdV>tu6KufjP{3H?WwA
z+K1y02j^|18f_KBU}Yy$I6<rB^bJ;FpR0elKc@|r^=*pHuY!&3+^EaQiW~lsNUR!o
ztJgDs{v`ZIW%)_u9}6aC7X`vDsmbHuru;ll<Gdj(C~S6nd~o%bjYuukeWwET2RC*X
zV?4IBY*pg4s9+;2QF#bxO+(bu<#kA^H9<p;s)p*%%Xb<GsCV}`Z-rKtrn$bxg56fe
z%YrsK-vRj>|94_pABq_^-~ymqfQlEh=Ni;8<AvC0kh?1mvr)?IycWUxX(`MLGD^?f
zNDFDHKJe+nT262GsRx@~=rPh-Sqe-Ueza_E)6nJxn2o{rUwnxnE-G`X+4~O>Srs^v
zS%y?#Q%-acrFHU}rX03wU*{XN_Rb(CD_4v|AP;7Q^vynPeIA9<$CS7IM`_ouL~jCn
zrOw9;=$-c5w=tYF58iqEjx>6nJV-n_taWmVe4DuHl_reSE<(u4HSxxNYCjs5O01sE
z&xDE;a>ccV-z8oT6_<h(th*_6h7Rw0*2Q_fZ#R;R?nOufP2Q&4>GYh_{Yb_WhY2|V
zymT--*7LgUElkM{Zgtumluy%<U^Lk>%Ge+?gLKz%HvP7f{+R8&s_R+Q?2@3)>*afQ
zzO!}}v!WM<bDR9eG_cuX+mrW?G5nLV1!7FgZ0YJ2uYJ{3?d4RJg9S*P0Zv+r7>vK|
zhbi_W!g2&rbIhNx%ja^&vN>lt(R*O_2O)zHsr}41|BL&bNia71SMeL;W_JN0O?%_%
z5A7LVr}8EdF1W}V3g2-C8<Ke_kakh@XVCG?(ca@PKRavvBqkKz=c<|~KU!s&><5@`
z(~VOeEiBD01&m7_2em$+h8rIM?00PT$Qq_6OaYWyjWJFp;#F6{zi{MpC8$PT2`+0v
z#f6W5fh$^J*@}aAKNB63;QVI^UYuG5h(nwne7-fe!(9Azy=$%6>M*HrH#@6%P}+LB
z8+^n7&yV>ZgMLZvvwu?$x#!>!ONWmL+UfD|o)6vS<J96U&oG+*u2fc`0{rJDy%B>y
zWU0#De`uL$ap4ib7T8(#Gs5>o+l5)f>4Kg;AADCHgI|*ECh7mS3M2{Clqr>ui|CY#
z!#ZMDj<D;g1XZjva>5KAQ?gTE2X%VOJt*--6274hYsdCteN|Y6&PXJn*A^onLA4UV
z?YJ)YMA{E`LUi*POyBnn%Diy<=yPSi3$~>3qP=5bgtO-B5)9+0=y)l3oym-97I;IU
z02{`CA#JcdT+<KAjhhth^z7+1*M}@;{C>?C&!Yo$|3y-{Z7iRqmaRpE#|Ty8(57$!
z#&Ur--_`!Zix)JKK)esEo{1PJF2l_3(uNTFl8F{!K;KGu!hxwP{f9)h%$mI!51gQq
z%+HxO6S%>c>8B&AgWz$(R~F}?MF-LB%D-5dLca_|{wNWDD^UNX8I2P6rA*c3AJZs~
zlTb;$?4h<OX59Mm>f`TJZGxYfh&*4{4G}&F;^r4N{s}%Zbr;<c(h{U9`)!B)9*Sde
z0KJ^jXlamN6`H={WnweAeB6}iZb$cT2o+PcB*rAKh$H1N`$Di8f=;p-<Pj(?z-K-E
z<A;<OaEY)%$6@c-;!?x%)iT#ObMmUVuTu((2zYi`i~7{L!{9mLlY@e)(5^+7EhU=2
z-=NTF<@IgaC<V~O@9w{#Am1g2#!yHt&{XGCj2de5=L135i+Y7HP1nK4>ul#LA^Ku+
zlblWv?Zs@1x2N5f(!(yRYA-3L(j>#T7w<E8CWjEZ`*wLF<#r79p68sF;evXgm8MK?
zt5)TeR!l`M6}T0X@%wnzDE0k{Wm^e+JAD-_2!M1n@$Zp_<me|d0Xy4%mBUV8E=yLK
zq(^Umvlg&S<rsf^x|Z<44X!RL`eA<2LAl(&rlq_OI02N#LmYEwUF18h-Q`)U@{5ri
zvuB-f(Vtk*eNbo0(9`Bg2(m)rFUcJ!kLH73jXdkegc|noBq|x@u=#4y@Qswm@`F_9
z62LmVy_+{k=RBg;0xOSuX`-$dx>T>h9IIv*B(izu<)zJ&ivoihyen6nJNg~ZS&zF<
zh_Kq2zZREf^`^U1sKDu<$$qlzH=!$9SU6vH5&Fi5BVaf|ACHKmf{wp@mNTgAu2u0q
zL!-zY6-kU?k*PYT!FxMUW<>xu$#Bv8U5g6hNg4hrY6)X)*5g6EZB}t-n!Bp4Y3b5{
z?V+I>8wiO?XK6B{C%g(5w)~g!P6!tyn^hP&8f3Zjj_WU50Eh(m@5;2r2k*-K8UNdB
zU`J3O1tfljCb$Uif7-s!n4<xU@p@ODHR+?-Kh7lf1xB%m<e7Z$Py@EyFM*(O8?^5F
zf0_yaAB+z4OJShPesA(Grweeosf-mQ5rWeH?KA|7VZ-Hf(f)sZi}XcCiS`d_JH!+v
zdb>2PooKmP_ZtpYb%C2xCg@#IXQ04q^-lvyuSZZ#XPZu;8imfrd^Wlss9IlX@|a9h
zuSAuxCe{<^UwS}SLiui1p+m2gZ+0Eiy!bX7%Uu%b4RuLTQ#5CVUS&MPW)+06?_By0
zU3is%`kQnE3@cCa^Z*PIZCfDgF&nB*S9c3Y;v3ZyDskLX+LD^O5IP;V1(%I30wdPA
z@%ehEXB3eL0CD-F#pY*ye-JwlaHwc`7o6kvfpCgQIuOU`2V^!%PK!atIQ$(uz+4d_
z*Y*X(b@SDA)MXlnA?|5+L)}{2SH3p`MrN1$GyD_*819WgWXGwy2$k38W+SQaWrNEe
z{Z@u&87q+CupBG&-Ce$zHhPV-Rgmcu3&gUPrul75zrL!8PBQZ04?|j%Kk5fK3trb4
z&iIl^2w9$Lu8hRoGXFRRlX7D+pmp1oSBsD5G>^ZGr&ms;1vJQVu&eu~I}p-Ri2WQB
z#Sqx*d%s5UZ1~(%*DSc|-2mPucd>v2!pJif`a}Wj-}W9)4_8(F`N?W_O;1SV?paDH
z9CR$hfR-5l2fzK3%EWZuf$Z}>>aVq1Ytj*137lo1cFE^k7Y;!a27#e_HXtX;Hn<(J
zS*E)rVoZF2%nekSbo9p`jY?wtHoHIRzz}7{2|0ZWBw<+$iwur4M6c$Z$OT()?3zww
zQmApii}(OpE-AEMC#M6iqfV7O)C@|aKnQ|>x#lnR+^yk+ZzM8+gvSgJ+^_T@6LdWv
zJb^W6a(AH4g}~pii`;JKhEmj5LvWcFQC~ML$4M+dv)mW3UOi-PxFW|2ApOR}hGNX5
zDERa%AV$WtmF|`ygJu6L**^5#Ex0XRv@=7xMN2>*^I9*a^acQz>D4RU>S*Bg+wMp*
z5@^EadaNSbNjptDidWne#p7{^)C%T{wZh!USU|>*07lq1uD&EQQjt7BPHO-RFYyfT
zOP%`N@5Sq@UtcGhMhWx$5UR2miUk5;UA0$r#%DlR4*&@mrip^LNtPKNGKL--cr&(*
z2Z1%k)%GhAPu9TEW<ca>4C5CE?JOGH)$*@^pc)6Drv`wEkjnTJu7l6_JWF0q9?Skn
z>%k<p(JLJx15_cTvCyU*N|)WqK9)-mZ}rdC$BXh)pdW%*Z(P%&7!Z*ng-+%8m;K<^
z)7=(hTn}dZZuiQKSQh6>+SON90a+w|;;pEGiog!7sJ7o|u!JZ8=omHP5!ko9AE#R%
zH-7vv4nPmQ1+NO{FLo^T#?W}*X;+$sMkVH#%g;s4du*k*2#?ilyNO&c!yEBdL+BRo
zM4xU(P60WrpUiE31qqF@!{_dxK-G&xX~J-u7qHmk3L_>KRrMuT$yVpa%8gn}+!<@3
zN@#HV0C{b>2P_Gf_o^rw6w0YweIF&O8D{b8#u=EPD@A2fA@7Z!ots{NzrA}-70gxe
z!MX_JpCxr>3I=`4d0CddnHQ+|W}73KnRuHrwO-_hrb&2)I0OTUuLk-r&qhSAnjQ}9
ze80A*A8&M|E#n69XAgy_;`c`M-x}fmO5X>}K#@osv7BXPo*7!eSDBM?_=VUs0w~ao
zDlIq+y|<S0C1Bj&O#CAIfEW?he%^_U_GnlRM3`ks9Rd-j>vM4&z4;fuSgF^bvemVP
zL=bZ`qpStqI=0(e{MOLAxB|kAze?^@2uJFoT!JsZeWh1P+s<?e6DwmRDloPLN^nwF
znwg`<tG@yRb!2NRaO^a;zkwIXK0bx3Ip!X73Gs_hi*5xDZ}ZQ-vH4k6O*IR%;ZUWj
zN@V1Dz)g`)=nJ6a$JvPLq);3?PE`aqIG??eDwS^JZ>sjGE%Y{`5>3(b)}dQx027P&
z+<JeiNU@2#$f&dwa(zZjRn%-hGSqERv*hC$k?8J>k}wi=sgcx+j(hzmai={im$vGT
z|6QcF2OY&j|AyzU4v~||D&#~zq|s$F`EK7$WGUsKB64G9H+u%C;(fr<%BljsyduQ=
zMH;M%oqv&1??;GiN{jLyH^VbL{x_~kJJvfut-Uwr(;X}URiAW|+|l?Zr;2?n1B#*G
zR|wv14W~cQ<;so%+@Wcw1Bif<b-kdF`C8t5sY}5yl!=4p#kJv&I;bwU2dHCxfB|3L
zg3YiVio@51oUe1WJNtsB5z|IJl!)w0%xN?LKnxuuGASG|I`}QS;he$&o-7TdKgnAS
zvaL?dJEO3c^y6y`o~@%*1l6aO4FwkR!$3od&XUk~#2MW{T}bY;M0A^+39&~hr}!BO
zr+6KBg(n0-UF(UuqBB95X7cdKOM~GiGYgiMzk^NZA3Z+)qBH&MnNJ(I8Q`V?GiM?y
z+PJA40u(TeFj$wL_lz32Eo@jyCkHvgmD_zi+47dUN}4o0H}n?f0g`=aiw>;+Y5eW4
z$p=mW3N57{TOjV^b*=#H@V^Zbi(o`o8rd6(2$ucw1kCPL8KYdQ{Oy#=uEnK@;jbip
zr&sJt;$PM}`a-@~flTyY=(S_|Er1eRwMZT@X{qpd_}y|R7jY@mET%Cycsk?*OW8l)
zl6&s21bM6$8n4RSSCQ|`)t2|*wuW-$<tqt)I{%VG6J64LIU5t*fxe4Q_tO6&E8u0=
z>eO$XK}*W65GUS-0daNECaE~4G<p|ITsEJ{B7zH0{&NR4=8b&CxhWvu0$c5&^qZX^
zEbdgoAWJHjc8$dUI3H+tV=IhSP|h*D5hD)>pUUK25IWCi#EVD|e4m#J5chY`lq+6Z
z?w_GsMBPkKr$N`Yjv+n|3fOozFGb9LN-`<X(Cwrw{b5|r27NQ&-V9>|FUv-MjlE5E
zpU?B5B@f|=^g!b>>y6&#p@%Z2p4TR;9n{R(K6=;P?iT0rrL9}NCQiAG$i#;PTYx_q
zMI}qDl3W3B_anGjK=Y@FL<PHK(6GqeK6{<S)<v@jGeV*?55g`SoRS#soKWA&W`Ptx
zxzpgAF;Wmci3%vRY=QAhR1cSroUPgLHy`76tb$py4BXRA3q?SAU9cle`&EJzmlMt2
z6=s@ym2TjO=!BD@c1Cvyxy??&S^2oTE|XS2sA%ijvqi?6?jYgM8#IPc$lwl+d!t0R
z8wLqYd?WO^cq)`+VT5mka?K0Ii*OSUbJJMm_%dJq<GVJ_?4r=rT&%;uyl|X0Z?CfS
zHgfalI*jnMZ^!xO)ozDpA|D#0oK*@%(ma#n_*-McyImr~L`u01+L3ou1%}<azOfpQ
z%QQ@SCKpthT;*%buWjw0R2>$qAa{Q_YdT6c^_vsEn#VbbJ~zB;kN-nKuk|qpQ8O7g
zDBn7H&kyzhRK+`+(BdnUBxOrKR<INQHE&KtX-{ZzIuK)hSGJvR@~)^LV|TCXl_u8W
z=~l81L~waKz8CF)C(r93sLob~d(Erf>nVH3<PkSkt>pd=Ml^XHYvkrAP*vf!N$8CB
zl{VF`<izSXl;Crs`a+?M-A~@@?VSK+yN&d7V`r=xUmpC*af|vQ5b$|1q|(x8KA!(@
zj0`3c8CLkI3bD_1$S~qXDX^B2ML<Moj>gQ{9+mX`aHyuDK{)@#^jc0PXl%4Q8Dz`V
zSI|}UM>xP?9AgQgmtR!&#JkP+Q9=aIcMG&1B!VD%-OiTcWL-9~j+bwQvITIbuRSs^
zNbm~NI*KhbYkHe|D13$5gRHC{(>aTx)@0PZNqH4<@rX?PTr(!8xMQajwOefyUS(=B
zlID=c7LdT6N998Z7{3TfHf9|#WfNjQ0(+NLv}pTllkb#mA;2>9WzS-a`bCHkW+6m0
zZNgO{9h~n`j&;f}%b0IH>2Q6R6aHE7>YB{;kI5sk`74!4VTXBZ>3dCQ14E<SnEpLk
zu9xrxj6c_9Ny6edmCwBzW^l^)3tNd@dhaOSc^uG%_tby0x!mc#>trYAxti4ka$LlV
z%rj@F2+k-Py|>E`Ijx9mr;9(mgqZ%g{1B<O{?AHe{hV>>wODD@@)BC(M6X9+644E}
zK2xQML=pU{3<1klfh+FpLOF>n7uQ5%y|fe5kMHi5p320DEDhIXY1cT>i2Y*}DvZ~8
zY4PsmGASgi#^1gmveaCgK$ew~#|Zy@f-P%%+J7O*C?BWQYgv@mxM9v|3nh3BTFpa4
z<a;6E73gG=BY|0m&EU+nrBhm<PtyyU5emUBlrq4<&osy>OMc@wUX8kXa@b|r1T1iG
zGFzC1I#H2L{F~ZUPq>62#30teKW#J;Sf7oP6T4ZuXlUvtm9NQ*D#<8>#2_75c>X$g
zeCf^Ol@n~qT@v*ucPPyxZ;Xs2)4U!)4su3#%XvhIt6AVti+F-a#-4coX-@frpz2j;
zAX#zeT>Um`X^Xfkt!cjBR=!q|twad6dGumMv+MLvy+&MrSCLT0D#p~$+b>BY&$+i{
zUIYbA$SBHW4EGk_c7QshjkO?O{soWHOUIbgsN=l6&sv1tVS-+?PANoGe%auxKt`-!
zFhTH6vwlzJ0*K?rm@Oyovn^r%)p>8Ph5^spamQfUKoSH7CE-|*8hQoVP@MKdkY#c`
za;VkrYl0t+=#ctm_8i7!5N0=PU~DQe<eTCgoAyCy_2hi`xpQLwjr5089a$Tr>NilJ
zspgLh!#Il7f4^|N({mRhng>gyAetf^>gTm<&C^c2vf9f;pL>@KWHN%UZ*H+x`L=8_
z2`p6VLC|<t(lU!YPO5q(!`NBHcemfYdMsJn>$Wm|TeePeqeM^w6fc{Aris3Q%|GXC
zs;Wn1#O*wFALsj2Vs(+dQJD*nKQkI|tU|z?6k|sr+2`O{!(w#STn^_j(2rc%s9~m2
z^A51ERx%+aE&KOy$E;Q`rLkGeORs`-*M9cj!#$aMqh8X2b}~fV=)$K>*aL&H6t1VB
zl8++oz3$Q=SZs2z?!1gzM(t^!>A1{xY4~{Csee2jZQ$=wgaD3HRc5=JvE!RvK3PIp
z`<}GOF<pa?Q>A8coPw*W6b|ATOE21qPG}6Qf$K)4+g~W+#l=R*&6W)LM1}r~lXe!}
zrAv+*r799L#D-P5u?-21JWVanPB1wynr)L?p&AK6!U77Wt@f_RbLI*Y&0j%&oNsyM
zV@9SXjUavcb*|HrPG5pR0QH4W)5LPuJ}hCbHj~NlR@x^{+c6~3kHen%oS%q8Xp&E@
z*}TTjxgob$vm`>%(*Pk$`@OVnO?%I-CqX(3bMSKGOA7NY_%5&dAjoxCiP+}xxe#Dy
zl>#evZmoR{<QpE{LBpvARcU%^LfV~>ys6i$Dwdc+(jH3!+pUC~lXiFsVYQY`$&#Wr
z@&F*j`kA#+hL6u|QH$`cFp|2?$mf=OrEOndn+#(T=we#SeZAwZH-c|s%3K+<`I7Jg
zmaO7bQAp-wyDnj}gi1+r<YtsPn|%c-`-<;<C0JGBPY{wa(qg+#RNgt5crsFC$ES=P
zHyg}OM{YyQHDY{DV%V;FslTJ4RhNpyL{A#>VsDROV*n}B9b`oHX7zh}bH401knAIS
z<ESTTQ2;nbsN@&$-2!O+&))?we$JkRvV^qSolsGI@Oi6U2uod+I04hX<1UTUu~Z_b
zTd^Z9)fDn69(JXQ4U01v&F}*e|M#QlL{6`x`nOL6k&8GOUiC;a^^D3!arqQuDYMa*
z!WeTDLY1%KAF39&gw*#ZY6x)OD+kcnFqe@Uxz3pDIJ9vaWyYfx&*GNRo15J5Te`wB
zn|ViPFnc=0$s96mtcAL}{1m~37koBgIN;DhuDJ#87Eis)aEe{l5m;N(C=?~>W2$2C
z8II+W*bHb}f<4~dr@nOhoL(_VJ)pYDtfotz-oE0-M$T^R$7MUucVKG1I>x4bS@Ski
zWW2&8qGVVvxF8kHo@Zt(w)o-hs@$>BF}K^NUNf=wH=V&xuWF-XH@G^W`d;%34%<NS
z$@z95YO6Y(;7;(~&YpK9WHyM?==dO7DDD=4-Iz$r(vA&qxXu-Of;|GaHcr+X$ew(#
zkuugcFeaEToVtD?{~p8c)1}`f0&sD4k-kg8iPc=pr@W>2uVMQM>~bNqsOFYT?+cXf
zcoHV@<wE#6p}x-jnu`zUA4xR3mKlt^al?>`qa>&xJA`o1AhAAXbm)sJ2djSwW6y!v
zcs4j5FZ~7=%e22E*MxA=;(+{?(`KFWZ`}*})sG1ss^|89NoZhP1SJR~%3xyXzqB=Q
z0}#Ch|398YELWmxSa<HPczz(L^Xde2H1jOpP5K~M=VxC|$2k1ESYnb9(B&o>?-T#6
z@qzEK*rER)-@={4WyX<0{K{F`tmS=ISkp}YFFIv`RMpgTF55EG*?W{2EjIFxdhS$j
zm~d?&9tr{Gv(>;?Qe@EcGpQ4)nQNlbK)b5IB?KI;TrNpGtzHLoHUdHIMP>P>d{e(u
ze>xJuaYu=%CIkb7UW2W1pF;;QCZBTym4JiLtM*n{<BU-q2kALRoaWF?R0#wG6oIU`
zdGp0oT$16ZA|+d=t>FYfNzt7BX~pnxdl3iFI^Fh2-?1EN$Fm{c0RV=tO~5wBsR+pD
zzs<}@z#mw61mHv>OL^0~XXPu(MKds-0E9Aa=vlf2^}f4woK%(V0}|s^8Eoo?I-ygm
zUr*M02I!LXR~?{iMFl|MSbS9xZR8v8$TYu;ls7p*);Y}k<4a=4HlI$DTIGPMH7HPU
zciJQRsM-Pq6HEA(o9ApB6K;S@)>Cb&K7eb98>ij}guSrw+yiSsBvRG$286C$&rcGK
zO#sg@1AM_mi=k8xQXmxD#@Yq5_*}+x)B1ELx^%hZb*AhsB;W(2mqUfReuR-T{)hus
z=Z3f6Nfuq85>uJ^eKC)rO-t@;5$u1i`fg?|XIow=d-AUY;^c_LUTr>Cr2%1FvlI2{
zIYmvAf%>eG0-MpRP6!a`Z4ueYbQcAJZZqz<uzT*FjX>36f9HgszKzt86Khlfd&YCk
zg4hlXSRJ|oI6OE4j3S?buz_eZrm*Y&pK;kYvdHS<SpmL4)gctHxGRCQ>vhymk@oF5
z07g;0Zn@480$dkIK!xm|vCMo$pl|OlgM}~#h=+Y;jfcAlpPywZ@*bK<KI*WJzR4mr
z^KUjaCp$nid`>4!R<l!xfYB9D#qzgSLK!kJ$7caqsS06TVk03WLD$K1y*K6*a3lNs
z7&2Lb818p#puru16qX09fEhO3N_Gf&ERf+|oW40;>3Bf$0?JBNs{sd3Pt(ZT#RF*1
zjYz>mOay9?Q$S-2Tg}33I&McbWTyaD%I$KEVolYw>Vq7h+Ik!?<x40ws2=DP6R*#o
zv*1gvK!j#{Kvr*}Qi`Y{4p7P#P{~6`9ltaI0*&M2&6Z*>j)SVmB@h66)>$`z%q=oS
z)G))_jQz8txmr7Qo_?T&R<$wjOQu9-p3Q4xf^rgE$pgXV;^>QljWm9z{3Evy@RQPV
zklBJ$EyE?5RA6EBrVOBDOnR1#^?^)}&ob?LQCV$LMM?cI(IEd}-I`Y`HicO-@U1f?
zt&g{r;|eT&z|9}tS>cXr8G6vM@ZYjkOg{dv0QLn5`h;VrPMxyi=tX&LMB|jllg6>5
zAz%p9Aq3n@pP4XW0*8Xo(xpo?Nk2DYflX4z2}z0f!)F#OSdej}7v;utLThIJ{P~G-
z=FFLs;TNV^6X#Cw%i=*XIOm*mqHx!VdtBTHGYWH~{65T$R{Ag}*k$;IWpEz^9|s&X
zoczFK@#4k)-g)Pp7v0l1`Hqt&P4dq=#UXn3{rBJZVQ%y|3M)ij`FN>K_v3ZxW`{B@
zTC~Wx*`o*(C;WSy?z-!)DD35zU-n`5-FIIU7O%^jZo0|Gae`PB$IZ(5FeiXVVR1at
zwu|_4w2`&eS}U{n-g`%B@4ox)zC2F3iNc&Blj+*EYliocG`{}&YtPf|x8EMciKorA
zQ)Ya7)pi30@pw*&_G#sMhvV@}?K(|+Nh;Im(W5;sZg8axW*Wb;^{3C%ZuZ-6zl;+Y
zeO_VPBu<+4882T_-<Gz$C|`P{Z91hotNeAny=d9XJM!M~4Rtg188-lzg^yF-{k`zJ
z=?19$dmh#&b$-PaS7be#lxMy+qR*7alg0^;1rahV5ktTbs4)bJHmujZdv~vDVYc9e
zkVLi1eTckihZDx6W5<p%bm&lFnqSEY!*bPCS9#N>&Ye5^5jywr)7vs)#0YO*(WXrs
z;fJZk7hil4-7oeDqdp<Q6BLfjGP?k6w9!WX$RB1~kQ282u|b0d!Y?Ep9ctXTaTH8`
zn7dSk7^Q-?-+ucjnRVA)SEv(4#+3?@m!?$;$9P?00>I1><`hcbfB(Hd9WrEy2FZpS
zZs^aMNm7c}B_?1H<~|jnwCy6E=AC!m>1RzC(N$tZl3B2D$5zZ37?u3thaY@2ZL7?T
zl}BI`FJpP}>*BpM?Io$)jLb9Zbm*anCf)X7m90Nb@c#Sn7fjC{fBbO`$~wmDE~$OS
z%U53Cman~pnPZ;DjNT@jY!U|7`}429`pTP$GUH79h!ZA*d+oKC|L*BMa0g4wmMTj5
z+6d|8<Ef%~T6{ym5UBkK6isPxBbI`B($h~r?LP>ZWW?#Bi!PGuuDi~gE+H^ux^(>U
z$4AjHYr;+%(}>SM|J<8F=*Ly3XHiHn1>(o~1{-W3ZYCv4qb5((MjT~G>Bk>`^rkAI
z@UIG1F4HYH8z2un@PN3{VXru8T7@`{8#m4;v+~L-N68@|xY0jRp*Xrxk54eAK%mBq
zM@a>z&~!k39)AA$XN{J1bfa*BIS3Rg8ye}t3orC$VwgGQqcrV;`aSW)69vIj+ZpNh
z>C;Cri*dsjQ8vmu<%Fnzl*bQ0{LuS$pfF)yO&hK(9dpbvejvb&uva#Tl6Wquy~G)3
zK!twGjoxM{Z(+Qu)F0`5_uY5CA9k~0GI;RdFzLkeeC<;oO6uFvwih?5?foiHVtJo7
z+iWwT-6v6fX)j^fF++5-f^yJ72MPW|abmEIceH>1{#m#19facFzkBb!*MCPD1OWRu
zrD-Gic#5ZK*M@*0P&W`Ln$iGV_%mR$kDW6g1{453N)XPIIPhY#x#ymHdYf?^aB=3s
z)|>m-+9R~!D7?lRYk1{jyY05~&);*;J^n}@jF|Fc^aWycrl#qcXP)u)-`Kk25Us>0
zP`7T~^wJ-(afT2A<D{K6GclC4(1h{@DG!=HGdsBiUyV~=jtI$&;_)HJVN{7(1==U|
z!iSep(~B>@*x#NuZJI`ljtJ=Zr{IYC&O7gT+jYH3I+tI5xu2<EmSU@|wh{!UP^j0q
z9MiT79P_bb#^)d;1QG-flnH14A}A@?mqT1b)IW;&DW{y`{T-O~!>@t%jQw*(2p%Y3
zhYlUQQ|F*TgXF;nAFN0cB@;JSdkG8R%qiu_Baig`UwOiORqBtrL&;-ymio>|<+ac7
z*<5{F+V(=(4?p~H?*NS-%%A@Br>xmn@~ThmCB&9<zLU(dGoY44>NL*63nA%@HvVTy
zr=EJM#}_4?K7c_r@K;LPM)L8b>Dkc`Fa+ur0{;)5hgPZMIoN;z0000<MNUMnLSTY1
Cdj`D#

diff --git a/docs/index.html b/docs/index.html
index 6341943..04f5501 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -516,7 +516,7 @@ <h4 class="anchored" data-anchor-id="example-code">Example code</h4>
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-1" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="1" data-code-cell="annotated-cell-1">Load the tidyverse library</span>
+<span data-code-cell="annotated-cell-1" data-code-annotation="1" data-code-lines="1">Load the tidyverse library</span>
 </dd>
 </dl>
 </div>
diff --git a/docs/quickstart.html b/docs/quickstart.html
index b83dbb0..e2bacc7 100644
--- a/docs/quickstart.html
+++ b/docs/quickstart.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/r-data-import.html b/docs/r-data-import.html
index 1db0f17..764677a 100644
--- a/docs/r-data-import.html
+++ b/docs/r-data-import.html
@@ -272,35 +272,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -317,37 +323,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -364,25 +370,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -604,19 +610,19 @@ <h2 data-number="10.3" class="anchored" data-anchor-id="import-the-hot-100-data-
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-2" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-cell="annotated-cell-2" data-code-annotation="1" data-code-lines="1">Create a new object in R called <code>top100</code> using the assignment operator <code>&lt;-</code></span>
+<span data-code-annotation="1" data-code-lines="1" data-code-cell="annotated-cell-2">Create a new object in R called <code>top100</code> using the assignment operator <code>&lt;-</code></span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-cell="annotated-cell-2" data-code-annotation="2" data-code-lines="2">Start the function <code>read_csv()</code>…</span>
+<span data-code-annotation="2" data-code-lines="2" data-code-cell="annotated-cell-2">Start the function <code>read_csv()</code>…</span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-cell="annotated-cell-2" data-code-annotation="3" data-code-lines="3">to access a comma-separated data file stored in Prof.McDonald’s Github site. Use just the name of the file, in quotes, if it’s already saved in your projects folder.</span>
+<span data-code-annotation="3" data-code-lines="3" data-code-cell="annotated-cell-2">to access a comma-separated data file stored in Prof.McDonald’s Github site. Use just the name of the file, in quotes, if it’s already saved in your projects folder.</span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="4">4</dt>
 <dd>
-<span data-code-cell="annotated-cell-2" data-code-annotation="4" data-code-lines="4">Don’t forget to close your parentheses.</span>
+<span data-code-annotation="4" data-code-lines="4" data-code-cell="annotated-cell-2">Don’t forget to close your parentheses.</span>
 </dd>
 </dl>
 </div>
@@ -775,15 +781,15 @@ <h2 data-number="10.5" class="anchored" data-anchor-id="introducing-the-pipe"><s
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-6" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-cell="annotated-cell-6" data-code-annotation="1" data-code-lines="1">Start with the <code>top100</code> data frame that we saved earlier, and then</span>
+<span data-code-annotation="1" data-code-lines="1" data-code-cell="annotated-cell-6">Start with the <code>top100</code> data frame that we saved earlier, and then</span>
 </dd>
 <dt data-target-cell="annotated-cell-6" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-cell="annotated-cell-6" data-code-annotation="2" data-code-lines="2">Use a function called <code>clean_names()</code> to convert them to computer-friendly names that won’t require special handling, and then</span>
+<span data-code-annotation="2" data-code-lines="2" data-code-cell="annotated-cell-6">Use a function called <code>clean_names()</code> to convert them to computer-friendly names that won’t require special handling, and then</span>
 </dd>
 <dt data-target-cell="annotated-cell-6" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-cell="annotated-cell-6" data-code-annotation="3" data-code-lines="3">Take a look at it!</span>
+<span data-code-annotation="3" data-code-lines="3" data-code-cell="annotated-cell-6">Take a look at it!</span>
 </dd>
 </dl>
 </div>
@@ -836,15 +842,15 @@ <h2 data-number="10.6" class="anchored" data-anchor-id="cleaning-up-and-save-for
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-7" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-cell="annotated-cell-7" data-code-annotation="1" data-code-lines="1">Create a new data frame using the assignment operator out of what follows. Leave off this row until you know the rest works.</span>
+<span data-code-annotation="1" data-code-lines="1" data-code-cell="annotated-cell-7">Create a new data frame using the assignment operator out of what follows. Leave off this row until you know the rest works.</span>
 </dd>
 <dt data-target-cell="annotated-cell-7" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-cell="annotated-cell-7" data-code-annotation="2" data-code-lines="4">Use the function <code>mdy</code>, which converts text in the form of month-day-year into true dates, to transform the chart week into a date variable called <code>chart_date</code>.</span>
+<span data-code-annotation="2" data-code-lines="4" data-code-cell="annotated-cell-7">Use the function <code>mdy</code>, which converts text in the form of month-day-year into true dates, to transform the chart week into a date variable called <code>chart_date</code>.</span>
 </dd>
 <dt data-target-cell="annotated-cell-7" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-cell="annotated-cell-7" data-code-annotation="3" data-code-lines="5,6">Pick out the columns to save in the order you want to see them.</span>
+<span data-code-annotation="3" data-code-lines="5,6" data-code-cell="annotated-cell-7">Pick out the columns to save in the order you want to see them.</span>
 </dd>
 </dl>
 </div>
diff --git a/docs/r-install.html b/docs/r-install.html
index 593e22f..bb8e759 100644
--- a/docs/r-install.html
+++ b/docs/r-install.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/r-quarto.html b/docs/r-quarto.html
index 774166b..8f0c66c 100644
--- a/docs/r-quarto.html
+++ b/docs/r-quarto.html
@@ -272,35 +272,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -317,37 +323,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -364,25 +370,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -567,7 +573,7 @@ <h3 class="anchored" data-anchor-id="including-r-code-and-its-results">Including
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-3" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="1,3" data-code-cell="annotated-cell-3">Print the value that’s stored in the variable called <code>my_name</code>, then print the value that’s stored in <code>my_variable</code></span>
+<span data-code-lines="1,3" data-code-annotation="1" data-code-cell="annotated-cell-3">Print the value that’s stored in the variable called <code>my_name</code>, then print the value that’s stored in <code>my_variable</code></span>
 </dd>
 </dl>
 </div>
diff --git a/docs/r-recipes.html b/docs/r-recipes.html
index 268254a..beb5c7a 100644
--- a/docs/r-recipes.html
+++ b/docs/r-recipes.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 20&nbsp; Recipes</title>
+<title>Data Reporting - 21&nbsp; Recipes</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -113,7 +113,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./r-recipes.html"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./rdepth.html">Verbs of the tidyverse</a></li><li class="breadcrumb-item"><a href="./r-recipes.html"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -270,35 +270,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -315,37 +321,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -362,25 +368,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -425,14 +431,14 @@ <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
   <li><a href="#r-annoyances-and-errors" id="toc-r-annoyances-and-errors" class="nav-link active" data-scroll-target="#r-annoyances-and-errors">R annoyances and errors</a></li>
-  <li><a href="#re-using-answers-from-a-previous-step" id="toc-re-using-answers-from-a-previous-step" class="nav-link" data-scroll-target="#re-using-answers-from-a-previous-step"><span class="header-section-number">20.1</span> Re-using answers from a previous step</a></li>
-  <li><a href="#importing-reading-data" id="toc-importing-reading-data" class="nav-link" data-scroll-target="#importing-reading-data"><span class="header-section-number">20.2</span> Importing / reading data</a></li>
-  <li><a href="#filtering" id="toc-filtering" class="nav-link" data-scroll-target="#filtering"><span class="header-section-number">20.3</span> Filtering</a></li>
-  <li><a href="#aggregate-count-sum-rows-by-category" id="toc-aggregate-count-sum-rows-by-category" class="nav-link" data-scroll-target="#aggregate-count-sum-rows-by-category"><span class="header-section-number">20.4</span> Aggregate (count, sum rows by category)</a></li>
-  <li><a href="#recoding-categories" id="toc-recoding-categories" class="nav-link" data-scroll-target="#recoding-categories"><span class="header-section-number">20.5</span> Recoding categories</a></li>
-  <li><a href="#working-with-grouped-data-for-subtotals-changes-percents" id="toc-working-with-grouped-data-for-subtotals-changes-percents" class="nav-link" data-scroll-target="#working-with-grouped-data-for-subtotals-changes-percents"><span class="header-section-number">20.6</span> Working with grouped data for subtotals, changes, percents</a></li>
-  <li><a href="#joining-tables" id="toc-joining-tables" class="nav-link" data-scroll-target="#joining-tables"><span class="header-section-number">20.7</span> Joining tables</a></li>
-  <li><a href="#quarto-document-tips" id="toc-quarto-document-tips" class="nav-link" data-scroll-target="#quarto-document-tips"><span class="header-section-number">20.8</span> Quarto document tips</a>
+  <li><a href="#re-using-answers-from-a-previous-step" id="toc-re-using-answers-from-a-previous-step" class="nav-link" data-scroll-target="#re-using-answers-from-a-previous-step"><span class="header-section-number">21.1</span> Re-using answers from a previous step</a></li>
+  <li><a href="#importing-reading-data" id="toc-importing-reading-data" class="nav-link" data-scroll-target="#importing-reading-data"><span class="header-section-number">21.2</span> Importing / reading data</a></li>
+  <li><a href="#filtering" id="toc-filtering" class="nav-link" data-scroll-target="#filtering"><span class="header-section-number">21.3</span> Filtering</a></li>
+  <li><a href="#aggregate-count-sum-rows-by-category" id="toc-aggregate-count-sum-rows-by-category" class="nav-link" data-scroll-target="#aggregate-count-sum-rows-by-category"><span class="header-section-number">21.4</span> Aggregate (count, sum rows by category)</a></li>
+  <li><a href="#recoding-categories" id="toc-recoding-categories" class="nav-link" data-scroll-target="#recoding-categories"><span class="header-section-number">21.5</span> Recoding categories</a></li>
+  <li><a href="#working-with-grouped-data-for-subtotals-changes-percents" id="toc-working-with-grouped-data-for-subtotals-changes-percents" class="nav-link" data-scroll-target="#working-with-grouped-data-for-subtotals-changes-percents"><span class="header-section-number">21.6</span> Working with grouped data for subtotals, changes, percents</a></li>
+  <li><a href="#joining-tables" id="toc-joining-tables" class="nav-link" data-scroll-target="#joining-tables"><span class="header-section-number">21.7</span> Joining tables</a></li>
+  <li><a href="#quarto-document-tips" id="toc-quarto-document-tips" class="nav-link" data-scroll-target="#quarto-document-tips"><span class="header-section-number">21.8</span> Quarto document tips</a>
   <ul class="collapse">
   <li><a href="#yaml-front-matter" id="toc-yaml-front-matter" class="nav-link" data-scroll-target="#yaml-front-matter">YAML (front matter)</a></li>
   <li><a href="#code-chunk-options" id="toc-code-chunk-options" class="nav-link" data-scroll-target="#code-chunk-options">Code chunk options</a></li>
@@ -446,7 +452,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="r-recipes" class="quarto-section-identifier"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></h1>
+<h1 class="title"><span id="r-recipes" class="quarto-section-identifier"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></h1>
 </div>
 
 
@@ -478,8 +484,8 @@ <h3 class="unnumbered anchored" data-anchor-id="r-annoyances-and-errors">R annoy
 </ul>
 <p>These are the problems that will fix about 80 percent of the code that won’t run once you get more used to it.</p>
 </section>
-<section id="re-using-answers-from-a-previous-step" class="level2" data-number="20.1">
-<h2 data-number="20.1" class="anchored" data-anchor-id="re-using-answers-from-a-previous-step"><span class="header-section-number">20.1</span> Re-using answers from a previous step</h2>
+<section id="re-using-answers-from-a-previous-step" class="level2" data-number="21.1">
+<h2 data-number="21.1" class="anchored" data-anchor-id="re-using-answers-from-a-previous-step"><span class="header-section-number">21.1</span> Re-using answers from a previous step</h2>
 <p>Assign your code chunk to a new variable in the environment to use it again later. This is usefule when:</p>
 <ul>
 <li>You have a complex filter that you want to apply to future steps.</li>
@@ -513,8 +519,8 @@ <h2 data-number="20.1" class="anchored" data-anchor-id="re-using-answers-from-a-
 <span id="cb1-22"><a href="#cb1-22" aria-hidden="true" tabindex="-1"></a>  <span class="fu">head</span>(<span class="dv">5</span>)</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
 </section>
-<section id="importing-reading-data" class="level2" data-number="20.2">
-<h2 data-number="20.2" class="anchored" data-anchor-id="importing-reading-data"><span class="header-section-number">20.2</span> Importing / reading data</h2>
+<section id="importing-reading-data" class="level2" data-number="21.2">
+<h2 data-number="21.2" class="anchored" data-anchor-id="importing-reading-data"><span class="header-section-number">21.2</span> Importing / reading data</h2>
 <ul>
 <li>If you are referring to a file on your computer or on the web, then it must be in quotes.</li>
 <li>Use the proper library and function to read the data.</li>
@@ -550,8 +556,8 @@ <h4 class="anchored" data-anchor-id="import-options">Import options</h4>
 <p>To keep everything as text, which you can convert later.</p>
 </section>
 </section>
-<section id="filtering" class="level2" data-number="20.3">
-<h2 data-number="20.3" class="anchored" data-anchor-id="filtering"><span class="header-section-number">20.3</span> Filtering</h2>
+<section id="filtering" class="level2" data-number="21.3">
+<h2 data-number="21.3" class="anchored" data-anchor-id="filtering"><span class="header-section-number">21.3</span> Filtering</h2>
 <ol type="1">
 <li>One condition that’s a number :</li>
 </ol>
@@ -587,8 +593,8 @@ <h2 data-number="20.3" class="anchored" data-anchor-id="filtering"><span class="
 <pre><code>    filter ( str_like  (borrower_type, "Corporation%"))</code></pre>
 <p>All of these examples can be used in a <code>mutate</code> statement to create flags or new values if the conditions are met.</p>
 </section>
-<section id="aggregate-count-sum-rows-by-category" class="level2" data-number="20.4">
-<h2 data-number="20.4" class="anchored" data-anchor-id="aggregate-count-sum-rows-by-category"><span class="header-section-number">20.4</span> Aggregate (count, sum rows by category)</h2>
+<section id="aggregate-count-sum-rows-by-category" class="level2" data-number="21.4">
+<h2 data-number="21.4" class="anchored" data-anchor-id="aggregate-count-sum-rows-by-category"><span class="header-section-number">21.4</span> Aggregate (count, sum rows by category)</h2>
 <ol type="1">
 <li>Counting (How many?)</li>
 </ol>
@@ -618,8 +624,8 @@ <h2 data-number="20.4" class="anchored" data-anchor-id="aggregate-count-sum-rows
 </ol>
 <pre><code>    count ( project_county, weight=amount, name="total_amount")</code></pre>
 </section>
-<section id="recoding-categories" class="level2" data-number="20.5">
-<h2 data-number="20.5" class="anchored" data-anchor-id="recoding-categories"><span class="header-section-number">20.5</span> Recoding categories</h2>
+<section id="recoding-categories" class="level2" data-number="21.5">
+<h2 data-number="21.5" class="anchored" data-anchor-id="recoding-categories"><span class="header-section-number">21.5</span> Recoding categories</h2>
 <p>You’ll often want fewer categories, or numbers in categories, that you want to use instead of the original values. This is done in a mutate statement. Don’t forget to save the output to a new data frame (&lt;-), or you won’t have access to it later on.</p>
 <ol type="1">
 <li>Create yes-no categories . This is really “Yes”, “No” or NA, where there is an NA to begin with.</li>
@@ -666,8 +672,8 @@ <h2 data-number="20.5" class="anchored" data-anchor-id="recoding-categories"><sp
 </div>
 <p>This works because the first one that it finds will be used, so a value of exactly 1,000 would be “Very low”, but a value of 1,001 would be “Low”.</p>
 </section>
-<section id="working-with-grouped-data-for-subtotals-changes-percents" class="level2" data-number="20.6">
-<h2 data-number="20.6" class="anchored" data-anchor-id="working-with-grouped-data-for-subtotals-changes-percents"><span class="header-section-number">20.6</span> Working with grouped data for subtotals, changes, percents</h2>
+<section id="working-with-grouped-data-for-subtotals-changes-percents" class="level2" data-number="21.6">
+<h2 data-number="21.6" class="anchored" data-anchor-id="working-with-grouped-data-for-subtotals-changes-percents"><span class="header-section-number">21.6</span> Working with grouped data for subtotals, changes, percents</h2>
 <ol type="1">
 <li>Percent of total by group</li>
 </ol>
@@ -702,8 +708,8 @@ <h2 data-number="20.6" class="anchored" data-anchor-id="working-with-grouped-dat
 </ol>
 <p>NOTE: Don’t use <code>slice_max()</code>, because it will give you back multiple rows in the event of ties.</p>
 </section>
-<section id="joining-tables" class="level2" data-number="20.7">
-<h2 data-number="20.7" class="anchored" data-anchor-id="joining-tables"><span class="header-section-number">20.7</span> Joining tables</h2>
+<section id="joining-tables" class="level2" data-number="21.7">
+<h2 data-number="21.7" class="anchored" data-anchor-id="joining-tables"><span class="header-section-number">21.7</span> Joining tables</h2>
 <p>New options in the <code>dplyr</code> library that contains the <code>join</code> verb let us use the same operators in joining that you use in filtering: <code>==</code>, <code>&lt;=</code>, etc. Usually you will use the “==” operator. Requires dplyr 1.1, which was released in February 2023.</p>
 <pre><code>ppp_orig |&gt; 
   select ( borrower_name, amount, naics_code) |&gt; 
@@ -711,8 +717,8 @@ <h2 data-number="20.7" class="anchored" data-anchor-id="joining-tables"><span cl
                join_by (naics_code == naics_code) 
   )</code></pre>
 </section>
-<section id="quarto-document-tips" class="level2" data-number="20.8">
-<h2 data-number="20.8" class="anchored" data-anchor-id="quarto-document-tips"><span class="header-section-number">20.8</span> Quarto document tips</h2>
+<section id="quarto-document-tips" class="level2" data-number="21.8">
+<h2 data-number="21.8" class="anchored" data-anchor-id="quarto-document-tips"><span class="header-section-number">21.8</span> Quarto document tips</h2>
 <section id="yaml-front-matter" class="level3 unnumbered">
 <h3 class="unnumbered anchored" data-anchor-id="yaml-front-matter">YAML (front matter)</h3>
 <p>The YAML is very picky about exact casing, spacing and other details. If you try to render and you get a <code>YAML error</code>, it is likely at the top.</p>
@@ -979,7 +985,7 @@ <h3 class="unnumbered anchored" data-anchor-id="nothing-runs-everything-is-gray"
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./advanced-regex-r.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
diff --git a/docs/r-start.html b/docs/r-start.html
index 98c60d3..dbc51b1 100644
--- a/docs/r-start.html
+++ b/docs/r-start.html
@@ -272,35 +272,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -317,37 +323,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -364,25 +370,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -508,7 +514,7 @@ <h3 class="unnumbered anchored" data-anchor-id="typing-into-the-console">Typing
 </div>
 </div>
 <div class="callout-body-container callout-body">
-<p>Use the keyboard shortcut <kbd data-windows="Alt–" data-mac="Option–"></kbd> to create the <code>&lt;-</code> assignment operator .</p>
+<p>Use the keyboard shortcut <kbd data-mac="Option–" data-windows="Alt–"></kbd> to create the <code>&lt;-</code> assignment operator .</p>
 </div>
 </div>
 <p>The console remembers your commands, but you have to type them one at a time and you can’t save them for later, or start over without entering them all again. That’s why we’re going to work in programs called Quarto documents most of the time.<a href="#fn1" class="footnote-ref" id="fnref1" role="doc-noteref"><sup>1</sup></a></p>
diff --git a/docs/r-verb-mutate.html b/docs/r-verb-mutate.html
index 2dd2dc2..1756167 100644
--- a/docs/r-verb-mutate.html
+++ b/docs/r-verb-mutate.html
@@ -530,15 +530,15 @@ <h2 data-number="15.2" class="anchored" data-anchor-id="math-on-columns"><span c
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-3" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-lines="2" data-code-cell="annotated-cell-3" data-code-annotation="1">Pick out a few columns to work with</span>
+<span data-code-cell="annotated-cell-3" data-code-annotation="1" data-code-lines="2">Pick out a few columns to work with</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-lines="4" data-code-cell="annotated-cell-3" data-code-annotation="2">Compute a new column using math</span>
+<span data-code-cell="annotated-cell-3" data-code-annotation="2" data-code-lines="4">Compute a new column using math</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-lines="6" data-code-cell="annotated-cell-3" data-code-annotation="3">Print out a random sample of 50 rows</span>
+<span data-code-cell="annotated-cell-3" data-code-annotation="3" data-code-lines="6">Print out a random sample of 50 rows</span>
 </dd>
 </dl>
 </div>
@@ -723,23 +723,23 @@ <h2 data-number="15.4" class="anchored" data-anchor-id="putting-it-all-together"
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-10" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-lines="6" data-code-cell="annotated-cell-10" data-code-annotation="1">The same code you used for non-profits, but with a ~ instead of a comma.</span>
+<span data-code-cell="annotated-cell-10" data-code-annotation="1" data-code-lines="6">The same code you used for non-profits, but with a ~ instead of a comma.</span>
 </dd>
 <dt data-target-cell="annotated-cell-10" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-lines="13" data-code-cell="annotated-cell-10" data-code-annotation="2">The same code you used for individuals</span>
+<span data-code-cell="annotated-cell-10" data-code-annotation="2" data-code-lines="13">The same code you used for individuals</span>
 </dd>
 <dt data-target-cell="annotated-cell-10" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-lines="16" data-code-cell="annotated-cell-10" data-code-annotation="3">A new type, for Tribal concerns,</span>
+<span data-code-cell="annotated-cell-10" data-code-annotation="3" data-code-lines="16">A new type, for Tribal concerns,</span>
 </dd>
 <dt data-target-cell="annotated-cell-10" data-target-annotation="4">4</dt>
 <dd>
-<span data-code-lines="20" data-code-cell="annotated-cell-10" data-code-annotation="4">and another new type for traditional companies, and</span>
+<span data-code-cell="annotated-cell-10" data-code-annotation="4" data-code-lines="20">and another new type for traditional companies, and</span>
 </dd>
 <dt data-target-cell="annotated-cell-10" data-target-annotation="5">5</dt>
 <dd>
-<span data-code-lines="22" data-code-cell="annotated-cell-10" data-code-annotation="5">finally, what you want it to say when none of the conditions are true.</span>
+<span data-code-cell="annotated-cell-10" data-code-annotation="5" data-code-lines="22">finally, what you want it to say when none of the conditions are true.</span>
 </dd>
 </dl>
 </div>
diff --git a/docs/r-verbs.html b/docs/r-verbs.html
index 5525bed..d94f13e 100644
--- a/docs/r-verbs.html
+++ b/docs/r-verbs.html
@@ -562,7 +562,7 @@ <h2 data-number="11.3" class="anchored" data-anchor-id="summary-statistics-in-r"
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-2" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-cell="annotated-cell-2" data-code-lines="2,3,4" data-code-annotation="1">Each row defines a made-up descriptive name for the new columns created from summary functions.</span>
+<span data-code-annotation="1" data-code-lines="2,3,4" data-code-cell="annotated-cell-2">Each row defines a made-up descriptive name for the new columns created from summary functions.</span>
 </dd>
 </dl>
 </div>
@@ -606,15 +606,15 @@ <h2 data-number="11.4" class="anchored" data-anchor-id="grouping-with-summary-st
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-3" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-cell="annotated-cell-3" data-code-lines="2" data-code-annotation="1">Collapse the data into one row for each performer.</span>
+<span data-code-annotation="1" data-code-lines="2" data-code-cell="annotated-cell-3">Collapse the data into one row for each performer.</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-cell="annotated-cell-3" data-code-lines="3" data-code-annotation="2">And fill it with the number of times that performer was on the list</span>
+<span data-code-annotation="2" data-code-lines="3" data-code-cell="annotated-cell-3">And fill it with the number of times that performer was on the list</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-cell="annotated-cell-3" data-code-lines="4" data-code-annotation="3">Sort it with the most popular performer first.</span>
+<span data-code-annotation="3" data-code-lines="4" data-code-cell="annotated-cell-3">Sort it with the most popular performer first.</span>
 </dd>
 </dl>
 </div>
diff --git a/docs/r-verbs_files/figure-html/hot-100-top-10-list-1.png b/docs/r-verbs_files/figure-html/hot-100-top-10-list-1.png
new file mode 100644
index 0000000000000000000000000000000000000000..55a973b16783454f4f7ed8d1bdb1e2d141a05840
GIT binary patch
literal 173820
zcmeFZWmME{+cydb0wN{dB_Pt>4I(m>h=7zdNJ$SZ(j`hah%!ib4<ImrbV@f;(%tZ$
z-0$<;_jUQ~Yp?xbueJAwcdbE&0cZYmp2vB_FD6t?MFIOB`8^~gBy7c}a?g>F(4>%%
z?tI6%3m!Q%AD%!$LdCI?l~uErRgkr}v3JyXYh+@kU}k6LXl3$TK?Vs)C^Ax0-;z?3
zP$H$aob_R!{|F2gjdRT(TVeBjFko`++N|7mnk~krLi%Ip=1iIu6YtXHS1T3?LTS9_
zZ}TzpH@eNtbZ_#v!r8Mb)Ur_hADs@ID|{Xz!Q80jEw7KMv2N&|K3hrM@mZa9+MdPw
zSi`0pKg)>1Mqb3eJiJ)mv4TK|I5&F_V0#a+vwnzs9LI!AvDDHuxwS*vK09_?eR5;5
zQKY7T95NF)ULe#(zi5VZP1IJV{|;69;s@rK@hK)k)Fzd7@j+`Pk``$??rvFs(kGq}
zLwUAGc@fu$sSxy4hT~VL=uW#+AH?{-6|&HG%XbRoC#x_FH~3%pR>{v;Bc}XSqsqCw
z3Fq$A949<>j9&=G%h3Ea?Hp3fVO8FScVkv-g}KLoq+bsYbiB7GfiThZgj23L=jI6x
zerL@YNwUD3^M|@Li*>bc%UxD-8>ZfK@p+zle+8EsX@2oBdx^&h19geA)#H%vNG(f&
zj{T6+niXPm6z^?z3oDVa0pV=(Fln~!@TGnHmpKv$t=#kknJYu~YQlC#@gRfL`_4bO
zTrQ8!UA5AZFczbQNwCNx7&%{7?#cY{)v9Fvif35m>>IT>{X=;wrVBZvDT91jErup$
zOHTqplo;75)g0iyNyPP<K<*<gu9cVW>n}O?$k!}CuVq_)KK(hrIoHB+c8yg1;?4Ds
zjkQ588Q2ZO?m%A4j-@68{JxL7WxmS*rukcj5pG3J#c2OiO8S^zTx4`#Xrq}sHB2K+
z2SuLY@OQ-1JDDET4vx4ZGd#}_42OrWxk<NQTbL%!SM;|%%MipH?;r2C?ur|;GZ68%
z(+aPD@-m}-)M?%Mp#}LHcIO|HBWaHPH!X&%V<m>Gg0Jphqx;<P-I#h=&A-9d^YWf|
zoxLtX;Xqu)R3e*Q)5fU$OO_7*%X1W(ZY1woXg!9-UERxHC!Lbf)}1HOG)%rrv%e%3
z`kXQxGaTmL{m`H1a5Qu@bf9s{P<S)qzQbp|UHql=-ilR%`gDfOt8sR*N2Y<|Ev+r4
zN~&u_K>?UXoF1-%`2>+9`t^4^O6*PZy$jxv<T`EI6_k%INWORyf0x+{wzR{0N5yU_
zfI(4`j6Tq|-Xmb^2<GaeBtBW}*Y31uBZg+|W$48xK`7bs)}L%Hyit8~Z^UuEtQ%)B
z!YZwzGWmB{Yv->e<9o$CkBRRuf6sihQMXo4IS9k4Ot4%=D)HnIYOz<mz9N}&&o*<)
zxGQwYEPC>)N*H&(;LgB>Ww*dPi=~Dgqyh;T^!PcH8DTt!BJ|V0wOLD*yg#32<46~3
z$2f~L?$uy$b8+S2eQ;-F4yh*d=FSZx6syf-v>PM{Av0}7b7f_u$KXB&5^}H=5-PZd
z48F+07ZMUmCdz+5f+m%D=fCgY`F{J$D*RS;BqS*$MLFpgP~^=t^ml~v=Z)5P)KQ7<
zlS;QmAQOZktK$-JUkF~k=s`Z}(NPsH4}L90=!DdkdM90&^)A<Iu@`F6(s-TS+t(Yp
z+3mBl=_8`$qV)$pB0Uf{O_!g;Kl!KZeUCI1EBP4+kn!>U>E>ttG_!)NN>Ta2KmEn+
z;|fwJ-$dSHguMHQn^ZX?Epo*~`YV12s+`n6-B7-r|I-tI$N$fB{C#o%AFW5G9>!nS
zJDJ2UFYj?H7ArfuOYzE2nMx)a)V%zBWizvE7FJeXMn=Xnp@1$Mb5y=jzH)*|sop{D
zVX4fNK_l?<mzaqV<xl9A3@u}y<^R2hv^1zIy(tzxH&<>ai{%hh9=a0CLHaR^$B!TL
zC>HV2zh3Fj$RM>|=}Tj!r~inL9LY~+^gIg*d`BidCrx>WIxxN#tNu?Bfhw0dfcX@O
z{gC(TdFVgK27-wenNL8(jPRdh^pP{l11YSl`?)^!v?%{{D+R9=8L2|{pSSn#_xay<
z?xlfCmXVar%lP|l{Pp6bY`|-oe9?IFe{-y4n8AS<=>Pbe`5&)_9`FCF<xxYRf(_F}
z6lx9s&!upb&J-gF{B|Gz2fv@+9rR>0EMim5O?3q-v#QBAtOtwXoXNp7KI1m^TU9H&
zt+@Sp5qgY`i(#xiH`gakB5iSTaY>plb=sWQhuR9NXFXsFQPeI&B@1g`qFkr!YZmSf
zSJ_Ve;5!?K?sqb$xF4(}<rg<jI&Y41AJf^@ZwVY+U+h?<`bLQ#bHVSm*}?ao@Y+oa
zxlwx@Zzff}5}a|4IXfIyWf*q2o~*RCsB_(VbAG&K{ZilM=MU#zv!(0=gEvgd$)-5M
zzwpXiaYRjL3Nn1oJH<|>omL8~XL&SW5QnAj(MlXvFAcq2%G=4U6TMGn1Xu5~va<G$
z<ZASWu@-6PG*$#!3m6J`9@{wm&do8`yG*i3w;I*4NQ)R=r8gRZ#^#sOlN@P~{TXN<
zKY-|Q4yb5S$MCw?s#Xc<UYYm3KFd*(FloPka5QF^VmDJ~79*Zx<$JlOpcVX0rs;CO
z8?hpbD?UQ$)WZ$?`t@OHB<=`}Y@*fEi<K1n`_yjX#O!%vM!r`^V{#NcES8fMR!I-7
zOZZ}#Rb7v6Hj0~gL{|DGZ&Iro>>JNuwI{Qly~ya4uT@0WzeaS~hK2f^E}$RGxXmRc
z@miQ`UV-%BbTZ>sujS%W*{%EP)f*AFZ5x<s#x|;zoE2hFLq)jlaA!_3e?Nxt`geYD
zkEWKsnd`KJLhI35q%|(xO=$-;uSt8vUW(IopTp%*y5d6Q0{+GZ31yTc|4Y*$^KAS-
z&Zyt{gO5nB4BvNUkf2mgJBCZ+VNS9Z!;QD=cO+J?0uyw`;|+bzc41F$eCGYoMT#X2
zKZS+e<Fy`TJDBy{Nm8J8M^vKkeD}wd7-zP6x#N9)n9KN>(R!Mv8#LBhF_Vo`cJ)bD
zCv#0N)9_>PR~0@`T1lM!BH;Bnv|PoWEH-R<wH|hbMgQxdUvh_Qvtn2!$QoUB{bk{+
zT6AP+PzxUJj&)IOd+4F(PJ>%RsYB0B)6FbqW^>DK&oAimGWH3t`*NICGSN$II_*f2
zj4$vP8xP;S!}i>+&C#iOYvk|oO;D=b8RPGnz|Z_y=EC-7GK&;Ht0ZdTHhI<FQb)Zq
z20eHwbEuoN&kl$*Q+==;O64QZm6E3Hn`YqWzw>)vB5e(PmNmV=+BaR!e6GaV;5>9}
z*Lb$Fa(U1%Xy0^I<qpCz0(GFtuW7HH0(0;7Gzy6Bzz^m`x0q!V@MPv&usEG}#AwO*
z-^602d~aMa&M&NbWsozq8a*7!oSxXd_#*z##P%x=1TB0~b0JXbU2YA;$?VN#-A0i$
z9$AY<pZ}5RB(DZE^eVWTNMjv;-_f?g<H!-^(Xi6q*EbVoW>11?1s=OB1tv^BV=epz
zH^$f1ATtPo()MLY&V{leXqDUOw23tg_rAA=bR=$TH^V2q%pZ$g?uF^QPM9&2^H7m<
zzCtxc?Z|8sc7w1`i@FxDnzv#LIxpYt?)hl}b6T1RLFzg*>SHkW8CV>lF%QstP407X
zyw%&qTD*A<1!v>EJhKFELn_z(Lz?@XN=2apwv!KsPH|J6`ovQg(l{vuY!X>LK*Sd_
z_yk2MVj#nFLUF_1k`q}F7VE-Am{n5O5E|5pcpbcJz3Ufz-k29lQIdKdMK17u-?L?s
zYAS<R|K`wxHTLbwwS(2)Y3Gg-WYv}=Cd8%Y8lkNY`=}dA8DAo3wc1IAcqF(~dXjiM
zAc<VfZCZ>zti`5R(Pd;<p=muXrPt>h6!3FlNxg%!ZU5$~_42>(g46*6s268x8kLsZ
z$N}N#)X?wK9vg*~##Z!i48YlUfhE&Za<!!k1$Z3?%{z=61=pg*y+;X~qdA$uVhbTk
z3d`shJBTpyOMgo#6##+VS&Abf{)s2gQNPK<jDfaMH~Pvt>7Y-PGq)p4HT;M4^=-0N
z!8134d6CKdT23(BMo_?L6vzx>F`)kFxB|L1rGZKhR;QbHL8|1Y@i{drOw8$^@%0)p
zaw_W$wvx(RTJG0OJ-<1gvbQVE!<9JARGC>@N}Y-noUIv^IB#jXJ}M4K+;XG$DZ|6n
zIQ4g(w8|m-aEE}Hzbn{*<ErvwXO@|szyDoq$M@{-UyWmMO4A&n6wgJ`LYvV@YT7!+
zgG$<6wy@RXNKjZ?RtQwZ7F-?dyhLDg7xJBpdg0gFwO$ITj1Mv1+Yjs}wxtMk=}M@i
zzn<`tI372d5klpiE5>Agm)I+#EM~b-9<}<68u7eOzeWWkdo|VRL-@fn5qn7aDx7<z
z6i<Wcs&$pU;-tUva&p@1;?Y5y)|}2l&CrXaHv|!vJ){N%=kcf2UR67;y2{CKQj!Z@
z>%BSI*^Awwn$O=ndsc|<tW!Jn#N(MuZeBhlPFYW0`1oe%0mbw756g+3yJo|aJ+2vr
zp~~8CXcA*%_aF0Tn;kDGvRv5K#@ZTEenG%iPq!a2n!ymT{Hn9uFV>M)JqgdXa)@C|
zLsXC^gN>xCa~tF8PtTbC)<;Bwj@)vobnMdsWFC|}$Mm>@`~8O|sQ7YXh<M-YgN&7Q
z&u#8P-D;H)lv1Q99<dp(OX>Z${!x#@EZ^ZJVqG+I5IT+NyVYj#q}@|K2PtIO{Y#m#
z$ZQz2x)H>QiqPv);ze|7=K<-^$BT@?q<-i`>5kAe`pbsh7AzGx(>W@)=}*pWZV1(v
z6~PpXXe|yr{XGcjjlIuqJF_bwe>kB~&AzrSZg4Fe+I=d=)!J%P-YcNA*?l}s$MKe9
z&ih1vH7*DmQPL(O4=Gkz&W;zdjKHPwy)5mnoObL^VvN_OSC^?e>d<ih{tmfMm-rI_
z9jWT;t{lmMjwFt2&`mJOz3i*s!8)XA;hJ}kc-K`abEybFu9|V3l4~Geyz`#1y>z&(
z>F{WMwxRw;dxHvhwY*QH+jhfbso$>QXjJ9tV%^Gg`r8&v6``D+w~acG^~x~=-hi>)
zB~;wIc;f4;Z0`|w@R}^+eAVz*UlnK$QhGPnN7p$}SI_x)p2$pl++3Z^@PBwJ>~*^P
zYD_&<Y~p2tPY1NK<@sU%obSy{o1K}M{F>L==impkxq9ItS!{3R*J2uuC%fHwoNPgj
ziush@IWZbM-W@U-X+{%uF|_G<%2H+c6XWB^mnTDVDiKyHsPagaoXk&NeJ+E?U2F%a
z1*o!COk#(l^*;xR;<KkPOSJE%1VTj)avWD>Y1?hN9?&-wU&V&{+>;VjS*uD}SLO#M
zLW3C=pE&f=R@uc2EwFY;Xn|NgwoJE5V8{2`pX_2eYGskZ?ZbqcnYAU#<+jPeO0ScN
zk}Qp3hqVXlDfWK5g*kJl94!^-=+`l<d=uZbCv$yTyz2ATd;8Kr;Nin}=CVpwDTK&(
zJ4?m}J}FSIBa~6IR3)!Q4mT%;D>3o!ehO%V;t9SF-`@mLQItAn=sOcCKfA$2Knndt
z_tafhTN-YB{zxzk<2h6~o7Zj@U#OI~Pw2~Ng`wCe9yW*w2WB2rG&KY*qMzY|P|JX;
zJ;rOOhB~fX(7Z%O(0M5rwt>BVowLKD#m8{fgx^$jD_P+j&*Biuw&GWZ!}{c?Rss1I
z8Q!B*f^6@elN8v`HHJ!a_SKru(W7!BIN&32Je5sFu^>?QecV#@uKbbw-6*p}i4Q9x
zG<;77qVOVuaHx{;VqSL;n#w(dF|_)E#@DIiYtt|Z-dFh3A2_gOUKeVnf)Z*%sdcT9
z29u5tMG;<?{!gP8F!gC<=j97J?&46c6*(Esf@Hi%P5st(Nw~-T`$arjK9g$tCaRj;
z&v}%W&;$=;z8Y~*Wj3OJdm}>dF>!~#Ij0Kl1oj<u)>@=4oiCH0v<qinAKDJlzBhK*
zKjkTRSGW|r*eqARxxONpfDZUuMk#Qo&tmM7mg<o)jOkcVt*);Z%Of<3RN;~v)`vz#
z3GHRN+a4wHr;6flF8Ma--qG8|FEm?qKPq+1yZZ3aDvyOMa_HQO{sqVJ#IMBc<<atC
zSbo`i#;n|f>xwg_0&T^})Nku>p~9S;tHQ@D)l)2E+r>qCrfvO>DQ#gX&;XJ0j-sk{
zz3b$vnx)CIyshKk@-Im%-~ZOI_c8h~?P4<gd1elml7u9)8inb5;$s3o8_ZpbiA3Dg
zvclpaK?x9Jk-j*|#@JvSZ7Ywj7jbF=gP)t2=5t=UcL}edT@Po|C~PY})i0!rxV*{A
zHrSPUiV^OGK$Rmk@erd|7eINmc(3?TYf6MwmKd_K8*<Ety1f_Tk5}mENl+|znKF(S
zte$PccMr{s?<}HS?00k4Cw^KuSI@Z}K|)<2T2wryPYXytV)VkF>x^&UI^g+5)zO)r
z2?d1+8X#fc3vuY9auDJ^4<^y`qQVtFRAat}u5<3aKAS6;WisG(Ulw71Up>JD8m{5n
z*U8;MX(BM=Jd8?XNPXqzk1fEJ;eD)Acs(B#HO94Q$yLPeja8cGwA$B5dZN<P?M0Er
z78L9d%WX?zMEnI8H8OV$YHjjX0b<tWtX6Wkp49$fm3NF`Rr0sUzLiEpM=h~N9LaRB
z;G4%w3&DbL{RY8`jI5oLy*^nS)M+pfOndFMk#NAwQ^lHjc&{KBcNc5#M%o_go-Y?S
zT_=yX`7>|UEPcp#Vh4?QBL|_hP^vV&6T#U4AIx{k8moaX@~85m{OAbTOvS~BVujY<
zCiwkl-zD4oH>6+GVDR1bGhK`%5%O21#J|TK@Ko-7i94>47*lzv_whtd{D91~f><UV
zQX-Sbu(Dv1mVtX)Xk;#Nl5~M_vg-ieTH=GgEV|k5k#N1Q7oe5UcO>@NFJ<yS!^HZ|
z+O<o5U`Ai}HBa}2PgA5-+q#%uE6PKS3;b3*@%QP*P%ir4Wl#e8gleHx9Chp^ZKKv~
zYI~`<>)q=YbND3gf!L)fZJ|QJySoNnU%`&R-lEGQy`5dxI7CNqaEhQ`$VO3%Pgi(;
z2u|Qyy~`<v>fL8ww!t_ZujhO687W%)Q6`Pp<!<Yz;B0$O<PoQUaFSmCq}Pp|OesPt
z69yg4j-U-`zRZ;%u>KjvufCpjj-AD?ar&jedaTIKh;FhwOD0`iz&)!=xjJ$~@Y$|R
z$rGeD27)iXusUMlC@9;Q7Si&0Yj<s_&HP|+B6o44F6-#qUcyt4kr1Jj#Do0`gAAW%
zVsm@fzl@MMSJQN2D!M#OANcYuly;7|C&aV#|0<`>2p?vKi)}eP-D%XQx{7pOJ~x^2
z{;BD6?{YDi@VLhARpZe$;&`Y0vretkVU{b0-!;o46ukt*vYE{XawV+1vOI`1-fJcl
zTy*a1*@*L$a@xmv&j<;@F+_6_NtQDM$1rqwgf<ScDi)vj;DMCJ;9w(@@^{W2e7k@q
z?vq(lj8G`TXiU#ZS;)O`CnVv1<FuNEF;v{)4viV~>B3O}35mfwiN4n_URCV{qYUrr
zcm<-VCjQ=J|IO5fmDwqD3Rm9=@x4UuK16|K&Uzj6mf8jEZa>_%ta>tz47(FW#>>R5
zNxUN-;XS*UT2IDnE~NUZ&xprJ&@NR2cJ?5e){>F|{;FY)Cg-|wXJP9pTU(UG`47yD
zkd4`_^E-m&sS3mjGPHs0oJHtjPw*}Fob4<bs2?By)G`zvY&Xnkq=xpK<6-(x(33kT
z`LD?LmXKI;VUIDyLu{~N%16Re*4&yp&E8N)=7Rui*itzOnvpcy@QY(GO4iRf?0<bM
zIMHZIZ>RR1k@LOy63bY~&e#IaQiS0GkC5Hz_yXVAW#dm`@t&0aC_aX5YFiTN^HJsC
z6BJ8KdPXeNiQEmF@{H8lIp&i<Z;UV>T;m)fMtu4$0l~N#6-5i9hZbMU*E{=Bo=UWn
zcay}8XD4uJ5of{JbGlS)JEjkzaxb*1NW_Izu(YZOq(TMX4A<ozyZU{N-scfyY#gB;
zrtYGuexkrwWvNu&Go+L}Q)MuBIOpmwbtRl!@DX$O9SU-6a7<@Ww@gvf@6laUQVlrL
zvy44l0g6T@5n~Pd%u*OjT4Bq#{PyG)AzbAK;s;VG=VULgVjhGXm?Gg81l6gpw5LPE
z5t#CbXJ*qq=W*kT!BQQB28);YqEV6pI`fQ{*X@FjBG689&Lhwt<_CKE9XF%X6!1n$
zXkMRhGng|ByM+#S%G^)w9@{*8eO)p;*nT&n4AP(0^jQ}1jM#_Pjz;RoIskcI91YoZ
zS6mY!P}G?-yvj^u_3<&+(GStZ#PJKS_-^L?Ldl->$e&Y$_i))#s^7<tBpJ@uIcuQ9
z-isS}F#T2|A=8cXNUGw#r)2q4O>Ly3RW?oE6Nk<$)J4b?<?5@RicO*Q-8af7AN{nG
zrNd*G6D4O>Asq(3eeY*a7s_g{Q~V1X%(Zw=%bX}GSoAak=D+X0i+ePfdFsOTm$Bau
z<#vMmt`d?ryo;*soorPI_2Xd_V<Y}m{$y68%l+c35T3)nc4!)fHZjvZ&$vahF!#Zj
z{(^fS7AncjLZqQc(Kv%_mIbsQq`D7JK#0w;)B67y%2rBvT?Y4lGH>=q0ELPZl}IjH
z!u1nP-@mdRGGVc74J1XN*ZX++%&mzl>L$JTMCd4ogYDR#3t=p{n(k?rwr(!bZd@XH
zDnNVPJlav#MnR|^+g1|7&}}TzDQ4nLR_kSv?J6{neKDe;ntc?Y;%Xyv|8?V*`ue52
z^2#TQIA%C|UnMCA1Ib)N&LcXC7K4t?uBaGh9i7_o57iu3_?&8Dlitwi*AuyY7F=l(
z;eSO+Z(;q%D6TkhfEM`fx#dpB4f(a6li^DJ5o)y3g_pD2(z^Rz$qdnizzc6#WS(j|
zTcMbp=5s%3zFwLN(4tv<KHeTlu?oB7Wf!rTmBqo;6jR23gg3LBZ1d;<i;7Oj-M)SP
z(9IBcuFs~AYBrkGAlpdRvW=qbGuzbp(xCjb*5<b_AMzM{=p)W^`<f5r6VZqbjk$|6
z7mRL^l)7LfjJ<;SnU6uV-n-4{^|FhYVXO9!m+I(H<;HjI+SXWShj|l&&=3HL60c@~
zV!PY7m1u8P2h*q^YKi@ie*CnUv#<H^WwDQkYH!>m%Ee&%c7g$%s#tBF{R?5}yK^UL
z%cv5%uybokxa!=WU@I*J0Fui;XIJ#bBjvv~!4=zmw}%pJ&t_l(VPP`R&pUlD;a%s3
zkVEZO;;2Qm-8@WJEVG(vWP`)G5Fd644dq5YF(}ks#coAf4MNYjvG?P3?Q)l?!k(@K
z!kjyvP`1k2w*+7Y<?dpeP1%%3igIi*FtN;*sfP;yxCQY$K-4%hib}B_um1INPNaaG
zZh4_(sZN8`Rv^2fYJjAe%oROv^YwSaahgq(r%=zJDTPaAu}qc}c^DiFP`Y+`!e{$d
z7@4zJXJCE!0c_xkb%DB}nuZyQRD6ruX?A$39Rywv73vDM=IyrzVdpcBI`vM!ZN=V}
zMl$GWM*Dhcj6EwlQax<1Aa@Ngyeo5p)mISHz6EqxDFZeArh(^HrHh>5IrkXbdjS(<
zVrEsGd`cZcdWj$@8a?QJb+>^`X_rYNoe+VrKckKuB^f2-?nhS=bmRc!m#gOFTv%a;
z+cQ&b8gk!2RmYuiHVL5H$&F_r3P@aN4Q7sJC<zqr-5pvy&AtEWfkV926k-{{8zU4F
zQi*VqJ8fpgDdt4|<hnIkyBI>?z8K0}m)ph`7C;*|;IG9<|0q8b%n#SOI9ssn=Qz}v
zZqp9I2Qo+LrYyXZX^>p=S56`u0P>R=*|*xeZyF|O_PXSxB}28<Gu0`WnVnluSa8>3
zj|NczaFp+D%@5h$8458$3%tDgs;)7d0#Io0GF^pBe0TqKIjptLfMI~Z#J=ZtbJ`)P
z2iG}omblnll9L)*pxc!9CG}j0s=^`TLByf62DEyS&Bh0~l8QKgbKkCA)e{emRH89-
z>gAs~T+7wSVZw<@^gJ9?ND)6CXDE&2!I+S`uW{Z(V(7I%Jh14|dO#q7AZY-=m37<M
z5f+0u7&TP5<ep~RoYQ!tND~dleZ(8#-mXmiW{D$NB19OT;~<Vp%Jsbsr7izKv7(%{
zoDb^W>&*@!cSr4I5qW%j(pfM$*9!vxM(-((4A+2a?d-Z+MOz0}p~gFbKdD8jyQpBa
zL{d}?^Z$Mez6(rRGBAfV0FX3@BnYo;&gViH|EqF}fKId2?tK+mCf+@C{7XF7M}jel
zUm!bX=i(^%Sp^yUAJpD`56uiL@gof*UCfDMpq^Ln_<bDW2@Np9z@nts$%h1>!O|IO
z=t}4U7WIi4pVYO;-ULk1Ol~zYDf#EXNTj~Nnl8!~Q$-OHV)yam@f*3t^f_zhQJ5P`
zT0CTRs*V)aWGTI_&sJyL7pC7zfQDaV75m4tlRM}*&eSm>V;DvSYf6%&+{OXnSKwUc
zkQA4++$-WNIelS5Hb%@k!tjJ4SwpZdYNuQ;7-I(-#uJy}L{tTsg1I9%G5T8KOBWf)
zM5FhG$e<{tMYOfq=6LA`BKq8u-LE#r3yw2%?-xIX;02#prdo6Bjp=y>qZ{-9SSM{s
z9s}84j8J-dXQDi3?$_+nuUt(j>PAYy2H1%`9Ko3VA@+6RO@3hym{OJ|(NE^xJ7tMp
zeEkEI<YiVfMoa!}KKq1U6}61%yYp)poS}?sOMHze!No|p=bc&zGiCV6d~MLj<Xk!?
zx`*>D9mbCWhhQ&`6h4x_7N6Q}yiF7ujLNsT&jNop9~4yp)dbPfUSacF)kaH96JJtG
z*Oq@Q2=e>G2`4?l89pKlfq-;%T9BW!UQtH94DdO(fn|vv(Mzrp8U5t^7@OA^fGxF#
z*dTBawzOH5Mh?$*&+o5X>;ADUI(*3t=4r9-WaaM{kfW)XRnkq(maRn%K0_j#P8Y+|
zgwl(b_&OXb?|C_i<)7ITn9!MUnqV}A^WZ;KcsIU&x%GbUBp@R&mLubsL=g1AZQj>Q
z6$U2ClyABSZV5<;+-(BC7lbq%7kUCcp0G%{zPccRG`zsa$Mc)U6S}!PNc$1l;V49i
zf$NDQy2;f_14qy>@ddIufi%x+ksa$x@bS@}AeMJm#5=jr9w8P@TNHBZHwOdNhk%g%
zYVnnD9n%v3>tRH!3?N34G3VQYcBLUtV6Y_@h0AA2pq3smyki#Kdt+a>&Q$$6#-1)*
z5Z@_I(iKybc5gUlu>OzjqeKZqCJ#>D2WRV7<o8~|xy?mxM<m!60+KQo$XGuG9uXT#
zZz2*MN8+Li{3TpG$yLcTUGn<0wK5s4F+`N@Pu+muIUb;LXgs=DQRqK9p|Cl&-Wz~=
zZ8ctO+nH^o(#WBKEKG=sY}{6g0yjMU4Uis}{84)V`vu7tdYo);E!GgrnQ3DFrD+0O
zksp(i6t`5$X0Artf660m@;l>qPn-rvM^Fh9IQeO|3YDZB&VKU@#{c(9RbpBe<o3$W
zE#j-)AAnBBD*bIh{VOLfmx&23dlFS8I`7fnFx0=1cX&~N1o=PD^S3qmE0O;H-ug(n
z!)>}4Z+9Yz;1taKy<i;M;@NCCnS~iP`6OB3WWxcs)m!7R)HA3+V+(j+vdNSgt*-BO
z|3w~e@{I8wh2S&vpwoDLtcV>g)J*|saZ5q%O0s!^zAJZrLD{pRiiypTHwPN3;%3k-
zK>DqGzk}=C)^IqaWPZ3d2-7MySP5bH@M8tfUDb0vJ0V#yhABxFM;NhL(uyN2Ih)9B
z(gQfzSk2mQj{1Vy{VrDh^#9_*69J|@$_AcECoU!&A$fHeL&j&RW~$DvTjhL}2Zn$1
z;)Y`$)}k6?QxV%g1q*&Fasan?5=nUzjzRdyFDJui(hCr~MXsRrlWIU6=!{{mF9Q{~
z*rR%MEVrzzY}U+7_d93PW%TRl$2$O}H=*|2<aU}cOGr`{oZ-_Pia3_9bO)4h5ywou
zn`-PHplE7rtqV$@80&uaI$a2+>2aI$wlMEYO-gutR%M|9Es>-@?N1TxXM0ugD23iJ
zo_OzP#;R#xBjAlyeqD7)UK^<hFB1+I>DQYoR_gzGsFKByF*B8~3G<NY)y7frOm>+N
z1UQ`SfZa?{l{iD#lyx!%>hImnn*r>Dm5R_pz$=CPq9TEE^IjeTW)(j0BE@F_b?YUw
zNHuyAN&fRSwWoDqW&8xV$X-BqY319s?Pddn)*K{_{NWb?G65>Htyk6Z;lwUOBrQR@
z_6<Mhk3t8F@yzj|dlEh%2lw9cx&f2noQ*YjdHEv$qm3}Yc#71E^^)i@3jI9hZD5%n
zZ%*7*w*sr~hO!2@J_7}*{q^uIs)dN$ZKGC(sP#^9k3+6IckgpSBNi*G`z0;{Vpf(#
zQ4F<zb$9#SD^Ww7f6Lu4GczlyN)vT+{fv39<hM&+0J?qt%fjC_J2UmwSnlsi`tF3;
zR?i4<E};IV)Ta1bIs!_QOnh)vwy0|&Uoi*?{J@mf$M4q7UhQn)A{&guACd+A%wFEZ
z!=m1kCOQ^DB|4ft0ea@%=Su8xJmR8z1127P-1PW(^)4IcBF<|c1q_9xH8tZ{{mT;R
zRfY45>hkvJwYkJQOv&996+_p-2>Dj4L9*IvxZS3-jd%snz;D^IP}N|sd;x>eXA*mk
zFGy%VwZDHVdzs;Tz1}x?tBw$H`Tc8hp?WWE%W0*Dn}7+zrIDk;k6{ac?k+<g;CsHF
zQ>16CwrCb6q|oz>w7gKFedKg#=($_^uIpDlxW!bpz5OU?-{nMqJ4M(W3~`9L-_CY!
zUeqXZn6r)3Hs@*m1SeDuS71EjfrdUPb_P5(4<a2~H`V&_uvnw2%jWoCS`KLQdT*O6
z^M#hc?POL2-hCs9bB@1fakW#_#W^YxpQ#tJpqe~k+69^jkwF~w2()m#-Vl>Ba0rqu
z(%suBUB`_+@yngUe?=?l)xD-{1Dh|wBTj*zd@DuxmnfoNg6=(OhcK8Ac!^Q*3$!iK
z)wwM(oFPO{Uf(BwzYVky%P{g=#YVHf@EK%UgRLtDw^)tXL1q`5+;|I(L=VsN-USvX
z{Iwk@(B-$n1VE+r6ZJZ;8?^={@%D=i_W)kp`-$aY&|24`sFB2G4xCOlh_tUzP_7uH
zjFwaYn)WH30wU_5-nfCrbF=hr^KFiFqq`#%B`tb5Ae%N%!f1>>y!;!){f$kZ@G7M&
zM9n_r7u#c^%90mVSA7XO87H7GNfF(yX%!bb1sNetZL>SvN!cu=<AkN<-azmk4la2X
z+n)Uu-}StNH}Ri(4G0iP^~S~y0X>$fGLplcDf1ND)YHyYeYcsPm?;G!*S<F=zI`T9
zTeEm0cw)1+M4tMOyuE^fYoJ$2*0rl;%nhb%c_{PHAgjRi12fb6E`}(vMLBf@W0$f8
zlrP<RNU`sZEl|e@!tn}ceQ&&(7wO8dX(ZtBYOzH<=bOIZS7~Dp%>r`15-1%Nl=nHd
zovaUc*nm!wqmO}dUfyP~*&kh@y}Up*joSNAef1~6u|<;s@jQMBv_mUxB!<p#{7Y?V
z@kczj8Da&<SOlPg+hF)G{}L(SXz=B$C?MagF-MPbN!&vn6zkU&pn^j*64$3D2jFi-
z{8HC9LVU1|Ae^hG7duT&3pFFi#BQ_hIR`#W*ZI}$zulk;OQGR8Y-1SdWwVQV8V>t!
z#i$K>61iKCqPD80`XCt+Gi^;~lr`IR8$GqF=^OdkZund=J{h(gWM`kxHiOq;o(UgT
zqGbgHS>0Z>O!K>@yZ<V8hZ(EK<tAPQCrR2y>j-TKmcL$k!E4cVa(qBoZaHhC@oKgu
zw$joXVilUJ$2InXBh@-D|FN>mGQCE_x0bs-g*-oxB`@gEwjQ+l9-}YrkmM>o+qz+g
z<Q}!JR=38EyPedIP)B(;OE%bIX1n^jE2WHXmlqEa7uuXV*3ZDIgWcS1lOJ!a8=7sb
z;|yk1TKjh|9cw1K5Bt2ewHt&ruoxPzCp+4bnH5eZ6WfX;&Uo4zbBGsZSdGE#p5E8`
zcKsjakDs{;?WvYFuYF!JkiQLlM8&JWJamI&G|<^`p(R=khfn|&$$tIx#oFMPG-&W6
z|LV^f(e>c0)aTZn%=!8j9(pxnJs6B)IdYqE(JIV6!YERp@vd+ONqeS+6T+-&9%2p|
zv^d`ae4|dcl6_8^tK)k%7kX^^Z-SEnZv(2Ws@f}TiN9cycVzIGg*7waA`_Z$XZc*6
zGe$2chMqhWcsnn}e14u*zgZSeF!9<C$)jKV<kyHl&gc8Zy_p}M&fFc>&W0E<*&x&5
za=~N_p$ma_pQ+%oZsSJihRU+y__3k(<b)6cOk*!8B(pFH`Y$`80INtK7~JJR90+Oe
zkOuz{!dhe=?)Jt&4j8~^cp_99W!;LE>2RUnPNW!vr>YmxxW?YcPsAK0NxjHykjb?l
zl-cj{<H&D7#W&RC5VM7oFopZB=a(z#9qXos+{O$E<ASrfkz}Xy^&40cubt=c=1Ej<
zum94stybKr$E**lw98@$`0`UTtst6=EifWs_Jpo8C%xqTc-u6oA64>LD?$tSKk@M^
z5b+uezN44A37U~|&*}o`pwIo@uzr21<1l<LLj_V;G1c`d@quUR8x5=cj*L0ehql>;
z)AZFDJDNmTkR<lt28`5@vSWV&vHW~pU3};C(m4K~A-QG(X++lDuPXR#6Le>KiW2RK
zFZpLGzreSi9}mO{<SN_CkLMWPj8s&tt4h92k#qSX!9{JvYtdH`T)FirREiv>xin}%
zD5Cau7Lzhj7Ll2xVahCJf_#NGmA{moR<nN~6<$5}ZJGX~77C5`;U@xNd@X@FuLDGB
zVteE&Ep;ZM@;+9k!LCVH9Z89oIgCpZ9C4Kw2UF0KtVhO*LPd|L?Lu;(iF8G<2U2_A
zK-bdz_>IRk=%$QrMTaC`c0mX6HeAHP3wC@EW)c!8(P#C-2j<K$PDglGUPkOWzGBjE
zYQjh_vXKm{RY*p#l^XQWYO#?(*X34Q6qjZT@$P8-G=M%kF9#pno~a+py|WtR6d!6)
z#K{@sHc7oL8Zv^B*<5G#{H6C0Er#dbaDu-AxA!NfRd&0kgR_l`zW2JMG8NTAV=q$0
zuAzrC@CN^&RlzZ_1;({j2Q)RiviG0)-gi8T@1N#FwWmyWtiMUs{e&m!E!vpDGv&o+
z_4#VDF5Xh>jW}cXt19le1M{T)(Je^mQ4w3^Y<%2LUttHlJv>z<IKMj`DSdh!NsFmK
z(AzaMS!ViCj8J|VW|^!QzAcKKaLFmRbLN(BEV`-GI~r9EHeC$fao#%YKi$J8&+9zu
z#2(TsRu)Pc6x#M=l}*4N;Qzdr!#FzplgV|FQ|9-0F_VQCJ6@j*6UxK`OBEC4DqF2{
zC>U-LOD{KT`9=vnaYN>OFf;ciSaQHPdk<wc|I^Po{}w8pT?~07d4%joV~-L+VyYJH
z!$&q4^cYq^+NM&dKkI~z3qgDZ%%Y&!V~r%v;PVI?P`j<PD9{f<wRfFOVzbmRY#olP
z`KB!|FK=Aw<}<#BCw4TV!I1Ixtj?}#!rbNbIk9S6A`kAI^A?l0z>|XhNBx;V_XM<l
z6|EHEH?JX{Q9N_U-@F~F>Fx#F)-K08rI0_z(6z1n+4A!##-X&mm@fxK*&G_{6O21O
zhRj~aE?q0hVKKobMu$t^m}~WR5_c4~!>?K1peR+b2zFOkTU{qc&g+;^CD>Tixw7$Q
zK#I?{QRP^v7msv+c7q6`5W}*8A%E-S4V-;TVpTu1C#3Pmp<7Bj*?Wd!o8IsYj*J*i
zl7x9lP9N2;2OT9-GgWFLju#%MVg{6zO)mWImvnE$q*<-<ZLy3W8RvG}4|^<BSg&T2
zSPZ262L~Xe0S$YbTi;J({2!voFmYZ+Do@b6>3jY7=I<C2hUYQu&`~=B3Ae~&!b<1z
z7V3%z!n{{0!o;kKCbzs^q4xqOq~F}by$ntBQbq>K%|j&^SNdI<<NoIp1c7RSCSm?-
z0?7yVS_ru&40WzgB>eu14sRth_VQXU>7Q*Y;ap?hGizSVLS<`0`ba|SzkUYMG4c~j
zWt!SiKRU=C9$qEy!V73!30!(DPX1*MZDaYOY=_gKC>N-XM>a(Ci~@<?Yd-TjzQ3m;
z&$mMNv?_>%zu1@;+MCSZ%};^rA``_&Qg1A!>a?xDla%B{oSYqRX;nm85*LX!*$GJ&
zo$?Zz@Q`@b{C03!$9y59Y2t(oX1opIwLkCYc*4XjrBoobW^6m@97l9e@U-rGXkbXN
zU6-)TmxCb)zX2ZdNO%HW8RAIp9b%I^aeuqWdqY2L_l@ABmh!}QedO7C)x=&d)7?jV
zvWgbfb!E@eo^~HX#YB|4rfRIVe>tq%G=l0`rg3pomR_Wk%(Joe46dFq=K+tO$i239
z)fv=t)blNu$vL+3Eb~3>rMt^n&Jum~lBqj?t4%ReIEuf`{Rsu^C%H9wkdu>0SbOq{
zv%tm}v9Bai4k2cy!#9>FaE6|t06DHh3sn}u;>(dkV!QTGB6&zVXt{>i?S9!;J&i&)
zeF5ox^r=@XNOMtR>tubXVM&-QJ(;G@f7QIv2Ivyr4z~ckjx&h@k!lN&+189Mc!_Gb
z^U*&uswF1<(x+6|h1PiLv9F*MOZU{N(MhMn<~NCEj-G9I6?Np`&KiRl^gA4h+{PDl
z?kef?-2NG>{TVxrJOX$U3?F{3Mtv3PU`2>=S*lP97+Sj@Y-ZP|WaRV<u)5e-ys9q1
zr{J8+Za8y!nVn>n)2XEytIJ)U_7iKYDx9*{sfB4a10zU<uY&v`E`8qg3wv1V7)0U3
zAmK+$AM<co>oc3lHxg5BCo?`p?jIK>=#!#}n_~`LeU*_bC(ye}L^nksFQ+SZIN~ko
z8)nRJoK_1bRovZ-t$IgJd00bu7$l$)dya*9dY093^|_a^8)ZT}{f^}U`lpshaUv6Q
zFWiu{ZcJ}Z9cM~e6qrz~Fis_wVgj2kwmeAuXp!es`9@K5*PI@sH1?nJE$sNz&&tyh
zf#a6DW(mv#Qfw$b=c3`bfq|liwA4%&+x5k|mVRhq+NL4z+M)LrbA@du`A#NsPIj8d
zLv32iL$#x!eh;iGBw8M7)N{joaD?WOPHIc9KeHDk_B}qnl}=le3<oq6!uPwjP4W`s
z(~0yby2XY~j}BdI3eJgI9K^r0uesfFNS@;ZjozNaUleeI8T$AGrF1ej2Et1g4i>82
zHb&WV=-p3+&SwuXXg!JsYh}~I-0Td^q4>Bj{;n`gm+HMFy;0t|MTFE;Q6TOsZ5qPF
zG|$W-8ruL<p6Y#JZ@*Dc{^8<u(z^IeITW;aN$J#8063^q#>iD~=eq=W%EK*V(Z*DV
zrFLN>0X^bbOSUJ7*sCbXYazJk>fGK}S<2S#koBusiVt@R8_$1lL)mba#KrxIXk(G`
z9Zd={XS7%3r+{XJ(UIs)pK__L$&sj>RnuF1T!#0<2=BDATG-;Rz13?Fv>|$_ZBcgs
zG*hLT=k|Ai5gGqSGQjo<KC$?NU3rWGvFgyCW}+<SO$Fj`AG=`y!n|PuGGK{v?7i5y
zGVfF+S6Y-s%ypB=T15Db(LM1XK?*Z8Y}{4~d-&ib@E%kW>)wOyWhA|i&<=q@w<^a*
z@ri}61275n^!?x8-@`QYSZDPAk)UMKwKHLnu@xp)P<M4aMbLJOgx)O{60-`%rg+KP
zgEYK&Ylo0!>)_C@t3fk=Kt^$Thg)UOJe#B7m^$?yF7HcoPS^0MoyE5)+$KS1U^MCZ
z@rj`qUx9MrT!>zDwp|)bb#0wXQHnRW@XeLhNuRL^RBB1TtHp}v&ou~9lG2!8iq?<-
zGqMP3$ZE*9z|wP1ktx6dOjo$ck-?>O_PjO<s|BDEI5ZN|?z)TeVbzbQjoEN2hMuvK
zHCcvp(jbQt4YT;Sg##?tc2wI0yV~&(I6*9L{XwGun)y1s=Pe8;a69-Kqv9AYg#9lr
zi17nsdVGOnq!3Wdua3S_3y51SuKWF%dE{_?m|0t|L6YIeHt-bKm43T5l?2}Vt`vV7
z%@*qn<if)Ed!Q-f0uyC+N?*k4DwquHwXv@NI@YhKH*(JRtB1Fr&mzqw-u<9guor+0
zW}~2DwRSCq45ual%&crJN3dhvEh#l--eql(YI3|zT9%fE$27#~p4Y&?#<o)v`r|p5
zG5x;Q6p1L>F71{?9?Jn*0<Zb`gSG#GeenQ`8=uDWXJDfF7>|TPlI{~iHf39_KY54|
zUcX(Je=E6nUkD^g-JPg&v1Wcj?=?TTp3^~)u=^g1<L&SUjmO$&5i^D|+ls*_Dni~z
zKeezq@9%KKr*>7aRQ3urzAs`!mf9{3Z?&^ot6(xBP~59Ay$8>z%NYN(ii^>LQ3CZe
z0RL~LYRb+jYXz*s4t!<z<rUDvrp_yA?!x6y>1kNc1gc7F`heBt&34Ta4)a^!ZgK15
zb=he+alQJL>5mguS?6X_&6?2BuhoRbDb?-*cuD5!dHgZzbVMg{X4q{=-BB<`S_-|L
zi4sdqI_l7|5ACUKZlTdyxx%tPz^t&3@AO9;_Z;Kp$D#Z>0NX0T#5;(#+6_sB=9$$^
z(qPeZFH@D~0G(T@R+RbUvtOMtGqKRFW`asJ>}JxJA#fkRuQG+!B?b^D^HMehXTaMe
zhf*%A!nw9u7*FTHygzOo9uAPG{cqy%5js3U=e})jcKma<Z|Hu)PAd;(g60?M5f79`
zzVMl`6Qd9j5!vC(`0Si+#f3_VpafJI!8=at^Y!ap*UK!MFi0c$f1^%37|@Pk=gy9z
zr^Ow3^mr0vJ~cel-Mf*zM{z&D)w}H&><G+%4%?k1jG)jIdx6#49!0Zb?7Ns2j&48R
z9znJwPzZK>L0|r&0q;Wpw(e2G)yH007iNi{ys7z5_jQX#O8H>KuH<OP0RZw}czQcm
z<pqEuoak`C@KK7Oe=61{fBCS8B`d)YPpRjR#l8KJb`13wa<X;SW7U75v9!|f*59((
z@k4iCSC8Z55A)7=4}Dd%tK(f<z4bnM9Ik!(@dfw)0>}TT1Xw61MGdq6ZL}Mq``)z!
zv2LGoT113$yc`E8kGa73>pRts4X~y$>6d&Ic&$3WEYn;@S&{|vKmYNB*s4<DwESXD
zRDZGd?LPyLEfd+7v+H5S@J(r?R0mlZMZ-Vaj-pu?SdHcgaz+ko$B-1rYkf0*M?Ax#
z<^LbT2fWX}-ZG&8AZ1T?DvbY=*#9@rFExk{$SCI=J?uXbtbYSqev5QKk^}9leD&Xp
z|Hr!kKg!<<=?Bvs`L+Lk1%J7kuwYO=7KO*&<o<nU|9vTd^`ZiGU*4_j=j%W7?!S$<
zkEp@&hp+18XaCQO_^(IQoB`+k|FebJZ=w)$+t$<2(K$Q0>-~==VL$UHnee1a<}zUb
zA=i=t#evl@EV35h?F_>Kf<7SGtw`RSsTv=+ucldm+7XF3`GyN}QhF^YjWZlW(H&FY
z)W7rqW7uL7%*zmv0zII1rt$U8H~aU3(hAuaE$sbBCkp|NpJ*8N_P1}}xPSonBblTR
z{0!i?J-{4g-qqFB1K0(7kE7pa$FrXH<z0{SQOivag%fPM*WPgxC@1r=yu2tApYCMN
zu(_>TMUCfghCzn2DPLvuPPL}xBvSv(f_Yox&4IGylT}xDw{pGfRu4elmy5KDCgRz3
zpTK&Lo>0d^(nNk1)GUV8a!f*!c-r&;K&04h7V&1Ps!#_wo0NgH6f)Q7t*of1ICOY6
zxi_T5nKbxCLG)OG(m_rIm%t6uU-dxUWd6I~o!<9oSUmvowT?INShop(;qhk(Sw27w
zlR--3wfJxba{i36Gt8oGb#EQG=9Vi^)|~pqS3k(6mplrEN>TA}2lSc(2Nd4{5OXB~
zuV2D1fQxbgEZC&%SLhZ-`mOD!7yKhxjg9;n<m4dNYp`M~03#QO659_EUP(0R`5a80
z{3D?=Sg(>j#-vTrAL<0kTv7+MXHNtWWUAonET7#v6K5;IUUWS+lpwgcj`;YHRSyM2
zJt=~Kv+*>GARn04%*z37wpuAIcE9d-UjDRD%+~lv#WrFNy+rjCmW{L33_&p5f2f^*
zj|BnL2@i{eh1K=x60TOiGj2<n`0HtI@sy>n<ae5`j&)DI!#=j>+f2eL{yaM$G5xae
zq<E1(v0=|DOpAdgGJgdKv6SO`FbzORxDH+UbO2G5)d$II=nelqmFiE^pk2SyTHp+=
z5fR@*5LlrN+%i|GKIiKyA9Y=RYWA|^q#6rMzm;9Z?{)wJeVc7y3^>><?@zu(@%jgV
zH-SNiDB7;rekz(+ub-;iqF;bP;R7FZfInahfW3Fs6^u<}2kaU<YJl{?+8z2*ODpB<
zU^N9SOpAMRKR}67K`*2-t;%mcl7N@BB!$)oJ~`lje-;2}aWY&}b9Wqm^!q0=Cg_#D
zREHMb08~3DSR9(dOFTTA8SGtnUwKn0m0Ns@N#Kw)(e^6V;Lp^xubgQ__}QzPK5Cga
zkSq!UgDtWQ12fQ%FLXZi9;F%1oe?P){9}5&6WbM2_$;LWKktj65bW~;(11B82QWZm
zesbUYT?+v0)YH(MHLzQTN-yXLs7*0gLK>h6^m%|q8&*DwVB_m-Kd)j}ELFBU0L0Sx
zuipjtNT`v;u)n}5K^?Udm@p-9y44r~UEE;g9lhe-=6G#m%vM$#$c2TCnV5N<kU=I!
zOC&dugkJTzProaqcxNxx@Xt3C_URq5bk}^r1<OjVF%-n`jf>zCu!44GBA|*PGwUWb
zg`iFC!-|bg_uB9LipEoT1=`8C2ATjFenqv%oA~J}#lEq2iqB@eWWEUQdks*d`{{WY
zCBUu04-C2%lO{YahhL&39Nj@5$xkvyXkVK7>Fy^CSuSa|m1S9sfF|?<9;&1(E?n{{
zJVdFX@6}H@FhX%eyi;x6drFY{yi0i3{TsT~Zh@>&&WGIlJD>sc=&n2{@0W1eJAbO`
zPMJf;Mljk)JrbH=u~~gUosi4PFzh39^BL!B2L`fKlZ=h=Gq;+I(|xC_?fyYvBXoLo
z2nMUP5Q}gH%kg>0I2N1{7@4gO0HMRYW$||3n|C~8&&y)2K&M=94}!cZx^svwv6Ph5
zb8F@un$JyUo$|TZ(pTR_u=00(qu5I%QzE6|F;X>TJVbJ@VX`t2SaJW3YM1otHj)~Z
zXP!s;HP=0NJI7wolQ^Bf<2HkO5-iyoS%!%(3;pzB3pdK|>#`8YIc8Mm0h&g7**E(-
ze;*jZ(L7=V{nUL7o|(&{`fW!+6R1HD_NTxUd3YJ16A5S2%6DcpEN|^IX4|)%JfQbo
z1j8WtQDIc)r%O@w6~R=liR-`s+f~OUpXw-axhu<jPKQBT@oTtdp>e12q6hex^*5Mm
zDk?_QR$Fl;;e8`6#P-I1bcw-{@~;xf>p{J#*9TVbr~}Jotd1;5gjEhi{ba5dA6K<u
zIf#l=w99mlz#eE5zv`GLnf2V3B&R`n62<@tVglja?|1aSp&GxVmni>E_~aB7wpI3&
zcmb{E441r=C`#n_qt!r)=?@s+Wwc^fu{mpQ`8xL4K%ubH6wdCT%3@iGMDr<CyafC`
z;KYOo+NSnn^v`N<hPMU%UW(l6es<I+i$?0GPv!W7E;rh|A*wM@h-Q<5a34%DHEpvk
zQ+x5IpLxY%{nN8M&v~(~xVGKDNFIpHzq><O-ObrFjrZJt_kns)weTV_bqcU&%D;#%
zsKzg9)h)rcF`;A{<-`qmzbI;gImS4qHIk*y7SVCkU->K2xgsX<$2uHh#jr{?8$lAG
z^IsdDW_SP4<qDkZULeTRu)cf8Hit%z{TxB+3gi%@_*qGjZ-IYgthS2u&ffO&tA`J<
zIu1qws=(`Y`c}{&ngrL-J{7B_3h$L16H3y9Zv}Y*J3be|I%?`>RllnutfunVcs4fC
z;#PX?Es{c;E{%x$&Rbt_CfgIUr`kUz=@HPxT20E6CxX>E(y6L#-Dtw<@_2T2jeRvv
z^-IS>XRIX;N)fwrbLNs$H(MMQn{Klk9FfX8=2v**ie=HyPx6==yld22TvuUg0D^Ub
zKM;k*#l}v+kIfFoef7u~&Dd=`E@oB=Ctg=;yk%MM+SIM}*1-{o%D;2=P4y@J#SC?n
z|3rmm{LJOq!R8#Bt+GydJj^7R5wfFPpH#$3+@AHJO8qZGbcj4?6Yq|Lj~n<6OLS`O
zKoKk$qUJLmUf2qzP2yZyne@!uF6-vl=Ay7E@jF0~MQy(GFJiz23j4J*7<iivDvud2
zWOTi!HBj5VurjANTcTKsLI`i2YgsrlSXqyHH~||a=YlVM_de-70!L0IfCz51_jU=E
zgqEhQPHsf%hq{;i^9%kHJQ-l_f2j2`7B5=p>0O%3$2+mDBYKEM=i!f?PA!<sH59lA
z(n(0eo}&z^xq2QTgPA{ksAiVoH0}c#!Qxg)xr%)>Ad5@(GFgGJC&#S2uF*xYjc>|l
zaYIHj6L+Ca*|*sQe=o)QZYg^F;R(z<#KoudE0F1kt9zq%-+D#GH$jmV*DEj-qpcOd
zNNSdO&00|MUU)6j|AG4{UYk#fZIuHq8s7Nci)jEbqy`Hu(JY~S=HI1^^Z=U{AWe;7
z3o!Z04@4(e2Ji?>(lNfV20^=(+U?qOs8k6kc=e^9f=LlXjINK4Y8QeiHm@YIq$2ST
z5+%O>UIj$Tk{FV4n_H10ae4TI8!sPtDE<Ly27Y|2o_F216Co3il2pirE6juz*PIj>
z_p}i;2vRxUML$)wcd1-P?<thdBAZzgNc4bB)B8l~fARK~0ad5%-nJqsAZr1V(n>ed
zi;z;36h#E2yF*~nU6RsWB8VbLH%KkIq(iz9SaidCt^3~ho@dXz`}y|1A7(~jM%MiE
zy3X_X9p}9gq9=Tp7tO)}L<Iiu_122uVBXmlcB^)KjV}bNBY|_6^>C1xauXa=D;7d%
zB=7F=DmCn<GfkLW%Pn}^_viRp$eaKB<}<lOfeGf%yHuLVsZkR@-&>8BsvqR{f91}H
zU6v*DkTwq{eW)2VIUaC6Mx;q7jQZ7>TxY8FwMcmLf$FouN_~^U>(A->evsPJbQnLu
zn8p4<LnTI=3WBPwOC$(;D#t{-l>XtOQ792>&$eyTeJ$&`6<gA})&^Emjvr|G@AOKD
zdCF#1I5=!K_SF(WU-wTBXRdrodkd@aSgx%;M&B677_*c}F(I6K*1kSx^Ltq->V=!?
zVW0j?ZLC|{eU5E)pU9c-%FXz=Dwj?Bh8W*>KAZ`WZa{ANVHJI?AiHty@kn7Ij9<E4
z08~Fsg>n%onCQAb!Hd)!aOFhB`;UA~p5flc*cG$6+X{ks470Hf?kCQ}OltSL6i?&d
z=p;YJp}(^d5V2Hxs~aU4sMYv5I1OGDn<Sj{Hl`Z4+FSEhqa2s&$viV&?Y?J3zrUE2
zmoiK7TCURCFojjHdghRR2DuaAyr6gl*1@96%DYd3I*3~yI6V+~Zzfg0X(&y=ZS4f;
zD~I8=8fb@FB`}W!X?)PRwX!NsDn|v0LE>4D7p138Q5`?SrNR1$^91`zv-{Rd-GGNM
zr@fg2H0BkCb?(8Xf{>5}{M!$FZ<h#WF21WQ^t@VZWHO}4&pZ{O^~wN*22PRH?s#z~
z(a>J6ktQt6Wu-@`>QX<LIgw3zL>l5pn&F6&B*^k@Q~44}_uf4wuP67EkiK;f*QslQ
zZ*0LS^w#q~L-R+i{7-4J>wgTk3!p-&VYais$IEY`x$3is)Ydn)(%7iik%mt%EM$CB
z=%II@5_@BfD+yX@SzXkKTL{7;(#1zh8(n7|m;P7?sv>#s$j(-Jdu<o&d{`3Bc<3c|
zQ%Gir8{GdOnw~O@FAu9{pxo9}1e0E<6q%6O{<B7oRt`HTl~nRfQUA{hCh6`%mSnHA
zBtokmChw0;@fabwyWO?Nji?9Zqw&Gw<e~vTk85q(#RE~v>O1|#5-uQAc-C9-QC=fn
zq1)oCIb_E5zJCL3VIbgC-aY@a=I6}b{>kbW>x6Hkm(z2psSt_Dp&V=X;nJ(pyjp|C
z`w+J6lY3*h?;8)I9|NA?YmLBT0t5Qu%@nNY@Ta5IR*W7n|6t_ZhjrVgIs%?lp9%>t
z)~ea65IyDHUI@#wsS$Wut(CJ@!TlHCAkVEcHd%jxxQ0@_gmn(9mJ5>)CnuXddusmF
z>XSAO=}Y!UTp3w$vNbEmCcp01m;0B)vpBtkG~T_||7vy4wnJ@LsbMLc%AbQDdS^qr
zgort>ByVECsc33fnq%cFEMDyaXkZd|xGIf3Pt#xIwBxw_n0$25+2$B3xz~cH<iEC>
zM#)!y)1vl8TDH1)rt{#r37lRQCo9B|Oz_}|G@kU#lF5VO2k0+t#hoAJ<TYg<Uc78@
zsQ~NVtVzU9Po?$Dy1|SLR=aW~&pYq6bQfpAxsY(3TgKhI!!LpFW>{pA#(4gI*f^n|
zQsef`_3jOzaqNGEW;Svv5S%RL&jro<-P&w$npDz$ivCU}KP$RjeBvkjI(g5rwwB?K
zNCR-?!bz|V7O?JEeX>4`rFG0?O58<hLH;a~nffVLLR$CY4m>IMPyryP42K5+SPu_z
z>!n*ggRUAg&n~TSHHk0guaRpYS>K0~)#m;1(w}q*+6nSvDOPy;g7I`kbY2ec(EB=`
z$gvyZ;ZsON+W{2QvKN%)lORKou@Y2MssA)2I(WS&&Ke$sp@Mjt*U-8T&>miFr%BWT
z6dta9vVD83Y&I<OS_tk;@1q1Z)wT{LDGe^g1TG&Z%q4#M&E=v}tr3P`>F#A$g)6$)
zuI$VX+9Lf+ah4w~`A@alX#`LW2_i{<r!TLw8bc&Qer6tQ_D6<H30OAXzmN)ogb109
z%6xk|Rj{<fr@ObRZFTJUJI%{9{3N7BD1q{Z2<z%%Vx%v$poqlVv<PZA+5Y}Gi$iH?
z_H=C**8Gb9jNzDJMl)5+HoLj5@)6os5Ni5QnJ9%86oAMx9_PuO-=FnY6W?3Z|KN}}
zJeT}I=|Q8>ScT=dNS%g}U#fgSFVeoc#&m%`@7odv1+k=j6kXc<_t=7GYhuPot6oH@
zH6^tKS%x;d53ztZ!yi(lp2V=ZznX=;_ZlLZr79#yNBf@A6>ehcx7{|0-m#p0+FN>0
zs=TB){?))npZ7qW@<k$W$_2Emo<N6enY<X$%;=Qs<SRP~q$F|()hhk#A&6E(Tr{e2
z%;687+U-W;azEGySVl#4Vvs9OUJS;PD8C>n(uZ|cDgfM*%{Qbw_8qK-1y?oJ4${2W
zHy*KtkodDl@cdHzW~XYY@A(DP63W9v!!?n=k}mE)NMRiQ6no%@@t+6=F9g-F|52CI
zTqNPvm489jT2>Qva%AMQ2ga9?d!q89;5E@=J)y4SYAhHzocH5QWW@COqcu@>gW#y)
z2ThLFr#z8OSc~E;wiZ^-8pW~XV$QV%UoN%Qr(OM(oX#a_DEro9bqn9D889GLTZH)(
zAL(R{X_fv`@-Hs8m)r{~{Q3olMzT;@qAbXYKl4UBl8f!z2_gEMaEHS8p%a@j>K9_m
z{G0V+dsF4E&pWG5;~`wXUD#AKi~E}T63q{Uc9qe4qh&0jmJGtDmTF7y=wv8#b(~w(
z=4}j$*EQk3le-%JgG0QfG5kl_Duc4~ijVJcwFG|uX4irKYY!qu$OFsMRD>m!VzvBN
zHw84+lapkUN2nHp0M+n(jk(<6p_+K|b^lMr`g(=%R0*0^zSzTD{c}150fzl>%c`x?
zUK>5)^NGw+3x#(UAC)*68@IL`vQF0w-JYMaC(k+rOUJC8k1d|(b#kgMO%;(c?cy<%
z@}7Ng^I+<U)WO`r8MPBb<_=tE1V8V<-vyc19maS40CV^t*?}VGJAR)?Z@KtYo8#TG
zC3iIl8dB)QGecxpxtdW;tW$D^!3upg5c^B>Xz%b*BDZ^Fa=!=mJ@s2B+2|V8xVY}$
z^4Y?!+7L%eq;oZud+$xGT9@|oQf<DLRo^2JGZRF6QyMrsFL%L0@Fafg1Iw~Tsb%kW
z_-QZ}$7Oc^8LbX;%RcBBQGMh+%c};fwYD!M0ax}HmAXGT;CMG3vUXPc27Ck%jWp(X
z3{sK(jhsdJ`I3y-`-6+iIxePEm!jlX<B9^t{GZDMOp$2J<5<*v?RqkA$ed*?9oM!b
zCS{{p`x><hjhtt)cINo51mAU+Tm2pj_ptq%%zMSfXFBz8*q|A&j(8BQ%&s{z&!br(
z%+i-i^v}qi0~;h^aw0`Qoly<(o8LDm#Lf&^UATAGSS|iH!sU9&`?{dgFzBKhuGMFX
zF3)B#MY>Y*-N@cbIakFXBB6XaZBk4<snFo7`)AJ|zq=8a<vF<rDv#9j`~2fI?aRuI
zDN{>KG~FwRURug1zu_#Au&!5GTlp~9^#megeZDD+t0=EE-T0F7W;mgba}%-V{Gyz{
z&blGvY|XbpaeAy`>5GGQGPWJf;!U1ug4&1%=W}`p<Ii0pH0G;ErzbW->5XANOsMOw
zPR=lb>C=f7ihlMZ8Gath6@2T%$^DiGg1(-A4o^K?6YW}eFfi0PqV{JO5%+wfX`sK(
zg*@ShwAjd}IMjj3*jc$pYWT@Yvk=T41>31TR)gEL23;Sq=u%i>6e1ptQL%}D%G|VY
zL1mM@g1SZ;&6Yx@og1~z%)N2ngKe?NmJ5d?F`B}y*2k${i{wV#)A?MvC!o=3^k~ZN
zPJ~9MG&$Z-6i=9?za{#fzif5P+qCOH(suQrvbd8HAQ;A|N<t96X{n7x`tJS5)bl9Q
z4HQS2_XaD!=E5>?t2q>VW+FJ;)9cTn6gNQ}cavD$qm4KFYD|b5Q0JLc{BD{8SE*J^
zncVp@kVMTcIPNJry$@<pu`-vFdJTS=t$U@Dq@x3ANyNQmd&04gp_AnW&v4%BR``=U
zm~E?&t1_P#%{3#xmRt@LhJ2OIwq9~cf!D)~LThz!zc+pb+{J&2|3=Tj#xO6}N(7dh
z<R@RK*hVK|Eehih3!@)qe56FBqdlWdwASyA8^m{xp_GKVi=h&5yY21b3&73(-f^sz
zFnNoI{=~;T_$<vt(Y;=%=kc#Elq+_WXI}01-w0)6tcK9(7gOi88tTm0>At*dRg^G?
zK5ALb9}cb(Y;%#gKT^fyp}_SrxN2?RU3;v?h@vWUIYwHIB@Ck-A^{Zvt=6w1YIS>f
z?Ys>Fyyk-RY{NatU;D!xqs#`XV+&}Lc7(7a3n3sC7A(5khl1$GEOy|WF+ss-;qj^&
zAK{*E4A&108N!n6BN&W=40Qou;k4SDY{Tk3_#&<Z%RbSbaNv~}Zpgdt9OGmSPsE`@
z_kat9gutF;WP63t->bEQ#Mq}lr4_S_FjSw~{NZUshdr@xFu9%ULH>>15bp4ZNMdrA
z)J^|+HL1yDzr_GNpZ7l0o-dxKVwj6jN_>|ko>kfN*lBxe)cB-Dp}G(ipLhVuDWUCN
z8jqqig2>d2TLDX%?d4`N_^roXK-LxFI)3&ABlEJP<;WsrSG=Dk%tqGLw}mm2B$3~#
zn}68qLSDRv*kKaenQXY^)1u~ib!oFDL+tVC7E`j_CaH*&&(Z9-eO^=kS6qY7nKp`|
zu7gVv_IGXH;yvN*zI3M6JF{<Ha|IpNPo+Y(-Y~HyfO%+MTvd0966nKChnVJ$HA5L1
zM_d%5n0dDuz3+WO{47tF#<@Ha0!~9FNU(EVdv&DPu||~!EiQWZf3GI~>laxN)a|>^
zemWW-x`23IU09g*JgK+zRI_W(WM6R2^Wyk>1=l48!E^OW1oHGyuAV<H+z?fZ28&3Y
z^wHl?8EW|aZBmK20R?L->;Y<Z6NFGDsqGBr1p#%Wg6Y%2Bc3G)XZ<OpvBczD!XO~i
zAVM%nkPALoiSifOI`4XYOh9$E$DC!17Q+P+N`B&Hl(hXc(sWco_XK4ki}WRr5gty<
z%^1tP$>Z2_q-2|4aTLKuPYY4vw$8JIX;<Y6R3X&i%uAzw{tdTok;Kx+U^<0Bu@v&B
z)z>BS6$=8nXC1l{UG<Tn-r-eD9RQS}PCERpt<~gpMX0k8K#!x-(-_5lUyNM?)kCrg
z&lU%CxNeT7c*Z_JoL;VZ8~|yc)b?~$amBhO{lOwZcnvJKxv1e-)1`rP_WoHy7>`)#
z{A0>1;c$vO!$#lBQs=PHk0;Aa>je0YF(VvX*>Q^KXxK!nbu<6ojLhP}CLSe*%jC{B
zxaTj!faR$(KFE_XPs@3C_ILZW(R9D)?jM(xOR8de8;xYQDB6U6BpMayd^Hn=QCc^-
z*#Iq1x?Lx}h#ww30sZ|9n8J-w39&a3E=@q&IlIuuq6{wy!Cxqm;DJbcU6>(JF+`c4
zgtYlnY<w=r0qZ3Orov%%(3~KA%2gS)Y8CHa!z4D<$uShfPm_+LnE5=<-}|ey2pfIi
z9@crwe*y9UF7~*|20amCD<$~KBIxOeQdHs~&1vjy(CLSN(aF{3q*BD%kogAMG;^x5
z8qCEXdDJx~Db^BQak8&O@^~4{sDXrX=NxGcxeovehYryP)4ZdNAfiz1C4c#H*M7aH
z_ek_;Q6Fu**}j9y6v(!;PV9n$aoNd03QKUj3PpakvF#S57EycIKZKD&D<LqT<XD#x
zDB{WsyujkpE}lH@k<m97w-ZC<F?M`!=}Q#V?>|rU$5Mr*wXkJYw)z!J6-q^#MbP$=
z9YlFwGT5UVJh<2kK2W4g+U!5=BKt`ef@N0wgD#x1c(OVlR;YjZNSU{1G%2I-(r8*O
z!nbQbcM5cK^6Rineo%1h`GjFS*SPJa1-a1g*nL*QUK{%%{`haMSz=7zUoBtcT1U~I
z@Qkv>viNyXz;Nwp`<HtzUv<qZx_>qNc+d9L6ln&1-{aTy51`WW{@gzmMd>LPnYPIu
zn?;|2hQ#kI!S5j+ZHLm~GWB{ZmNs9_@62l6oDFzeaBw>z$W6(n{DeM0L_phqd@NpF
zvug%4?n2E=LxP$e<vqk@R1D7Pq8(x*oAHV@U=2z8*Py^Ae49icog~5>TN#uOzb*Ob
zMVL~0Mh05dcP=(umll|Qy&e!B;NaZy;O?u;G-~}+lWY8uuRd@uX_c-8=-C$X?3Zr;
zE2`eOOp|LDVQiso8xj_hg6*%rYuVi_I;>Tc-*n;Nx+Zcmn*r<9??oa`WiwY&;2?0*
z2l91bn_2GCCPYIynZK6q)h^x=z&*<=_e-wL0)0%OC6$&l!?9nuXoANOQ5~$%Y=|zg
z=Ab~3klM9oWT?mJ+G-Am2UTr@@X9WAXE?Rklq(ru5q&u-e!ipsS|}5&RQ`I4{tJ{a
zgff>HJasap((bT+fCiT?VaCodP<Z#KQ73O!@2rrAZE#6eHXUMrS>E@>ke5HLIxmE`
z&^fdJMNgtHDRahi_MgN8-L2AubS%2mRttuRtH_WrQfeKWPjV3-v=rpd7EAMx;85jd
z8jbukNx+7tZl%8VfDU~kISm~)YI$LIg`}l>aC~P7qAUrEg>F+$n~CBfnQed9I?0XF
z5ron69RH|RbF6q}sFx;i;Lhv1Z?yIrPxL&vkQ*DNeS{2<FVSVf9B`NFiQPEL?|Nis
zX19E!@fJO(B1g<21MQSm;c>4%af4*BKK*${BoL_qQ>+9Lnk+wtI1z1oH0EuHGOJFY
zQAkVxX=mH%9ZSGobUifx{<y%VA2+n)bZ;qLa88cWdRhHv(J1x>hB6=CZl#Bo_Sgs;
zjY8Q5)0v4{$O^+OgU5NnR?Sz2943MnIe$QQ-ahrpz)IDe!htf!wz9=G4Te9Nr%lB+
zN=CLub_sw@wtwV&r9j-dJbEB<Q@tGB!()t82i_c-hRRB$PW8lc<z0i$Ol$uOexajk
zGyQ^v2&r4e5;8+5Qb?!~dpJ30%Pj8hR@UEN{(!Z3fnqH(-1zIGaps-=CEy^-XU}y6
z!q4iBomfUihdtH3{W=iRMD@?_){uYI;$yS}a915htr>mTEaZjY>xmC0;Zll!|HQYU
z%ys~;z`aB-aBF|z6F&!F9FDbZVf#fHQeAVv-SHLbG=S=vK<DAf;^*Rod<LJ>>+Pz!
zT5c=_&)v~)Z)(}6xc+CX>|d}>gcJ&Mx$8{z>aVT-{!<j@5@qek`4(U<Yh|gKl>l?O
zw6os%Y*A>wV=8pA4KNq?ymk!(a~=Ur>;Dx`0U!DviV0bQeT)C+!TPVSsKlX|kQMVc
zpZ@-S`cLUWCQAqf%S0X&s{ik>j4DdP8<+ns;(z@#zKgWrqzm<zX|nvk9Eg8^8d-U8
z0#!4d`8EGK9DiH1eDFuXGC~6upa19G{eRr-#E($0jO)~>+22WrzhIgF9}i6Dw_CM9
zX<nw+?3aW7O%6PkMi^CIzcvP>f`<F2#=pHP{%hecnSm3xR_f6fAvotTjOqjfBBGIp
z&!m?dIMz)*$E))K)o<R$(5K)@&?r$bH{W?b-rfD2&uQ~j8E|E<ma-hNT8?~CT?KXV
zt}o#gW12mv>>v5(KBpX178w|{AC0$<vFt9#DhmK5Bi|cc&x(zS62tVXh>`{Ht`|fl
zMXjgGc?4(ShQC-rCXs6u{N*Ag?)e8L3kKbO0j`F_7;-VU@{Q4NiNcE^3^A+41Cc~Z
zLd~yUy-ETZy3~7AA|v@<<a#cKl_i9gK-lv8+1$;wDV$j<(-1ZR0=a!)jWc8B)N|xZ
zenRcK3@rmSz+}Kqu#dN!YExgi&Q#m40{l=w%>B4mNS6@@>)wN;BkAv-B1O8bDn`{2
zYe08t*$3d4^sWVi{H|k{g;-^A9$<pu145fza$wt&>qrL)tr~Cy-%U~4R5zDfPYalf
zBR+z?$i(*ACHD@opS|ZjipQO(Qr{|&b3c3g8N6|lf~nn@jZ={Q#)<F7w1J{|q3`dx
zP+By6C0bLobpnY?J<i&7`qsKCl)fcZDnLM|xbgg}yZvxygsSH@aR2eXdFK>A305?H
zfPK-NTXaaEd2$TujQw9=ZRJ4%ZN4DN*#`q=*`U-Z4@!g6nHtCR)A(IOW6kR{VaFNa
z=S-Tkv)ABNr&PXFCZP-5-oZUhKrL?M)>$~*91(msBujB<*6YHWsmN}JM}`KS5rM69
zbG7T;hk}ol#ExDK=g5ZTO|)q`WdP&P7SX@FsoVHh#{0`%MbyCm`=*~DBD$|N-0BY~
zg1+za)R94YXr|2tf|abO$qSB(SBTe3Q-@11Lkm_z#sK9oruW%oYcXy-)OmaotDj$x
z&SXZ3ld=DEkxQ$FCJdDX7G$$X(H$Wd!*ElA4DnQ}($E9eY<g5|8x^)XC_Dr0`FEQ2
zYnyCtFrc|u1+hpMu<gaS>3omEbA<u8qeflI1M{t(q!D(=W#Jm=+EAVpcLvOFhEnH?
zM*?{qxxPg6B`fU%01y|k9WqXM%czPA1L(esZrQBc)}zXe);aWQPJo<hS1m+K-{ZO2
zy}6QDI-|pE0N3^^xV;3F#g5{L;CqGf_r-m|Pi+stfYtxr1=-y)z!#+eS$4S+LjoWa
zM5gy^vIo6M_a~1%5`I_N>XuEKq>c)=p}-!9i0hh*gxS$m1VO^*fxka>RcO$T)PEvZ
z^*?ZDJVbSJQ+G&?bIj)&8z%LGsd`cG-ZtgDdyT**VuFf&DgVY5=bE)qIO1>?1uRi2
zv0v6{=%DLwr}5Ebzy1iXQ<Elne%^t2PXJcWK&uh01h+LSER3C}tp!hfmH`662f9@$
z!Uw&4iaiM&2ik1+b#>9e_2`yPk0MSh?REHPN1KPO1dw)otnZ(^eVc4)xYnQUbe-Ba
zsV@CLp1fIS&@WZDJ}{B0twXptfuM$nM8dxlc}w(k-akD{hP{OOHu01D05^VKQ?eM8
zlH~iuRyL24unuqFWq5&WE|U`dtUUY^_m%kMWB0tTF^RQ7w(Ff2fVZe&xoGqF1D5wx
zdu$VV9-fhEQ2^~rM)5qC;6Z~qgz}>cpxOx>%38l@k@};fqmh)E@=vYQg)p1yxWaCl
z9i!^g!u0;Aco!jWT>!PsyUS*L!5MC7V!VjGj^Elxmca*6uo3@*Gmyusb>j=H22t~7
z>2aWqV^hl)-i4dezoc(e!SwXEw&!5>TKs*Md2`vJIDId%qg*F4s&jj56O~gRiPZ>K
zi2n^=im*K5uk-};LeNHc8WZ)s+z%03`GddYrk*y16XFf0<IdgdKFDVI!es9zOxnmH
zPPKa|dyc>7QTKj>a3=ncZw)Jqu-AjJz6+wb&KF{)I&1tOaqA;D`pC!VJ6(t*yUfK4
z6zreAp{Da?;WVItc8B3A=pe281y6rPCl7p3y6|oN{1}K!2Rwyc<glk!o?ix#k^!~(
zMK?cbdys><^bF4s4O`-_cK+e(7TDn0n69^H86cP+ipelu@sIBMoro*B`N+*QpLR@i
z9EI3N@KEQcQ2m5IcPql5>VyA0qWrU@s^}*?j^6a`3~vtVWM%#L1(7e}p^EM8xOP46
zMLYWCTs~Nn26iyU#}<G^?vj(lA_e`x=F>Ukn^!H*qOsjaC<*m~)6d&j)s&*JmtC!o
zcbN@qIRw+#Im_Bfw5lJ@$nZuS0Ls!?FixB%hqyxotp6-Se|UKp%`J`Gu@Aur0?3fR
z$mo+cF~Ibt-RkHE+u9BoxFtBLI`$vzbsB%Dy}1TEo>AyEBu+ykkQYaPe!G()>iuvk
zYKv4W^uy}|dp6`AuwXd;De0&s%SvP5dR!7gd5Kcvr`aROaTZ1s<rUtLDC5ixmJQ;C
z#hq5i<mOA1_+W*})2%M3l6CkkG8qttRXj|sR2D<!@7KDpb;Ap)*mI@CgqT1+!uI<{
z3s6`^H7|Yi>i1)A*7$&Jl-Bx#R=EZS$>I!go`qbnw4UfTiT$pj{S9``OC(-*d5uKT
z*1#dNMo!3*9UdbnZL2lvu!~^G&Nj#c=&BIO_Pxxfx|rjX_ODwPx%{Z?EiwUrrimy2
zG&9FG07LJbn4V&79`K|t1ZbD`a9VCa&y3&9AA{HG>ARu_dAo#P>CEhlWcP<}HYl4d
z1GYuC%VW1F_4V15w1iy}wH-pwfUC^1eGK1y+CjTtI*sRJ{TVm)Ah6pD2PI|AuDi1{
zl2l}A<e@H`7I~)hWrEi#yVyxL=i^N?t*LtL>kMs(l@ng8kfZ<#J#B0X0=e~RRh{ZK
zH)<XOneRO4+3|bY0OJ~bjIEITT0f{syf6I1KIqJ`TiK+aOE(q6TaoX6S$^`SY}?)9
z$|i~UJcny5wHm#Qf}+yBgza6`imL%}bAd<4uU{+T8C+1Yl@xChUQFAnfwXyg-!Y#n
zS>lW5&woo>Tr}^@izc6T)?Mk$wp@EnYlXIX)pZtJr8{3d?^cjE@9T++uYLc>xb1?I
z=QD?9GDs~ehI(?9Y!pB{40N4d2U<sc7p1grOktQh*_1Uo`#hawc>KHhPlH<D+>Sut
z|Ivx#WmMt1O>C>U>lr|*!x;3Vg|$n+l@c@t`yEm*4mXz6LlAtJdx7yQZ=saUgj_r_
zEa<oaoUjfJeK^=@xMcWXb1A$<sNEZQ)ri&SPO<1jd?>Vp_-MHBB=9FtX^=9t(NEF+
zvKw{<5*CcG?w&QPEnhgV-p^ZW3xSD1ACckErYXDpWW;c&0A$ujLNx>EfAW#o<M4Fh
z-qGh_nMTm9(#@;hn|<~}xeATnRj4@46rE#2&6cD`DjXsD!~5ns^s5KZ7VMB=D}Y*8
z=0S=Kaa(E<JRMuz9$z&`JkYLYBQI5?d4LNZXPu0SdlAi50wcM9^f}-}>qx_t?E7dr
zhfRgwjNAZL&_(BMNb@xXHZ@U6$m<knlCVqKn{K@+AvX3hLeIgM>8R*R&kUfl`@gw@
znJacjsltwVY(%!NH>+FjvL*T76Jgdc8q2If(0Cx9RPBrAh_tY7uQbaH=AfIL-|<Xn
z^r>zDypHIuKKdyt!4pow-!1lYUO%*n{J^+02^WnMjgVfXWecn@t&!)ebayj+3gw~(
z<-C_tn8f$M%V2l&Y34z~HkhtJ7cfkEakRBobUP9sm&9|??O|kIu}fNIOmN9Mco8}2
zc>D-tSbAt2yu^9D^ILQ*X?;3&%LPnI$T+xbTjWw?eg0^TiCM;USXb?ebGa73hM54G
zYMFH1T=A^e4A$xz-R>`2WDV=80*lzT1>~$zFC*0wF4sAsAs&G-9NTA*=JI#za9Syu
zpnhbP@Ddf8=d0eaw&s)}EI0;@5dWvV&sNXvZi`I~>_!!u6DGY;wN~;n$V(O{ef=YX
z<!%49bzRo4R}<qcrFYBhj)~P?zh&@Ke^d4NKJJ<(^u2FEk#==;vp^=*9f(vq?dC}P
zEB3<b8V$E{|K2$q(y&x3WRu~e6%NIM`=DDrT`u4mfi%ZZWV1-1jeePuE+&%_voBCd
zU8eZ9rzWp(l<@es*BV*a!N<y}u?mX?zmz5Voc*qZgSBPy)k90Uso?p8-koV%?K0gP
z0TJK%3108(OSdbr+4Trj42p0cm!xLbN6DXL{9nsqSrDS8LH*0~Yh+?LWB*A9jaiMY
zVLf2|{&+RC&7a^v$G@=<#Dd#@lQ}Yk-}->XK+h`Zpg<#cYa6KbFLn-Svw(uIjI-Yi
zDHnYO7cpR8pKs_#c1mL}tfnrZ=f40HaZ&$^RdBLh1oQNxMBiw{cq4gN5PZpWesoIk
zFg1dyQ0(qkOWD%Fj3Q`wNjVMUI_~6oCXW8$%z<khYBUYE?sjOCIY|mF<)z%~iW|)$
z)gR%hrjrhlJ4G5W9LfHaBW^qB!Oe<`Vn+NF1Q9M$wXsBbngs-nJxhO}51Eh8IasmQ
z(VGVsM@h&D?83PjT3n&flH;%6iX1#+WW6<rnb!+CTJN+8^n1rG@Q0m{=wy=RU`f#%
zjD!01X4|W|#L3J*+IC71As!j#HalCmvipuUL*ve${5d-+jHrc!NA4Y37b%uS_G{t6
z_prmstm&Lt>G@hX(3B@2TL9j9oEI_6mf&V5=1`Pu5QQDyN%r_W?&Mg;sC8YMT_ez^
zJIM*yctrK%J+YPH%?pYhJf5!L#n9bKMzo>Kez}ZK;40&VK&PI8AKOoL^9N1HH#JX>
zo{Xlu%uXI*O_4r6HWg*WJXkyDTMP8sYUSX@9BOi1Zx%j6GpoGf|Gg_!G_5+fzup>>
z39G&fER$+U;=Qgx!K9MVM)D=-0rK#kTTe`|efqT|g^Wu0oO*yaLJ_7YzZP*VB@DY|
zz9V;`ui6_#MnNnSmMp7gb<xrEijzd*rpWVKBwK2YnsxOTffp$;O@HRN;2u-<Xdg>P
z+)UKDSOw#4C_z*Hk2B(9Z}@v9=fd=~OB+YM%p;NSa=GQynfx}X5AwSi%h0}ft|hIf
zj`VqcsU-OJom=yDO=sNHn~J_&nDo2f&nyqFypQ|uNPeKVZmoEs#Jb7y;kMGU3eVFv
zAK@_5em?`D=Cu1fSck9B9YN~<;80MHWY~4}y@8A>w9Mzra(SJ%1iy#0t?hu-z@Ofo
zKRXv2+*DvLTE^F%bgzqV#U5Vp%oycvKW+bQ-;k-C&+dnTE;D14jE?mh@ao3RP@f(C
zBlKFF^q@I+UT9}L_-zfLrRGLipwhSRdCkS)ml45&Wp)bBrV)azgL@-Vw?e%j54qcf
z7gy*u+i)D<(F-gA^Jl1>tYPVuVJjwiktG_(#u?NE?aOEVIvh+GUQoO8F`dX>R>+=Y
zN*G<|-mTOPF!rk2Vsb_8UXUo*dVlmXvwn5?iBmx#knY3%>3g6-z~$wTG1aEbhHGt!
z7HKWL-V%au9z;sy$7xU%`m{1iAZL4@$0mc(OEUyR#u#@yS~&yG5riSLFS*QJ#(k$!
z=~5c+w!o!H&#{^yk%TqnIL=x#-y~ypiKj<L1%~ZP7w&DI?r6hs_wr)em%j!|+Ygpc
zzwDLXvZuUFMxc%t^#(Rd787yWdA)vzymgi*K%PkADrlYtruY5(-q&!x)0ERLuYT){
z^fs!XRPsG5_oK~~H#JzvD~=pfB{R5C?N&M05btohozxYzH+5Kq2@G!A^G6g+tkYMV
z8)8G<Z^CV1V~UsMm(ZgUJIdv0U5<g5H<e9_*wOX5cPTy7#L12ajpESJME>NQKq6P%
zTJB_4m(6hT!1m5WBZMeOBqlp5h)Fgk8Nt$+SrLs&GvDYZa<1nq#=*<kH0<b*U>!+q
zjcYyOm5sLz;*5;ybDACeFu_=(h1V*NRbJ)?OvUkPUg>Sdd^J^JVo$3nHjLLQG@R#W
z{cEM|`v7O8>(1lCdU_?fcIj@v34JNkax1(u5LME0laT)PiJrQH_T@Z(5G^7=q=LLP
z5=w_`GYk=5YerT-@p-a-m!Q33e_;V3Z8D<3E$^^{EJ+JqC`c}JKiL^|2IoB~FnIq!
z1)r&Y2aYWl(EdliZVCstp3o%a@m95=0y}P)uR?+!j|X2rlDW47A_=92Vj;QPdy?pe
z6})6Fph~oFS{->TNN5rpiK&jQ$MGdj&GV4~y35k=;4gIVRm5SpHxl)I<IC-}p{k#*
zpe3=5EkEqS!rpc^N;Qzz4g@>!&OA_l5p=frJ+y<L9)H%{F1wEd(~i;Ba=KqExKwVJ
zJo(h|U0cH!o@@Bnzt1cD?bDt%@vS;%ny#v%wurXs0a^T@`iDO%%dcu?iMFc^JouWG
zM+)&<blmWVbU&@Iu8Ld*03eQECyAThr|I>fE&2g7(^hQAnLU>nGn=<sb;i13y++sA
zn`Z+x$VTf;1g`Lwp<+YP!ucLya}mo9{tzkSLb)cVHJWF_`=U9Qrkcd2v~AoEC(Ge9
z;x6}_RZd)=JW!$8B_inh$$Ce>+~IcXoIl;3FV-W2Poz)%^;f!rnqpsQI2nm?2fX0Y
z!Su_&skHXxl1)BS%Z)U|t`}VNi4i#U+-~wel03(A#<;1(4(i9s_zHJF=fyOCxrYBd
z)Rfs^t>xe&X~j3ZphFjn>0jw0c2y>>imlniiYF;At>S-i5j8Ei-tPU9xL&5bk*M>?
z3L6cMOSFT>F!T1kPGAsoZ2u%T`eH^!Y(lOQ+83PHY2gz__5{oDiwUc3b!O`7kQ03q
zKOa}~$zx<yC`Aahe0X+gT5uPQ!Y$eG_6gIVTSl{jc$0l5ZNV}R<teygdHG<Xp`)K4
zz!jlbL}2B~#{OXJv*SY+7oms>Rk)UQ4<l#r@LRmrI4ntvD8(eg9mfU3PNuoqUvdgY
z0k<LrndybD@KhtSbJa7DXQdn+M~3biv@ppLE8X{^TZa`21UmxRuV(l^7I2)FJQ2cb
z!R7u#S*%&+aW=VE;mNR$)PMQ#+J{G;n0hh932wsBPd}k1cbz5Mpu^z}8)s&_N+YEh
zs$K1wo66=Kr#F5WAeHvkz4^f&>4MmJ?MvI8^uw?piI%4x$k4LfleTI@+?qb2fP=@!
z3zivI_>;P=D`8sq+-rt2mteNrLKh{7TB_hvSI|Q^btkyWXnO6T=es68?jIpOFXYZ6
zS^sMBzA0aen;xT&Nv*BPCfnq`P54E%%?D3!Hd)SvIW(BI<{Vo)n1oP4L{{lYS~cAE
zb|;-HzuG+3%N|#RW1FPce2I0HizXzn`NzH_-NSokqnE%5f9kd#>_sN5X<&C(rfsBB
zma?9?J90#)IJrNAZ@R*An$zhw+&cEt9o&0L!uptKj5&<bqyEazHDs3`;^Z*RHA9>O
zhV%lBxjt|bKi4X^@^|r%qDtu=qlqRI+cP^dux2H=OMST;+<UTKTRv?b<}~y2#Ry)e
zw0LO{-Oo7-zMYENaWZ^QQwATRy9h`1U#44DXXd)YOpwO$br;29+g*dvA4+J<)EGbu
zdD3e4skP?&&$h-eijvtDv@Z=uI5!$m)P{RcV#M2MEud?J1j22YZQm39C){WBmb%-{
zLVC=MLo=Gh#~oR)X2nCCM?RJkefvn)faluQY6i;Ip)HsX3o@^=yW<2hC3NJMo@>@6
z3BD-D!ycX?jy)z2rOw_4iMe}TX$J=)M*<sQ=YPZl(|>{Y{B+g8{y|QQ(yz-F!Si*V
zF>3o9i#%R#*L$&hdLcA;f%(OQrVyi~xUWS)f*YU`w%JNY>x>|mh{X9Vc?EfrJc!lY
zF_pZ{UFHnlR^m&9<Fi1^)X!ieTomC@L&BkqyJ@p^Y`xu1AhwyQ{O-HWNajJ?JBfti
z^zp`n8m6g|z{^NBNA2tKr=Ex7)s9S)BqT7X(5Ke(R#%8SN$VZ18S#L4y_Z}%I`;UG
zdzjfoA7cqxCzZi0MLTt`$1?91%PesG?x)5LQcLk&S>q=@N(mkVeVOM`KZDq+Tquj`
zU#f76xAR=9+Yc<ZmtOU;^e4xC`>PQ$!sRQ1lSLdKY*D{VW)VsHVuX@S-pLBiE&Ylw
zxjiThf!-TOZ6Cbg2~*i$$fn%0+R}5#Li)4yz*815Ecm)iyUUWM4>+)5pVE2p23&yT
zP~qO#Xp#w`agLK8k@*sJlC_ZMVbZxqRd$bh=DP5AZ<(s`%E+H2_b16Tl{{n@)fV4|
zMQf+dt4zD@|Bf0Ad}>fVpsAy7TIXfuo@dK=E_ErhZuBy5{f8?|x+K^0(Xgu-gN`_I
zzU|7F^igv7!c<ZP+pI5WAy2`dBV4hUj?5FfaDN@I9i;D8n5~_Ij+#};R8<8j^;d1v
zvyuBxaBaowhg>BiKz{Pzm8k2Fn8OS*v7+p;t#3`MS`z2{o`G9LLlPoJVrj*$MhI7|
z;kh%dcC01wFy`kT8jJ0`lhCV~9pAJ>`JR|0MYN;>WuvK#hqnKmF^Q1*>36*R&Ls6!
zK<EQx+&9ICOHa5r+f8G?an}U9x&`3|RDJ1-rXsQKjb)D2GbHjx(pnP*2Lv}w3nLBM
z6>gPoA1f+6Guc5<79=HD3S)Ep^cO_n4X<t+8q1JX=*O<|bcipJtnHc12H4X_ruXa@
z57>q|o`dO&W9t%e{sOGIJY6A*O7B5U;InB6*RpWTs<Ed;TE;P59!DjpP6xRdYgZno
z+SHok@5TDF_ghpK8}+jH+^Yt8OatB?R{;hRW)po@h<~lq)@l`xzH-{+@u<4ice%`{
z7nuW<+QsciVfX@}>EXNH`h)%N-xQ|;p1O}zS1k%Bb;ZmQ@lok!P9JO+JA3yWx@1Q?
z(F*V*&|U4l>n1vS9;W_by*vq%k4!560EOx=Jp%6fjCI{!`zVtfzTWd?&{HxrwPC`f
zBub<5JRB15=j_nXU7#j>)2QGXZ)p7Y68T^p_Q=~?kL-A)gZ6S-*>Iq7<mIh8n%Jzc
zws)XU;pJ=*gv)yvX;{)HFx?v_)T{m?<;ShoR)(_9C1ZbX11YkO<oU#|ut1N!L47ki
zd(1t?uv8U<Nqi$HA`Isco*6FpaXT1TBHBq5KC|{V)g<_zGvw$guoDSTJMUpv=vRfP
znjrA|+vxC86*F3wAL}!4GR{Wh8@DBX-d@MT9@2-VUz23Kt%Dc`Z2i!0^<aNq@6QDq
zbQV{okhNziSkz;ybX`2!L=1A_NIy&-wC{ZuJp(Wvs7elXqc#JcY)6Y{pp%Zp&k1x`
zr7>UqXwqkDhI?FS4yRr}o!cmijzEng>9uK@Pnp&D?Uv|vZX;2-FZe}j;gc(2#_B;H
zJ2>_4f8G(WQh0U@`eNr>PLdGP$=?3h7w=pqjY++mR>;CK6}&=>1FS+2+Lexb>EMPC
z&&y8u>*O7w8ipRu<y<Er>$s9T#cP?~jZI!P`F}t}d}H|FWH(h0?P0yxCu=#0wMOZ8
z$Y&i8EvSks9UPS5#~_c*-3{yys>WavXQ}cq)$7VqiOXMMjfUedYgJ^m=cy5%Z8~}Z
z%I&z%kCP`XFG{Hl%?~;L-tU#i9&yEnG3jIOphly<dp{(rgTC?vVhCr<VaNdR2oY>B
z2#-95`d4sAKWM5f8m7}3?Qw@bC8LJHi@4w7(hq>os>R`zf?wqghWuXn(-541Nss$$
zt2Fh4WE3`w6TVo~NhEl~p?(<D>=!4yyAAJ?v~z%sN1T(M+8z}Pr!nZ4X6_&zcHqDw
z=)l`pYliaRM6m=&!BuV0&8Y)Ijmb$w7NqKsD&e<URkVX-h*~3s!rRhY>%#TtVRo@J
z(`JGJ%R!~U|2&!wMo)?tv9+MaJ$W=hM9eaQ?dox*=K%-TRoG;Y{+(%10%POoUjaUV
zj=db8tvkg<z?D9u93tbP81Q5&Y+F8Jq6(Y(5Vt1^9yby}766a=H5L*Z!oG?heu{tR
z9RG?V6e=2<NC0}lW$s|C$Yrn!5jHR0Z#ibEaJm6ds_huiwmGu_@?CL(4bZ1r?vbw{
zzYf#$hx@^MoKSM0X*!IX{v%C1QlC=ZW}<Ptv!syL9_*q5q{@cJX15czZh$@P8q!>o
zv6r`!22>_a0?Q)WoC&eBM4vkN+uRzjDk__B0}FAvjL9M+v6va~PX|fje7C|0uF@>Y
zij~fr$gfCM({N@(;(L0;cE-!IU7Pj=JJ8=nqD-h)b4vaG(Mo3bcA(fHAqDjz%k8B4
zoZK>RT5<=KsomD;ZmN8;kG1_3s@uc@5}zOG*TZXoL)RnOZ2HGuRr*#y_0QFf?rW=#
zfT;1(zg~9#?v#(nWznKBGd)3<rKJF^Xq0^JYS7aCmXXBX8bkv!feMHVyygq)qo<gk
z^dmEDwHLMlF<>OY<<R60@_zY0s*EC1)jUzzb_n>3<XaSim%o3}7kvkPWt~@&@CAQ5
zvU&3$mCgX+_Dby@U%T)S495VlwK@Dn{ln<XA+Se?G1NP9%&ve`aRU<=)UkjIhp*}<
z1v{WKqQ4HX6d$4VB^zQnpVzdaUbuwbhm6<&Yk)*ADHQU=S3EHKJ!p!w9sAl**KiJd
zLp-4m*5|M_JvMvHv;sI8Ef%L*AZn0ih)4{gZAC~*ab7I!gn2yiogv%`C3n#Yc%rv(
z{;5$wJH_yTr$ug19GeU7JTwk@%-yEaMvYEDnMWy(#ia@fHnThz2pVktq1{0B`9CfI
zm#F~d5{_}pq3$w3m0rjibf%aJgx7erl(}i0du)E-4$9~BtHdvsvy)ShQ|Vd#4xS}u
zE0$k&JuWO`P*(#H*TH%#APKX)aik^I41=Dzff+X!8ZDI(wu~~hfI_HGd~EQLJ}=OM
zPMuj_F(jK(pzR6F8}!C4GD#Lj!<&k+(}$XhXBbYkK;Hs+xcvN*YF->6W(0<JR<Qr`
zv7wkC`F1}2<?HY%y03ITguyYv0+c%$%i6GsrR_juf{`C7>>GT(?EteGxE(*iyXWab
z;0bMHZk^FC=omo_!D4nX4|Kmgu8F(DMP1(rk=Wywu2wX#lh%-m{*cmcDH}pN_fnu<
zd+)maB9iX2;t-^4FD6F}qP)*9MZ^7&ii=yuIAHgj`<!o_pNroChv|th3s>3b_~%)t
z0z=O6oOG=IddZ)Z1NDT~x_P_WTOX=86>4UosoNM&^DZYd2E=%@yxM0IZN2-BU9vTd
zS1H=3GcI^?1V7Q9ioQ(Ub0JiAj$GO`jxmI*9)Vdk?zYxNOxM?=CTafrLvLczK2=_p
zH<jZ+Y&u)mkd(bJ{}**5tnZ|W=4ZR+=yzD;y<Z5tPj<QvdL*gA>cymgX<Y)9!PjC<
zZy3Q@*HPQ#c;&V-+>m!TmCKXxMUOgbsQUof!9=r!_2yUTyH1)RJ{#wE$0%WihcKh%
zbBhpP0a2%4h95q1CORPf8Tw<3Djkg7%}kiB3nsuTYQ2GU?^k;dhXq_kAC4{Xx;2)K
zG-^QM0i+DN9P_~;dB4j^^eP#GUqgXr?HO90+?~)<VdnZ%IPrq4V@2}~x_;P++1mGQ
z(M)#(YgZCS@5U?s>A66Ic)>Z#xI|Cv%o;MTA_%V|&$W`txmNNE@#kP7#J_;4|NeVK
zgyazVU=X_R8$5z0yePnwEXwjMVF(z#OYNxvg$ib`)nwE9&aOcfDQYBRNKz7>ge-GG
zRQ6C`V9n;_d+ha$iT6`4`TueX|HtannJyViEQvCw(Hsyyz{3#(tsM$0GWQ+D9EL^g
z`GxL>^eCDm)6R0a^P^X_D4z|V49VUBaBzKrnsRY+SrhrL%1NHIt&S~zE&lt{Dnv2~
z^{0<(*@{U(v8AiY&-)c>N)gIF$#d?<e>NJ#l3xYGrF+w6>RfEAfNVf;V_`R<>FAYz
z(8DZu1L<GmHUFfvC{iSc7@t_If7rPE@1sLMsEuPrn7cZ)@AR#Tk+FcXX%$}1$Y>t<
zUFSq(OGnKmsE-5dw`Gb}7-@dvN_&@j&Hsva0kBH4l?V*DKw4)J{&y_lpT8gu1z;Av
z@9WCHKE(e#W8mw+>V#(2TIKuvAK&@wKT>&!V%E^rC;u;>=rPERW#P|Zviknx0RJ}!
zCixLXtr2P%|Krm6>t_uj0sJQ5LHPy!udn^%qsh_(Zta;~{P+LM&-(xD0AbLnynVZ2
z(w71_E?g3JyMwiXfh`THhkw5vJ}YDyvydtoqpF9#2GD9?y_o?+57zF9tkYJ3mHz0)
zW7PaCVB~QEzQfO3D8gvJdu<-L5+@?acv?V!us?!=pEZ*`<&p;lYSjT^>Njr>`w}M@
zMd3H*a(TLEbBbILA3zP>wgF=>x{_{yLyjsFj3F4?2D7O)-lLXaJs=+N9^mrIw#ubQ
zU|6aF10!ZU0eG<kaHnI>5<p(q#J^UzM+w)qK*rf}^y_OxA<EM30zw6~sJXs<V0wgg
z?MB|lc^lPxkW#5eiEVZ@4I}txKm(dMv2dyqSWl*bY^lqZL9C<ee_$#ATFhz$C;mIv
zFA<czR33T?(4Km|L(rcYuhGWV13DdNluD;BO?cQ7fXKqQrueF3K_X1scNjG8PeEw9
zDC@vhq3mv!TAsq5IdIhWS+xIHVq<0fYCTXX<qi5#byppcTZ7sDS{4PLX8^{&i1L}-
z(DeaSY*k%DKagUQ?m3ElQv{m!3bUaWlsd)+te%TLa@Oku`_0C=5}4qQ+6=wDpf<n0
zn7hd<>l~NnPxmGH`~M5=Q*bjlqUSCD*Y&A~j=ZGeGYLPrd}rT&|2OJMjFQ#V0LS2X
z4Dt2_iUdF{Z`ruGPG)oAnVbM>D=IAP4ckMpdUc?&`@rBv^(obg4H#8HvB)?J`ZXw0
z?qW^iW&`;^(E>XU9*&Y6*?`vY0DJ<>8i=;fbpHF@F6jvYQ*q9-3LXFboyYQM#r69i
zP#Vc;vriG1Kb%?50DVE+P!v)Wk_k{jXTcdee+Il(KoIV>D=W>G8oxkwNo&9`iEh*+
zhOjggi@>L>$~F)DeOL5;)|`t>SjnAXgDHIw$wXlEImwr}V$tH={;NFx?+3PT4m29e
zY-o_x5M1~O1)qg>UC*O<!&A--8!%_X4t`3<Zp??y7)d9$2r$8ILn>Z&VGH=mLQO~X
z67W|&4S#OH%~?5Nu*2nj7A7I!rcg4rck7|oLmpM=kor5P*HwVUG8Lnet_&5N%u}S)
z?LYy6U@ll6$`d(DN!ZM0p)lAgEJT_nL~s;rJF?jwEI=#3xGl%tt04EAJRx%2sLJLy
z@1vgTtAhND{R;trQ&u%&Ql4rQhU9>|;YcT*oS?AvUQje}{v3&88BCaj<;$mYHl7-|
zbhG5csZqWwQn@aF{e4ECUs(DHxWQmrS1oD|DQfhUc)6suM1H0R=<V5n_BwK-^tEWF
zR}zQBKMvGwhUhE#*_RtyD6}>O>WczPdC2h%-huuhYRp3wIP1ZJ`H9}A3s^27PatIA
zmqN#k|3GNa+lq1yPu}EL4vJC?Qwven7hR5)o&lavDBHcD;Zb99=suXoVviaa_th1X
zz89?Nvv+{a@f+PKp>(WBB$a68Q~_!m)~>*)Zy-eziBEZCxX2>2^1}&4UGOOddU4|y
zw$JuaL*!;Ss%943#d-*!gN)>**2fVVuHfn6uH3C1Y;glF?;X&oKO(}j29scB{T+a?
z|0E+>8MYT-YPQC*ENCoeCTApf>w6zF(F#bUjL!g)6-45KHZ#E{Z9^I=cQ&t@5_jh+
z5hoUT>(OqtJ~DwcR-7&l@mBnbo0H~Wn3Jyo<O`k-<O<gCi(|5Jro#Gdf_VFG8A<ir
z+N+v-)|%Ck-+N}3$?q3^Ep2(-pGj0d@xG|AzV|hRGVT7G`-N55I=u09-L#6)+^62Y
z-@amtH@m$>(s1RZeS64qlg~auav6@?`m-j8!RGPV`4z^uf;)Z<UfyrXpXPVVCp>5^
zd19}99}6=d*^w7>S}r~vXzq_MPGy|)=Q`~nGStK2MtQAWr24Zc6G5s}<9`e?X3xNL
zf-<JDAMCYN9MD7N%vuIP0CpRLn)(4GJ*jiW*83urMm@YrcLZ4~9XAE$m3y$u1<=8k
zQ4-Hx{E%m40*tG468lg8ENhr7uL=doFADDf59osQ-P&;Evj!n9z8E0)?F(x}S@B3^
zp&WbEk3XgL=0={LfoI30`_V$lISOAwY20vY69g_Jm{x&L&!fH*oX}r;3%+-spt1t;
z9r<ks{8lylU|*<wX(Rzp`GtwH>Pp9XfuL3WDMf!BqzAkO$_G#*dR`FF%$et>*Wc<^
z*agFwew^%cwAHk@CUHl(uu)gb7n%-a=;Rtrz_y=h0bX!#2h4anm`;u}2ee@KTe~%i
z>?ZmQISiG4|68@aygk)`f@U-1aoi3j=lgXJ#fCF;`_f#0v1`pp-D$k|^sB;hWu`#u
z{rglG+ikgq$6aEV%{13mE8l_mo{HqefRU1QwcpuW0Y3K{b8dx0tHzbj%L&)yV`X=R
zRaLd+(ARV=?hAr9JhpkM5{@aUV#Wr@D|fp~k2i05xF9NPnDme6S)>?)IWCpHf4(_t
zVOlhuf=Ega^^x2y()o$^%T75D&zEqkTdUl`m_rjYY^6DPSMQ-5(xs-bx6$8&EFEfv
zI~$E@u7%-bzNA@hzSbkH13p{zmGdUJferHC<COnr?ebXxtg~3EBD+!I(PDHm5)Y83
zJ9i{hgQ^Z%bx^q+O09yv{szV%3eIuwne=Le652MNtsWwsWQ6LWP(E^2DkdLZ1!`Y_
zOIM1s1>FS?)lu=thdKf6V7!5Y?4{8@<z`zIIXyLXEDErj0T|u(l2$2jXze%xmrd;!
zK*C#xhr8=RnNA2@pi(v>I2ZWs;ClyvH!EJn3424^I!xZLNWr*)Ly2bgQoNjMwkCj6
z2)w+0?Fn^7X(Xjx09hS@>26}xgaIy%!>!_RZM=`8K~3P_=--g;iprkH<9bSFk_6B0
zLsvoH&y~B-1B;`R2et1v7G({__GV=U^O{yIfW&RDMequJcOJ<NXVfrAa4lkRzZJC#
ziZtP?s?#E#U`o$RDOqdt)J?AzH6)l7@e&Q5+>=ImE%#0GQsK%oNvx?G-`lnJVlV%#
z@2N0T1Xk3*hnny2!p64i-!ReZ=f}I%%{9S?F_bcjjtpL48LU~o;kS*{;x<hoCG44g
zkB#2dEeZMVV4bkO7&!1@l25zFavKBDP1`_91Lc6yHNca_KkNRK=6mFwcP%@!Y-@Aw
zMY+5i^==)%+c~pOYIk@<y!39HdcdHbX?NITh<KN)zRqi3$Y%T7eVl!>Apz_yJ_(2g
zX|Lmr*RkUGr!7L8wMNTRUTAkflmdDT;a~$J3(IF*BJfXzPiGu_#gf+g8@#Gr+o4=)
zoKydrb)@KXi#k-|=CX{Yn~DjOhyB4&Xi`v!oGK-^o}}hwxhGuIV42c%*<31T`>9MT
znZY}DLxe4!J~z8pX~bz}ND$d<tO<7B1kkO$`_zjH0!DqbFyE$((lkh=H+nh#1vSA>
z=b0RPt3RKG{29bDhGq!+EA}`2rKn}_9m?4_9g{^G7z^~4BM*!TDx9ck!BPdJOnC%g
zp7|3p)CkexY1|sgdJT5c+vT8py+PlG)4Y)`%&DR)qZ9DT#GgP5e(DZ)hYM$Vk%{Z{
zpcYFi3TSgmxH3(FyoOoVHLmxKZM4G$!X?f8=;@7moWp{G7+zJHkz=Rg1}gl@gK_RP
zvg5*7rq7-F1D<VQDCl&5Y$RdW4G+p9_vj_z$hHm|_aojo8g!04!FnTsC1q{_^F&gG
z6f8BpP|P!38>;%Ry+Iu7ZC9TJi=Hpq;4>U4`GmS~2UT?__O_-nnl#EKr3u;RC{tdW
zg2mvy;&Wpn(!KwOy|)aDa$Wnz1yMmzN=a!c=@euLNs$3W=@gLe9;75Bq?GOsQR(gm
z2?^=$k{r7EU$=YjwfA<t&vP9AZ|{fq!*Y!Xa}Rf1*Lj}55)I^<{~#Gm)l4)U)XMpl
zDJj%HLP2b_`CC$p$(vnjW=lS<nqf7gL=u}<n=pTemc*dqEl)nbcO8{!_@@t-!gI|N
z!qvu+Z7dOjU_}-BO6EA4Dyh>*6G63u{%b!{)2H*{lG)HBB0pl)ETQ|+QmXrJl#fM8
zyx0mQh;L$J2VFKj!S|C8BlYRExS%l+#-fO~C+48qdMo$VRbSRGx+vfQ`}ZLbB0sFL
zlt+qkisj~aIsd+*%3x$QRNG?h(b06TCsAeJlhQm}()_-r+JsX(<M>(?dh$e~$z|E(
zIBi20iCK;SRicOcCQP_+Z`~~=l=TnJ#V-p*ylI}T=TE$@e%#W=Ec#hGBB$k(hxbC0
zg?|w7;xbE(+nn!vfp2+FO+&EDvbJafcb~W7X0kqo56l)%>nOv?NY7>Jszx$wjHxhD
zrp$^O`twv#Pqpv^B}nZa0v~Ou*U!PN-Di&xGa4;hpciG2#n$(k1OM&d1{KZ0HQcH7
znp?W|UokXeLjtmT9TIhEcpn}E@yGYKd}1*xKjK&PO?H2<d(d?T6jz*1SX=0&;@NS=
z<K}aj(zSpKisV2UY+0~=8-+w=iD>Mt(FMDY%@IH(+YDvTK+r3P#<TR0g|^_YRJ=w4
zeuYbogW@OVYj$IP72SPK^MT{+2=N1`$&wuPv#vYYZ-%?EC0N<VcElZXKTc;c9~DUI
zncG485&ZsFe8RQqNzjSdfGoHkM815I#)7=_>2T!MhpTS;?F;r9mU>lZoa-gt>r<JL
z@d^j=)d!_IQ#Cg{%Z(;N@+_)(Ew^e0WxAZXvZ~f*-0%5&H|i@}_a>g5Jya>B^j{~g
zVK5(Z4-k3r4c)i9cc*z?TEOU6Xr7)NrPs}cC1Pc5Y4z?F`8s`OT6f>0_aJh2fO6P~
zkIwkl;~QS4l$xi<$&4riY~jKa7aWX8e3ZfDCph{6DCz`=Le3iDrA-nz`cPi2z*Y*`
zRNlp~%vILx4UJTtqE<z%__%A}vc-we4BOprq6)Up&oMfdU9|@x@GtbgJOU37679I(
zN;D&r-G-xPN-yBeUQL3)<Ac|DD+*DeBVt1Bs${p3ZXjLu;fGW`^CD+5ykf5KIerZd
z7(DP5aSsh6MLZV6s|mG}X6KdaZus80wUmD*#q1^#PYWFy!+ZsvVT1UDah2<E>?8&X
zLI2}CtM;V;KbEp_)35g@N%0hH0ay13U!6mN2x%LR<$)c|1w9<;2EVHYUeBJ!?(OK!
z!44%w<w^p0dR^b4c7E}M*_24PW`#Dn|N9K-F}%i+N0`xWia+*kO{pS7FYc->Mpv?{
z?cBe#XyP;>%+!fst+^-ERz}a?wyR?w#M0{FViD<qx-e>9k2!DGr~FUtnlTYn+4)xY
z0E_7t*l00@2^BukI_{Og5C7mxVn2+sSz7_}#*+_cIm~L^Z4t;SB$<BQa^i!6WLofP
zH=f%0AhFSmY5!|52h59YCW%mkORZ+60g~$8Z;n}J+59cnEsfY{oLn~g&6ICFm%<>x
zPKaF4k<S7GJNxYD!auap1`9IGN%|&m+1%4kgtwG{$jUc@VXBz~q1;bixeQ;o8<mbt
za$Y7Drhz$R_OvJmV`A!#Ee~LCs=7fMToE@s^uT0C?Us6pFyq<TSb}HG8eA;lGKl0Q
zk}EQL%eL%pE3?AA-h9O;&;Kd=yDigzT6>RIW$CvAD6Pn~!n^q`82EIGOn0=s{1K`M
zz5-dFW;ChgHDV!a;Flhv_LXB^i`Q_Syk)8d=gWLh>eMYV`;XP2ce}8JG353jcqH#6
z+2D(fIk$z7kqTg*L9!~S5B3WHql>eCbNqZRE<wq&mQpe#AExd?d#IBuLbW6yZGP8a
z^^z{$eBMbA$g^yucc&$%!Q|af_!79$?1pm~bOzPAgGS@+Tpar~P#Xor2c~qeBWgs5
zr<26@OD{i_P#Av6F8iAnnHOBu;{JXpEkRQD^^lUv!IH;ns^3p)hUP}d+gP~cdUIHF
z_?9`aF$~e)nQm@og_RIfnShz2_UPJhZhSIB4o9=DDz78x{HmcLyt>j()B_L^pkIcx
zh7Hn$@HJ)YFK>a2TtPQ1?GVI78%8LWn-re}tPa>Fn`Jz00F>HZ*RvQxqbCaw2#mvl
z3&B|#P47;DuwM1cUcNKH?VbWLrRP3^HfQM*@N)=#H4A1kO^V$K{;<>UYNuuNoo~9d
zfC3Z=fEyzGlj@lsh~Ad#*DI#_Xa(|(gt)JbU_^;u`;kp?+I{dJmL~t087zv3h|Su)
z%gg`|dvIU_5J7B`0}8mXZa>h-{qlP&p9C9|?-7-i1JN-fWfl|mnf{NRK}TE>F88EE
zaQ8dS7LfYcfXs|SLK!6UIY_T~P#m_Ppz=K?Yz=kL#*<I;;`E>^t_%26uP%sC!w<J-
zX66#VFN!}X2e#e6he|HUIsR(U_NtK1*8)SaC%$eoKsvP*#MNUBFtCGg>N(tw9fbHZ
z!-J3_ez3hxK@H91^Oh%io?$mb6@D(rI)fqeut^v}j|V8Q-&(yyv(nS%<i`$o#woS~
zUpWNq{=4A^IMkkeQQ(1fuy-nTfPKJ~;fUX?QTO`(JXy%?T$~zy6Kw@w9>x(yeWgTq
z@H9he@Qqgv6zq5b6nsa`Js4Q>@uLMEFfPBlWWR7M0!^^Ir6_&*-DoXeep@}+Tl4FO
zmHd%eNr_qtx!03YZ4#p38keB?ZwJuRG!YdW(roE=FUrB5)0qX3kf(wPVfm4=EAUDG
z+0O1t)Q=fjwNr4iio8=xeyASIBs0pj3wgG|KYKHrg6YW#c01)wYVw6d!bHI{(BJP-
zlZq|se)#U&=d}1qstVHNV8*qaVUac`j5z^24VV5su$VmCZd-_=+TqoN)zXCCm@k?)
zcmh-^Y`3YgsPm|cs0x7Yr_On?Y88ic{D+T>2FMUiBO{K&adsLvp$8xE=;SzNKT(wA
zpmic_G~T}seILEbP=>kCXM3T<0FUEmNY@c*PXd1a;KS);$0XO&RI;yD|6Du&lKVar
zO<_h5l`&$5XQ+@s|7aY6=C5!9p)A0zx8b30J4JIgE{cU$#=~}=Hg33o%SV%%8^ni}
z#a~ul64r<klL!N2=O82p2Cawy`@^I?oZccLU-rOqHtn3j9RAS~lYs~L3b-ow8}a+e
zdq>;eykGST&Z%o<UI)T`GB;G{m-@A+tJQ7$X!|EMyT_&LcI%JY*l1l&+|R++(&f$d
zq%WsAZ$G62m#wchix;HR(Wbb_M)(i!Q31ze*UxK2&_W?~E&qq6;A7|GV!pKCye58v
z7_~X#vV<Q!LTBcWO@2On49p+4m5AlK;0pvY5vhBzSozPyJM=W7S+yP{FTe5EbR{XK
zUGodBbbs_`ly@@@gH_DHuWwN444%i}YpK`$b0MLjQv9GFhP!`yN4%Qhb$Z^<|DlBW
zYgqaG{X3Y5OeU@3KbmSARQT%_94EaT&YXK<4+GLf5h&)lrNg^OWtke1M@fGhql<9S
zfC+mnPsNwtqxus($~;1@27soI%NvT<NYllQn-~9h5H7_L?rQ%t!?iba;jbhA`$ftN
zNXPymV;4enJPk$O5l=%&k5?u)hm`LB9W(*uN9#2U)ZovL{+w<81t7dbZvn6eNw0*)
z-+%nC<AV1AfyF7fN;m$z=l;6+a6A!Uv0tb({tu$op9dAehs5z1+-YC_)x!VdLj2dC
zz&8Tb7)EGu|6fqmzYp=RuNb33fQF}};{Crq;oo1(g8<gxm9*Uddh`E2uIxJi=Fs)|
zfARX;)%xrFh(I6=QnN(qf7I&#D;b?cLFw-w^{=n^(}Q|FDX{E+!+TJ`8H^X9`EvVz
z`OHW(^){<vWn2ae)M)LuIcHFo0q`SU0aL8%p!ZPp^c+aHQL6W7c!y#MhD%*Byd5E6
zV9}Ecau%e+mJ^ntWE@C_v|$4f3@0)HM|;3-bQ0Jo@YXQNjMN+884d=j++-IZ{lvdt
z4~9=$U?3?UQPvmmVeC?Q7&0ok1>`~pn-e3@+>|$Z;|PjEZ~{}xYPrNGm5^jVqce6b
z&a^-6z1dV%>DfT?<>Dz2jg?KmHES<BEf6i+h6Xc`M|?S?j073s+-68Cpf)@L%jD@9
z?U8KkpI3yN9y&R%%pRv!$tCdhL!gEwsP>4>9Dpjg3M{S@+rprBS%2uRtf<Jh>jJ#E
zrhqq0G%~?q3<*rmkDnNHE!1CK<mcm~wt*e4h}J^!3}in<tFK$$hXN;9h9{idO6rw-
zg=A;64I-7yUrBpFF$2*~0klyJm9})1SvU{3!5s0Zy#C?`W<YD6IMu`v;1wNiI5WP8
zzW~~9p-mu8(9(DbZAX7yc<6;E9h8p}ZuU|M9aR^(2O|IM!2m%ev4L`T(q+3g&!B5w
zhZ8tL7f*nZ<(EjOMz0%N0JGXLXjr(5{j}yOU_C{fu92K|g4yFn2#A;XHe@`Kr!{=h
zI;xhu85$$up#>&afeMj^u}1rG^KavT`s|3>18i3hl4>F96L5>4m;`#?6~m5ak$;+3
zcxJ&<j!orWWo1`IK19J6?&|EE1oOycP$iUyF&kxxI&4jj9jx}#I3D?uTCPItu<(Wh
z2!NNByJkS-kX6-=YT`vYqUEI3E~U(e8k?5`G)glEpb|+0o8=ZHSc3fLs)2KeCdjC`
z?PR<3H)M~xZx(u{5%{04y?^))VFMKG6}!8(Z1?ckxZ-7jjVVV$Qfg5kOIeHt2s51Q
zWXji+y+(DXS*j4g8;^CH(CXpi+x1t|Z1{xSJpv^~ML$>MsbRe9rh{hSw>UsuCDd`9
zTkc6*2PyBbSTuS9q$WUru|`x?3wVX6F{w5jcXKc&0vJz$!mbkXV1c|OSU^a%P%r!}
zAej%?L{UUX@O`j>3=OJ)berE-l1AkIfNf53<aI~ud2c@>a6tt9KzPu2*zq=eQ(ib#
z@SE`H+vC?gPiIoCeDvE5=Wjc~+=%Idi5{?a0bF9CS4yKF`*$$MT6XfO`&o4Td8Ycf
zHJD+zw?xzGoKk6L?UU}Ngx*WDlI&GmbvPxpXbcRtoT|Fd>~gvbXnB)Bs<@pW>v459
zvjsA#zI^iddD-MSHeN2-mU?F}^sQcr-6Rj`bL~Ef8ohc@0oso_<9=GJi^_1{F9nJ0
z-Q{DzY<+#=n@lvTS{Hn((#l5Gmc@Eo-4)qdOrD<j=PLBk2P~HlBO<aCaM+Rf{j*?1
zX$O)$mD+6O0ZIw7rzdLx35)0F&a4|4=2Qa5toH5d^}OmykO=WB8aL3q=o6G6o(RfG
z?ODC!4c-a!S2Xz`61qL{?XMimV4h)XxadoCF?a(G^RW-G_zS>ft!Tg`p0lH}rrH&+
zWda@)*5!aRRRy_kxHM8d)B|Ymq1TJ7i@}5UgxWjQcT!R2vE-G|svyJ-N(i%&zFovJ
z^2F{ZeG;MxAitd;@$#!WG0w;(koG>&Ya<zfZ(e~8;R#)K(d7=P0-gIg6R6UbQl5I(
z#E6Ze7Q;2?<cMQx{B$Sw#fl+A4vzC=-SO-M+vn<t(hI)_w*CUZ62#QsaQ_t@{p}G4
z5UFJgf7tfsrGGlWOp6R&70p1HVKs1HxLB0fuN74X<?Z>RUP(VdHC4r~{m86;kvBg6
zIB!Ffi9XRj%!>ZB8<->zM1NzLn0DQheQdk<j1A^1{|?;+z*mFl?H_a9(L45MuFmSO
z_Kvcx-6wCY4(RG8Ureev+q7mlFvlakJl8P^Y}jI$6YHQYJw*uJ^>V0uYwOCTE%?f@
z$gJOzV<gSNW#o&PkqU=LS-h~@7bdg*pONL3gy&?~+A(Y~+aOxNPAPIvtF?yTvT<Xr
z@l;bqsgUv@=eE1m4K}V5#Ym;f*^d!QnbBScuZ3eXkHT10*_h<FWh^j$I%6J|z1CsR
zq3AdlcBfJ`ojZ!OF?TMg=j5~6(Q*FDn^$8HPuca_r69k?V10Jr(KAw4zG<bZo<JvC
zfdj39&5|6^y$rr}o!<+nXB*@Tsc*p_Sxo{KCgQ5MAQ{duaR;3(9p(B>ggleD9C;`5
z(gk#f7woFALtoH3P|V^s;=DxKzA+ZWUgSN&IRls&y%PX>&jLQpqrGUM53+yH*%`o6
z+|7A+t_{nIsgu`g8deS#r&DabanQdkG2B8Ohx(}<0<4LRvEo{yAo>U7D)5Hl4!Zb^
z@Rm(skR}0Rf}u-O{1~Xg$9)3wzPy0~Mf-bpPj|wPjK_#5XcRoXZ~#iPql3olQ2}{7
zis>P$S%-cqTFfC#lP3A_HV*i(_bmoJ`svbrKd_N^U_R?&Dc3Chhz(hVSZVld0+eTh
z1LZEh+}siZ#LFt-$8#nYSDMq;prJCPIgHwFvN@WwZVCVgL?^%w1r<JW>NIIdsoP5L
zHvS{D31B-j1)BF?XM|o{;d%WthydK$5U!zZovYL6mG6NW*Rf%J$flcC2Vi0BueYAp
zWWPhoX2eHvL5xzWQEsUn<OUjbpvI610Uk1ZuFvGUY8dA0MkkpyjiOFmDDgeV(MM(c
z`R!2qSt}ghFrsd|^YxJN9ix`No|7RFZ|0TBo%J9r)5wydw14t=&#%z)8j#qU&y49<
zp(jOp9tYrb?$(%uYqd_s<`q%nNUikV+5gcw>!E!b^RCk0F;?t*km@8)H!qw2O`Ww1
zPx}CWVr?US93fm0TvJ@W+GCQ4Oc;)>Mu`c<8+>!@p9+2ixu{f+am2d3z+jEA!btn7
zrMB+Eck5fTY260)jM;K;vd-pvvy7YS_dU76OztN`$;XdCfr#;x$NLTZLlA3tsPwfT
zesLj;jHnO+^*2?Z%jd`l@!3Z6X|n6CG@Co;SO@TP`$=vGJLfXDgK+5PkGN?7RmTLz
za_c9=m9p3F$iMI@2#AqX7cbd?k0K_KFG61s(1)}|vP|&0D+*og58MqE$HIQfg~su#
ziQ;BW@y)vrjH}N-Ivd`rzdT)cBN6W_phBlPrn<D>7%gloIEA0h-H?G%rK2)Tu#4g>
z+S)Z?LJ1j|g!$vPps&^h4v*($uD`n34bkMV*g6ZMcc5YrO5|$K;ER|M|7^aw&DFV=
z4A(Tn7r~8Ovzt8TIQ<$)DlZY}EOUu|0&CcPNvSykeD3-&=lw4Z_`9FTD_*kaNt~*&
zUNA=Q&D{^gz3p|&<9K;MXCpy+BAO80KN;TXL(ePfkaq|aDb6Uu6E+d_rjzMAXl>o=
z6F>W;R*n24(NiX!n$kS7PGetyOw-mPiBY2_UQ8}__jANm(R;f&w@x=adHbH$a^fSE
z8r`eC<SPV{TSwk@E6^v(xJNR1nFil81FeF-g8-Mv`OMn{8J?>>iPh3w#7j2mrB+|5
zOcrfx@q}&LQ2u1t#kXh*^Mik|e-J|X=5(qO;{DAgBj<uz&YhbQ)j{}Akv(wd*Q}4k
zf@#HNYo6i0@HTKK%#@HXE1n@W@iLG+K_gGQbZrdix{g~?M2+CThn~>rMfglw<^H+l
zXH(pW+($e2?)a$wQn-&^apgXD-Kl}stfD88Q(iL1MfO9*rL3IDQQYFU2wPlmUl9}<
zhseOf7ti1fC?))dF?5PXEsy4%IJt@BX==)eE<+eh#bVNj42X_DNs+I`T%&z-xw8Z6
zZ!5nQ&4kxDKLYK-dO<O}E|>tFTzkBlUBr&$xwtCa)c=D5ZjfqR%V0@)Ci+t>XWr&O
zOeBs?x9^(?p*p0!cmq^{PSGd3&igI;9f1PBagDRhCU7G3nv5E8NDxAg`2@uh)mSy-
zPBPm7r`LmIxFoXzzQ|q<mj^ShKk*?M0otakzxjU7Lkp|+!6ktZGmDe{I-q-1XL>xJ
zNW2-8QV*s_!=IU^^R3wnJ3|<ktn|`$8KcH{o?k%;I9h-JQuA6d`tqg588sT`M&hNf
z(ZdlFA>S8W^ePKNVp_z%$lVL|Px?l7$mxfX01j79r)VAO-BM^#=0%A($BDsJoO=+0
zVa}F6`+z{0-`BkgaA%yi4Ym6gncjs(|8B{Di8jE57j|d2B|C8*sFfYOk0)JZawdLB
zD`~CurcRDSO_3QZvV<Ws>Yx2!3K9o!`ViyKX@jZW9e1$rfDT~$Qa9Gmm~OXW0yqB)
zpFV>Tf{z}116>MhLXw*krH6cRVHe#>Y*Fl*#6Biw%E?`i_V5V#LBv2~i%iSsH_N}Q
z;V{y&E*xV$qBgaot*HxZ2OkW*FyA{U593=mjfx)DSi~0ioLT|ukq{=`gl^0M7}{;U
zk2F7#JhH($=z25Q$ejFy=Ck;olxx~1^wgISYW*xLFIqe!YcDOUnGd0QfJ#7~_q{+r
zNO#-D*P2a3{9Q$Z;qo2zkxRT0WyiaD&0vn!ADE0)->Y)|P#Bi$jPdD*zjKg3F>keR
z1;sdK(2n~Shkwi@1*6sA{4YPEUU5l8FdiY{w!9FC)cA1FQ_i_V5FJgTP+aCO;K6N?
z<sh1V?-i`qe%g1s2a^xK&AIUY;jG1K(b{>Y%)5FIgdhIGldZZJRkUoD9XH7e;$8MG
z<Ok@bThGiX_~22u#HORsA22XYP71p_vP?K)`G%BZaQ6-FIM%S>;=gL1_MY4Ee9SQM
zOrB&rl9(oj?&uQhP1pORB$I9V=fog(fF*+0LSsj^yZ6f@-<exRk1&1PZfPem>(^O&
z`h5w#Phzf*4ij50pZJ`;UMEVJOm`l-#iL82F83Tfco`WHJ?N|2`VaRL^N;8kAHMr?
z#chfT8{|BX_A?8r@P&AQNAqH{R0M_za|hxNT7ZX5Z=jCu9K#63Ic$ffY2Gt{QU<1g
zCgRiJs`r{xF=B31_p1e+gZMM7-XQ>+3%MEx8*~!!%pVvO!_SD*o}VlE&c<Df%irlb
z;l~!U3RkUjSiVIlW=hN&q8bvNp_T`ppQW?J0JE}lIL__tr?ZU&&Z&HUPE>~TpLH-U
z#DosCX8P7-Ec>-xXH$s?x_<rCDJo@S(>Pv`o#0R#2mQs?g4Oc%o4QOJO2pAp3C9L$
z)NFE`u|p^T+#9RgahzwONt!K`4$>CFvREN+yNJ1%B;?Q*8tvN$VOrD)vcU#urtZgW
zL+O(9-)|C4@ZgONWvCD<6A_3sbEL=Hw<=XW!7U&P3HsKt^7iXMhs^P;R{6xFpskCM
zQ<qz4HHPpx3EZosg2`&4v2?584ZD|XUFh~&Rqm|%O|cnVimzU&OZWHG(L{~dyItMC
zzK%KPRvB|gzOMC7rVP;CVMKD_Jf}I3YrVoME26=9981U;L(gd7pLluoy7~$m1%Hl{
zHWsgVATVCY!BmI1i9UJOD#XKBd;?2d?a~a_zm&x*CRs@vjpb+^9<=DUO%rxAoYq0R
z#yc5pEhUlHK3AQSxvt8lx7_89Tbrb!qI@O~?q_HT{sk*h7641IsZqwgQZUm}7of>x
z&_%^+<VJ?;%}c!h_%0itOOi`5;W$c{W|>@+F&r^FCr+|UDMyDSk0pcsV!SUTR25`&
zr+E@Y-fY&2AVqm_aeBXLO3?XuZB126$%h-0S>*y=nxX6jAXt)|nYn!LD4{3}OFe)`
zp~}0ff_Aa%GBpRsM&9fmx1kuTgCymf(IbeQWv6G$<l8iGxQ~K{O6SHHfm~pXH<*vx
zjLER7=@yHjz_js=@u`TGx$@EB!49X?$Ja`qSm;sfoPc9t1-^CIWE&eN+*(O;$rBr`
z2AqKhsFYl(eyt;VdqqNc?kM*Jvjqp+4#F!=`>}T<0d>r6pim(8GiLrF#8aAK%Bx%^
zk-}OoAOp6<Lp9^gXC=(g{4-M|rT4U!jN-3Zi^u{1l}l{O;-A-k6qyt|AvQ{wyDPB4
zegt?gmEhNy0d8ClL*(CL_pcXPg0|WLK8PFaY+`dSmR}%;1B)Ex9ugZR;y78A%E+Tv
znIQf+fXYl^yWhNh$bkZt+UCTc_WLcHyNwp>ue~z#F!@ggV>^@~e|6mNBLMt2Mc=_{
zwKZ}11wZA(?Wno^H8CSeIVkjCWz_`fC1;9Y;(YaZ`Uha%eshd*-r+Ia=7ix^x))lT
zkLv9cdbeNrMWgxmwe&BwLNTE0^m#?xt(SBeH=}Yw3wlqbEy%Mn?vNhcq_dy*9u0|R
zN`0OMkDVt6h_>(?{t<c;R-YgGb(|wxx`gbgKPRD-I;eBkI8K&R7sOLC(05YYL3?|`
zNi{hMCM0XS9I-fs<b?o=O7lGqU^f-}ic8nsz1~wZWnS%u|H=-qARO`O;5$fmVs%-V
zOxciMZ2KxC7hLe;Jn4ndxe%dKW537BR6l@5hncf?>=efx?8>?r1+tH)N_>wy90TKt
zEnUit%sfbcy`LrLSvTw7JSeoZvD9%zO0Yeb>J31T93;9OkWGCd(AlPEBuQT&9p}WD
z(Ht1-a-9taOvOU&KkaU^tzi{0k9LvPIpa1zw(;2{>9jR&bc`jWTj7@PwO)o3v@vh0
z?i)PiOJu=6)&KsHYU|zgyF2*7V~#KhB-}_qGUx?Yb;#zDw?(<t>Dtp<WP_jSHj0YY
ze^V|0J%3HHhZyOJWiO;}KTDBM7BZn)0I1i&wA7o>@@x!j1u7hrYgA3XEDNP?47Jnz
zg~RFP%YF>i$)Z1`V~paCGAM=bYU&d4q2BPi!SNzpuMM1pMSUdIYmUYGj{qi&ced&U
zW}VG`6qIETr!`wa@f+0ep6Z^IATOh^+YQWe!S%8689Fbj9^KKHzW02^cMRY!B?M-t
zz875Wyf_Ba%VWMu>8`c{>rsxNSheb0kGZ*1z~VhNPeUcvexGn_yX`YoMPcYAQ7Qb4
zkFcF`b$EzDnC%J9ksRNgk3AXU>1!|EmG9;Ci|Tdcf^QL@LKiCBO^Q$1^QM$kw1yzQ
ztowXLLB)>)&vG^cfMIHs1}tz!1|y1Fi0HIeAznF22OLg2=ub?(Hwv(s?FG6C$2z1C
z?L&<42!p^^BtnwEsXhNWs>YE5DB{9rV~uc1kq1xc_(fMMR?`9}g)foVe$gb&#kenH
z7|F5Oe89wuI4$5YDJoU7LbE76IUL;u{RXT6io|nAFO^+JVV`#cQ-vRwuJDRRoRL#*
z!`k+<+w$Egjp8&2P22_Sr)$7~6<J7%i@+3AB<39)@+km~B0CEwVfAI@a4c0jI!g(c
z-UuL@Z_!JoQuL|dc&}n2x~b{niK22ZVwEnzhUv(5YYrQ|_8rbMa`Wbah?tT)2H8n(
zzW`6R0oO?h7%A@f>Ik|Ff`Q8>x)$fVavMNB-+JDamxg_bo%NmDL3DGTQ&^U`jN-(y
zRW9NsdvKF8i`C*j1HroANsyLYn_@5M%uy^Fc@hGlm2L>FvY<~&s3jLN7M%hFT{V^o
z*E5+?I9^_#)2f)i8zfOhKE)@GbK-=>%kwYYbNVSwf3i*(>P*L#E{n^w){cH&63~kI
z!;y0V<PQ^8L2N-k)HxCnYtTxS*Yb7Na-G}y6B1vX9UUQ`#x~?a3QIaRw?G$8?UVf>
z<Eu{PyMN;$doloulw^Q$Z3m7>2|r--)m&@dc_aS9w_vX!w44rm7(Erz*Mv&@r8J9G
z|48d;5t^gg25z(m)C}6z=o$VWbj2vA-h|7Z)9ro0+GacC6qAk>e;*r#TY;0aW1NwI
zt_Uv*`)VCg(%r3TrA*O0v#AMPpn8PZysU47E%@XDHTvzrq$@Yr(ah0ou&DUWW5+$f
zmyf6j2{g<(D75E5T=AhZ&9A-{lAk-18FDSU;@|aFIjE>(Rs%2)?0dG~aucw`Bn!kR
z+nI0-nHLP6o41Q%u9rsAM_*{R`zIYf-EY<S)OUXnQ+^8E$Mup?4r^>YoL>#DH?blE
z3&C#HQwxTUr4JNhrr0Zhpumk;2Tgv9YVt8UMuQAlORr`VsX5@s3u!YWC2;MKXiH1z
zg~HB^-tWi{mc>8Qw=uItfL@a%Pk21APUeal%lmJr+%D}40vM6<l1o7dKA^>4!}Xhs
z^QUV0CI0T29>R0}PU|9>f&hT)&%FhjRb{b)vvNWmeDgW6?z2e(r-m^-k|qr8Y=kLQ
z1&`0tOCz8A^Y^JcDe-&-oz0kB$_}%IIR+ohF$WkKSgbr+b{9W?>V1=+PNqqWGY6L%
zHQ%%e4bZGxxY$jmGvo4DFESO^@2Rkk+XL-`ws{Fg*!)l0u{L?kJhnV8SPt$iOO$2V
zLM#r(Tm(i~g_GKGfO)&I4$L!_%jhjtd~Hn%<5KiJBv0S4FJ*&Zl3Dc39cd3j9&vt$
zo}W|=+vM>$pIpj=lNb~6zC6r)2C5y%5rs8;izjnRC^*xe1a(GL{rM6zABua%!LK2v
z6<4l+07NY3BylgNp`_I_u?T8Q9k(&iH_XYGf|2a-2%}j7ee@F=IcngvIK){hd@?Jc
z`REYb5^t?W!2UTw)h%_0O$@Rh4&3Uj3K_|kO$z7H#UEFTzWf=(PU$$yqK&T42jgno
z>u5z3mVnl9+1CH(pXioIE|}^Me|Ux=PRWG=V<vv|A^Q%81X6Sow10TcPYilWZ8jR4
z?{REEe0Ss-4lB`JR;D|sKDQ``*C6vq3AuSz&dJ2`Y^;2DAiJ!MHsy1DY7(KD`CstL
zKhPG>Cdi#zHK{W<7MF5(8DxSFl(yNKB)c6-=RaP057{Ei=ZTRa51#<78@Fbu56J^;
zh>_6WUIFnU&V#10oQ#g2h*cp?Yk;sw6Ct^U_o(gz^`9anP&wUzpxoCCkqUMX`+%9D
za8ME=4b<)KZ^L4i8&K!WfsDv~^s(4X?5w!5I$F$0EkMzxC+wPhFoYyWZjzeMzCGI7
zG&<{XICe1^lKYzx1B4j8LF2}F$tm}o_TPsO_I(EwaNM#=FK-OBJft0Z5S+#y7n7mD
z{i@^s%cHn3v#>C>k6S#nGvdK%@0NnoA|7S@iK6`{BD?q$%vQ<!?EU}eZxUI609%`q
z@aO-tn>qEzTL2+-78#)W6EFCW9W2-zKnHF8a*zIJKXZQ(a&WIKqsaRG=ZE~`HTPYB
z0+LDz<V5^02Xh8Jf7~r@vj5YS`{$4KL1yn34>mJi{&~m#1^m84|G)4AsW00E-8Ioe
zYXb}WCD24=n*o%?s1ouyuO%<v0sC?lBwyL4W@*1S0C`bAb~!YE42US@gD+*4+`d4D
z&k#sM9BKt1Pe|Ui4$RdjfM@p<;DyF;<T@(Sv+K8EIRSeM<(U|RO^6r+W(J*MX?)d>
z9zFUk1cTh6=M!G*=IOTu*+ABr>mWFWl+XILLhn!v=M|U*O#z*_+b5}z2W#WyR>_6O
z!k5SCq`c-aGs32W3;-vq4l)VfItEy8+=hc2Aiq?CAv6c>ha^FlDuBNUDvXQLb{+^_
z0`Wu@NSSU)z67}H!O^m4q{PXeHv}653#+)a4?BWDr7h5;$ZxxVUEK7(NHPGG_BnFZ
zOP^C{e0vYc>j3!E#iCk}B9|@l0%L^dl_A98|BV#466$OK|KJOFM~R^LHC+x24FR9S
zZ&qR(P)j!jfU1SN{EHvclH390oFrwzhW|m`AGjM&*Ep}aizas%0>z<hb}o>hMU@P@
zC9DHc%E8D1fS*;&PTOKYILjK(_pYSVGR^G;z;cxy$u6sVIpffO^6#qocMJguA$XF4
zzt(<GA17c@=OgFTdAmmIG^ikD03&P~0!wC;jfehdf4zqXghA)3D}4R(WfHJ*CV`Ri
zBv?T=!GQS{q^IJm`SM|tMKzaF_Vp5&7TbY%<PP<4-`A%A$*%^=^iIqaHB6OviHK1^
z>eIFaBqlZy%g(esw4E1qU_PSN6(xJsJN*~f{QdTvZ2SS*_}>-*TJe-MVT>r8jhAe<
z#k!EUPl0*+6iC>Z0*;U*8gE+UYbr+NtJ;haxtxQ7rhsa&ueSn#i4`cDd56fL;tJGj
z#Ij!wn9)Fb;S!xdH=j)D*G+ZMWkOx);A2m~Cjdsh2X5Ml@_+I#o=1Wu<)uaJ%}0b{
z-k`veqmW_^zjg}Dw5yAXILm=*%;>%daa|scv(LZ00G`(f3L$i$<g4FeML=E4QxK?O
zMT!_)0Wr=ZAWg0htDEJj3$h_qB??**tD#-hZ5?>#4Faa=!MwIWjrQ3EP$Tq#)M{Zt
z*4TjFl7T~P0oDyPC8R3a+FR>>*^CBl&Gx{e!p`HQONTQmz9hq&3iJSv{VHO9xO^4r
zjJ&3+Elsox*8(?|%yuxGB169KF{fIXyWs^IwocI8TFC^k!`X_`w3Ytdbwd$S<x-kK
zYQ6*uC+9#N*g+MhpRvMs`he{&XsZOJTyQy49oC0`(*;&Aj$|g<ILztKTL6z>SwD<Z
zF-gAcBOaC&m@?*T#q(NBl9+v)i_&sb+b%tafb5<`foj?))u!lNp{rE3=5nE4s|O}b
zd8r^gp%M@f_%8J5NF3j#4%T!x7}JcCFR0H~am&2jR82(cWwA81Rc6U2lTSqZ>%*}W
z^K@Vtd$nvny(<1oaM<_wbxg;}ktU8$m}-vILi{?%Z!xe3LT%}C|5n6>fo(f3LXk@v
zQt&#XMCk4S$6xzk^}W}<4@D7W2Kc1Oh<S|m4BCFaQ1on%i*=o;!gg{ESw24CpIwXN
z0A+Q9M{Pu{rXHj*RR1nT`A0=a^z5m>Q;V~wKBw9Qve=`sRzbwV{mv73OvxVga3b&%
z;qev3e|fFnJ0P7%*~-hCObXTsb@6dv5`dU?Udg<nEh8fY=Pdyw#fhwaC$|VDw<1)w
zhD9KP-$}FOTL&|tvjA(HnDnTUWGYuk4!G7#0gOJ{jR2@`C+3TV#LuNX2eiZK5Fw&c
zV8b<zOBhvVV9UG#ISRwSNVss8kf)M}`JC5Zoe3X73HG2g>nRVH#gnW8S7#dV?*W7N
zoYO0T7<B*u+^Wqk*s1$~xVgi2lbL14$2O(cCif0VbpUz@-MOiQB3IDbIHQD7jR0Qo
zi`s+%J}}=vUXqjq*lsLB5}~7;KE@j$D|(lWOI6amxNZkP-&yp!olE1y9C|)3K$|=X
z)N5;wz+|ULb^<Kq+Y<r0t8*Po&qR__bHB|j+37q>{`?NJ&3m*5{iCkfYKE0fsKHCN
zV$n#!3AYC}r0HY}Tg}P@zZuaW0R-FA19lWpy|9LT(KRAW=RRwA@t~$4Nq9m@N%1Bd
zk+<s9+6*_(f@Aazi4q-H<ATT|laEOjFp7-Bs7lW0w;XGaZ<r}vN*ELfj=rvseYVLm
zsyv;R$+Yl_poSUy=jpb+JEk2A6SiZi{Zg*i@yOG-FJEq}@Eb)jSxvF>3GuN$(`maT
z$kkGlS>|Q1i^N>6Dstx1N<L7rQ@JBJ%u)_Nlvedff(<1)oaAmzg$g;X=M{db-Uy=6
z8e!#06e5Vy)cHrfa7P4`O4mi$mR?cwBA*87HQk;k^4$btbtFEn1>n}q{TPbE4f`!J
z)G7y^j*Glp%@WSXosBoPR5i(W`DWjlwvRDy!#3tm{K_W)(fTDzaPCeYq*r?jm=ylR
zUdSG%2#)S6kny`SCgBHS8$7P;vJ)NSX8Z^@AQ;!<W8*T!UJ7M-3ux4pcX&rr)NMXf
z*hrlqeMbXc90DfDS<Kg?NxM+6#V<`-R&OqcKIM~q*j$Y5>(A&zCgN)7XrpINt!NGw
zt2Mhpsmh=9OgpIBAJIPz+5qHvA84%Wz?!N3Ox(N@STfX>e@lE4?cN4o@Hc>n3}-;l
zMus{D_&=)Y9vah?t~9r)qU@$-S^T8(F|1;DpHeQ7G%vgbSxYuxpyTLogZ>LW5zxZw
z2lV(ZZYV^Ut@L_a7Eb_ErW&U?KxdwVoL#<6H^lbLD3*cuc^_z93#een>+g?i3!aup
zrnDdmn=d@2Kkgi7<m(R0RcZ^wTa?8Tid2$=xZ4qyq|Zc~@1;yeX%bv03F<8ED%p;`
zWmc1`qH4>|Dlk$M6nIt8YRq5X6Sd=Nap(Fa@6sWj!i!4BK4+4QmuV^>dKQoA$w6_x
zn&-Apb0tvrO;)P>_>0@rIS{Tb4N{<H`&4=wHHqS3O8k(pH8QOZNd9(1`b+GAKuC`5
zWHsv3BGXic&WHZ~QQCuW>130B?(d)V8eLTgG8oPJA06L4j8?g!XNWoDQOiKOcz?tC
zzS5rC))LEgTy7k}uJID5J)`qwb;ped2~2zj)|v(%b@-E`&{Hk6BtC9;S-G#?6+W<a
z>K!stWU{k2=ri$f+V<GZoH;-Lx?QP5clB$F?9(Y3p6pkU-TMW9RXct9W5l(3(jV4$
z=Z+G;Ipe)v)GfnJ8~Y4G4<)l>x5OeEU*Q7e;M#|BTI_wm77t8B$3Q{%#Bb)hI|p*o
z=*2N=N^xA$P#LB((bLhNV{y^t^W{@qzUB;n>}(H-VUw}+ds=r)-vXM!S2C~Tww=KM
zeEO{QHZ)!Z&+7mnew#f{?7tJp4zf{4VlkLAR_5^Y)qall$+a-+dp%AE5#ISAMO~k+
zEbLYjSK76poncO@8*`aLof5=hL`k3gLt=xT#IlRK%q!gTfO$gUH$p+DoC(DF(myG^
z^ePXzPWyo~_QLz;Xni~*ngjXS@``>pTI5@lcH0>=EPRihfg4ixMEw4KF+>WndfvoJ
z`Bop^^xk^-s|wq;ps}0Zdc-<@Xp=AZ>Gt`_o|fk7JH2R3Bv*IR9#YiHz2)G?L9>H2
zT<7!?M#`{qM(@h_8YpP1K~Lz4z;X{7V}I+TC{CZ};vO+~AI^XsL}98qQgggIRb4*Q
zCR{m5`196KQG>IybXW%hr9rSJPdU1XzA`j^$7JRtDw;Pfl-J1n>2YH!fANc{aNOA&
zA#k)NH0`Sm!g|vtz8MI_UcNvEGHI~0(Z0BO6z5+?+(Uh3(89^~uFa@%CZa)@dz9Ee
zF{9LAc%ZBC<x&%&SccTYJ3f71KHSiTE%EI><z9N_J7P@f6Si_I*?lSN^3ZR6kmIs^
zb3e9tVBX03U3sX`Rhw)a=MX?qPQTxt$ltyPV3I-nU8<BBo|6}5Pui{XdqzO-C6k9i
z;ip1_x(X;Csun~rz3b$G2m`~nluM};NEEB^f@714ou8uDBsPMI!AiwPrq3aFzGw*b
z7s87%RQ>)qCoO{Fujj)%hzvuB0#<a7H&Mw0{4=lw?glbMVT(x<p9V4b5}HJyoXIv6
z=?7sH%KzxLyVa{U9fbGwCstj=olu41(?`*X`DIoxmSW=h3gR?euRHSvcUnu`To+IR
zt=}8{M!5BLhD(#*@8z#^Hc^oK8`@k9AC5h44ghtdipIf=dAhs@P*V!s&D(4#CZc&L
zQL@e61+y_tPbJzIJw2uiF%%>2M$f_J^>Gq?z7ytS)+b8fGvE`wSLx_qYZ~4VO!}CW
zY}ZQ{No3I#NVyeF_imo0(;--dv@ZLj(3+cPiF8m*BEcRvtDu!ECe^zzCm{p*w-n`P
z+FA9FZ5#%8|3O`v$62j<Q*Ga-ZH<N2@qvG!RyAfE7xwYyL?;gJ03&9}M~QPy`<pF2
z4X)$wyZXbJyW7c=aK&*&y`Ll4UWGizsL5H7W!K`w;g~4Lgh_F!`Z-zv_fmV?`pdf2
z6GNmN8RzXR;-Y@Te;kb(0=(Wd1ar9fMocsdHun+#44lL5^foJ-+lCcL-oR*)qq}qR
zB&Md-_g!2A;gWYj%#R(^5*SYZF0OC~MK3?y*Tx+zQD(}Ulm|3<)EvA&#QN#gkFk@|
ziJB+Usc%m3`%>-s$0EWcS@PVwn2Uqt3EZ!LXeTKiv=f3EAV-^MJ_f{owo|nKl5Z=q
z9&VC|_&5ebZ8p!{K(9g={RYb2o}tBXp2%yRsZH|lPPHBAP2))KeYeMIB0>G^{M9BA
z2a|w{ipBNrDW!JE9maRQU9V=kH2vo4-mJ7Y6Ve1pSr}fhjhMyh?f~7=k=Pva9&l@}
z6hdxDUNg{Ru6fH*ZG~*2F_qrN>|fx_eU-H8#+MMap7SC~ek^>t9g_zwiBMfy8-DMD
zbk3r;Q}LY(BSi_`{PUNL#hv{-XvZu~;+{1O9lVS>&C!aqpRG{axBAE2&M)kbI$Clr
zoq9JqW-8Re>l`EZR<dABxVW{)0q1Ij7~D7t-d8`B0o0Y96^cUg0H1lGvSXVQpWi!p
z?HdreBSq;+AFXunaEM4JbB5598}A?<n1B7EIZM!;v^8F0IhV(8^=uZ=XP{ekoNwlP
z$I*lINz;!6p7v6!i8d{+4b^hj?`l++pO>qZj^dUl4o`*?K1nT$x7u-1sO>#`oxZ)B
zQ;P<q6FlDfMA}`=f@#?pDOesWRAINS%zcdkt>(O)BO_kz=LwYs0X@k$V2Z4E8v3J0
zc<Q!cUDN_7QSbeCK+nVG2ndL2^7W%idt`GTH^qED&YA?Dr+GSh?t)o^%p5Os=aNsM
zkDQ(0;)UI%uiO;F_4y4@bFT=Wr|2Iu&$j^}EV`$vt}{|dmj%-w=I<)q-6G&4?8{|T
zOyPb^Ra?z^!JH_KLh!BDuH#W2g|(|<_d+X5Dhkd})7-tL6;W;rH~~RgMM$yp5r_FC
zvulC{%^S%CV*ETZ@f7nW`MQz2&9fLzp0^mo<u8D{%B2K0`To>(Ph8}+P=jCJd3rDj
zV2JcWPeu<-5C3YmmLVQXQnw=B-jRxh-LZe%0J&D9PC`oyL|gHt7n>cAw_ggeo-ce{
zTK=9AF_Ep<)fV^~?v|yf@p-PZRMn4QYiq#KJ{g$WpKbXiPkO~N-K3t&vfN6jY@8XN
z-F8T!*lo(OAcfIhN4YnLrx3w&6}gKg5wc%Ra63FIXj^eoK3raU`_<V#bHBhV4FW?v
z*BiBw*9qN=A4k@YyjHuT7Z4SB3lqy3yRJsAtKew1F6K!T?A6%?9$2pFeA~*+1|G%q
zeU?ho-I%bf^K^0958m3#ueTGd&R>nff{7Z$0<*L5*x<t!{J!lt1nYPlDHt47B29U7
zh^yQNWFI?&=$1Or!#Ves<a+zLxKF;}6o<vfCUV1JqS`GpG`>_;rZMOY;Rbll%z{Z;
z-9bIpuYF~&)^)BPjdl;92S!|g(aRx{2U!H$X}G>QcsYUgrJ&rf6s_??>$@PAQLLGY
z^md7n*d5>FP{b#W986J6Jt|Qu3N#JF*|uNBn*Fu}rewWp;+f`#9471Iy+w&MmdU`H
zcJpg)ZyQ;7359Ak^}GS+v@b6mW}q5+f{)PmH;@rgHO}2&iL8>jO@e!gWdZnd%_b}}
zn)xQ$ST-IDLgIqV8NL_^h?u>f{$`ObbaTuI4jh%Fj0-BwtyA@$;Uh+|-wScJ#cRSX
zt^*~jM5*61gA_W?U6E%SQx+yyg|+9BRGRp07bKy3%cd=7lW(R5KYv{A6ghx6@t!ZC
zz|IJbY?LnFTm7bAbC+Wh<8^mhOj?TWco+#IZ#%C_tv<X-(QMc3Ys6q^b<uPd#*7Ut
zkmjb|XD<UAs7?BNqTG;geOR-ZY;*T)i=8V(qP!L~Yv^_JR(?U#ZJ|)>njYh<7WCGg
zYhS-or2(20wyzXv8iV$imX`)ilr`7at2G^RRgMvR_gZB_1B${_Q_o@phGIpz8fvMq
z<n=p!K}PGKO}LxEdf**}<?$YAfxUMvI0Q|4;?4X%p^rmn;7Lgct+rST`Ut;L3(Y5G
zLrde0q)F=^qwmqZQ#Zb|MLLN&W<;GTs~(lZe5W*V*G3*-CxQJ(c9o&{yG>iTcUq-Z
z{?OhSpQACaN-yLGn7*(^MY!nsM`68R31-EoG~hDiZ;M%_dE{O+rMGK<7xygu2fSP4
zSwrzXUH`G(y&r5_nuwa+hT!PA`zo4qMs{WP2m?KMo?cQ?S*X=~F2~^gz_0Z)pJtOi
zF87k%siT2jAY0wk$7DVn?|S$yM?9ah&e9j<7}Q*QV4<_Z?YN=3O-uNLz)FZNy=gsK
z+MvMoEEYpS0cE=8e$yPuw+9H-xX%_7zt%~y`PQX;BK%2~GTC@}n<ofefbWW4KFuLj
z8`ZvkTSLCjrLJBoUw;qXT+t5M13img2mG=sO@ZY?I6p$V<)GtMF(EyAN(t@}>sDZ<
z@$gmbDyymg;QKPotp&J!Il#6Es?a<Cm`SC?0yW}zHwi5I*ErTPl(R>%+_57YC=<;{
z5m@RMbN@aaQJWi>2b_xGvaXw=dm_T-GCN-!ixP_|g87quGTF2Pz{512n8E6H9zhXr
zQ{X=xSwv<XU_I<7Ss*gvu7wHl`Q4zqI;M-chGEqUga}FkVQSgLrsDh-yOF1lR7$<>
z8Q<W*^29_Q1ChixDu3J$p1hJ9JWrWs+zp#U)4j4}7k2?8?R2zUna9ULl=a@r*X1?g
z?i4mHxI&C+6|x!4H+>a#020cB7>Z(Zd)pnuu?1!ipFiD84RNFUrr~AY-Vmg}nlBwJ
z?ZvM5=B3}~=sDnKH3dxz=<&sv^~=uq(lr4x>r|?R__omaW9JBnefVp8$CJt)dK80`
z3|DT$vG~<1*jOXnK`QT{onc*zhv^I^k&e7Mp^4Rsb-7pg1Kj5RAnfd@zB&GOUBaf2
zT>bY!6-;d$1BQR&Ay9ALmw{o7?iy}6U8NLq3!ztv+k6*OzABv2h)y2-_?cVx3)tb#
zpg4D{KDInVi&)FGnp2#b?mdQA?qg+K4~{DqLiWg0fJ(gjD38;s$-JF*%#RZ4;H&%7
zrK*7$PapOLy)UEo8?1WzLpm`XT<e%bvQSZ&a1N<n6BwPwzv^x>j)j7wj(}0|TaLTi
z^eu`l!t?HNM>I%j*})u84Rxt~c9_q*OU&8g_kswgJ1*FWY&sO(7TpZ4XA6Y2Sw?4F
z<;k&5(?ICECZW`l4^T$uI)9}WRWcZ&!0MH_vf)8i^3S%G*B0mtBpE^INTUQB-}Uj*
zuh=dP4-Ndi{WsZ*RHI5GKSkq1_F}>WA-d@^)Q^6rolsLuL~jzo_Wf~#R=5qM^?c$e
zCsD{#e7I(t-#jV7c1AY73U?0{iDJ|_6;2~NV#rg&H+63GDlqlpHxR{HdVn4YWC|68
z3>X7Xv3|Pdn`~JiS{JB7-`{vKTxn?$mj9CMQjR}XG1vt&sTjVBCbMN=-fYIhToYdR
zF=|t&X9lZgXKkgwPW?M41KS^xs4RIA@)X(n;B`eY(wp)y8Pc*i7BSy&&QCann8aHl
z>;bnZv|n^rxdV%WGvd*X>;36@Wdm{o{4MwBelfJYK%n8b{FrO@%NA4*koFrPw&BOc
zAj%3cVoHjGq-=!33wk`uFIDK=<{Fbv=hQynAb=wfpYM3s5&ufUf<%IK(t$$W&dKe{
zAnQBrka-jKV<{qU#wPMKe&%m%w$rX>#N?g`UYgH+j=UM<o({mSUwc(NK?=QTFIKe-
zagYZDLhrv>_Bc~Lry8chdL;x?g>um>fn8b2cO&gWexqrTTa<_eEi>(I^l<dpmTmF#
z#VScn9;^NeJ-l$Dxky4>^p+B08<$F<Zf`-?2Zi*NS;(Z;R}q)*!IE95yR+ymWjxxX
zJ)6n=ZntR(B(Lp~f5CDF#$|pN7u6u(Y%490^|ne5Lsx4}C}Czp=YNDNL%2dpKQ03%
zIrrCF;%4?rFFa*FU2BHpa>le0=jWMo2}bLkghye>7lTgzRJVXP_W4i!@T_8@2snxN
zi<mc(i>C}KKX+H+?lvAlv(uz#^fh$lSmqgb-`63tlGFVKce@Qcajf@By6v}Kj)kZ&
zkiP|Sv@(mCT920Qf&heYESc<3uAC8Itr%XKjKAxJa}Urez8-zZq@Xmmd(b{&SRzPX
z-10M6UBAvT&7s36F(A9*CZeMu5LIn~mfdc-+aCRX6Wflc6eJd&-KMFZIzIDXsez-J
z(xj-(n$cuj*r3_x2c)N2i~;s%cyP!eEo}<bj<htRmK++wTJpGrs-2%~(jRZn`kkL1
zM=+6h<=xHvP}}=od(dCS`og4oW37b6_Pe=@`H~_KXi*;$L2j`7Gd{7sbnVfaHw7Qj
zm!;JnE%ku-kC?rVk=0uWce+yZw71~nh7AjAJn`}Pi#7Kz#E_c@6~OR=>WT+IoNN3N
z4H3{_y$J@odE<-Xh#XGlr1Dg((kSbpj5w{xzZ<eLPy)gT0ZvvC9#n~l;uGP(Z?^(@
z@h@nu?5;}vNhV7!@t3@W>&y9>&OdOYfBf}71+Opg(4IJMtkNGt!Nti~fw4USOYK;P
zSVP;ujWYD92HyMJY|(E(5$Yb6u3h_fXn@$^w_^OiV0FN#5DBN`Sss2h_HRZUB?Ll$
z@Jy2}#p`Q%SW5lU3mrukj)#&qBn~c!w2yRvDEL1{W1rqM+P&-CR`yR`{E6KB`_rck
z0+BAu0l~{Zk+6S%iwN2-sLKA|`}_t7|MRr`_4@zc`AH!_1>Ds<+V_ruA{Y3dzkRa@
zx!7N?@YjR>4~RMrRUc<jo)WG<?G8i%1<UV`?P`d5X4}O=7;iaRpt}W}PxsFtg6T-k
zt89|BF94AQ=*BkAmfVxIhvgu>8EmlvJ^9!`9mrBT_>w0ir_1+GJ63(=h_f&mRvXB-
zwTrF3MgW+0Xibz^OabHZi!!GCnx8;{%D;egnBoAbbjhvXKhbEN9d2Ii0_ToNz)zb!
zJ=|OeN}3p<R<9eF57r>z+z}+?1f{qDA|I^OYupZ6;G2D+d|`YeJK!!~1#Wv<Ii1fw
z_wV1|0#Z<<h4N?E!fwYPfsIN0dYM)I#qiMxa-qE%V8rA7@+z7(5+mtS*yX@nv<ir6
z^D_DX`h>j&;6|s(aK}j-sd50E8nI9u@Pi@z6jTran#lrW@XzBWwp9aErjrmRx2-!#
z3c`zrKDHuPSeM=E8kw?D|A4e3xIr^VI|a*p#phEG3OwgR3!numL}j3cT<;@T_aS`<
zWZ|=4#R9l|kGH@m&YDHN#2Uy*j^gW&=P@6f4KuUFs8)%(_5RcU)kJ}VRi{OsyoX*k
zZt(_b6s|Bw%Yzp6mQXzsgBG0qZfI@kssoN8qjXnp01l{tV0SheA{tBk&GAz8xT?Gt
zcbQ&=)fm^Gmf0+{Rf4NDX#v#bTOeba@z)wxsf2?3xId={DB>nTvACMA|KSJbE$w&K
zwR_G$seR!P`dRz7-UEZhxHT{~bpc(`Z_FBsf>u^Hp>=E>hnSF%aNUEg68PEZ4@5Mz
zF}}*G0A-j!jddYl@we*G+|er2n_eNUvV2ni-<%Z@6KMQ3zWM59=EQ67JClHgxFvMD
zO0&&!pXIk{-{U#t0RYIIw_ErCNf}=Z!7kr^d2x0&S5&rzKR6()i^P9QVAW3>tyz@R
zlXSrilGq`IcC-eNRez;*1=9A5Q;zlr3sAJKin6k@g-5CiYxM@_NTy8mx3PI1jhVxL
zLhE0&TN2?G3NsF@t$Y19f!;NXZ5(iA7Z1VV_j!Dz4uIM0tR?W>X6yWsEIi|)68&5d
zA^;x&w4JZ!Degdi_0|&rCUZT(qTYg3@_}vu;2eA_$o=Lb$?Z@L#MNCbWex&KtvcJt
zGYbdH?`N(S1_SPYR#&nBu^O<P@fz+y>jS!3cHS8vQtPrZ`pld3x{O>l=oVucWIpFa
z25^|QrBh8{UWBfxqT^Po0o1?i#pr6ir66$fiI9RmI&0a<ay+5d`63KFibY3329=fS
zS!Kp~u%?4HCcFvY%1f5!Itql<3y#yyM&-bR^7tbm1RBw3qUw@9Z`|3Zo4nCH=3EK6
zB90K<4yC7ML&34CT>dQ2I003?fqoTT7g5o2h>Ey&H?hdt)UCZcN;-)u|Gjbii&)HX
z67~)JE_Fc&QQ73ntBu54@9rs;`EFy4z2pB@b$i(-FU?F@{1eX#4<?i0AWuq+Th;Od
zk=7uFWgi;@3AkAge;_7Z%d;@{y<gRW**Q5mTX}hFNo%nup}bmJ4ht&frd3x@jwWZ;
zTaUyk8?N6g)CV`Z9k7G?7K(BVy8M%4kGY=MTdRP`5IrRb^q1R(K1ar2JtBL>my`aH
zrRY1gB`}I&j9@-QsGhep?*c$SZzQ_{4#$*w_WIFE6DHfnLUfPd(f`NXTSiqG^<BS!
zAX`E@q#KcP(@3`<jdVyzNsH3mjZ)I3bf>h^f^>I>G)N1`S(kS__v3xW8RyeG-ZA`u
zV>sNKy|3$9YyIb(zd7e1dn4Smco?)UC9jY>$I+S4B)bbxMsMYt%^tyBEjUeQ-gra}
zGYD$5x7nEND2&w7SrQYSq&bL!?oT6>MejDswMsghJeCy&+yq4bJDVtX%c;oPAy#vP
z(;f+_T0&x${4`cH<wz{nK!`D9c8SIWRrgBl6H<dHr!&Go?W16yMdze?8Lbh^JVu-E
z!9|PZOG%BFAyjaSQY?eQ&m%K{5AZ6qR9$UY{ncJn2cn^&+qG87JHj815CW%KJBrt-
z(K0?t&qok+VNi5Nd^oI&%H4EK794&Zk_J(v8zD3U1i$T`!|diay9H;erNPin@u-AW
z^U`NcMi0Ux9E>w&gjkc)xrnidvMKWP?&$5zq4&IqXeIPoTlMlM`r&y#zX@%F{NSGT
zG7fC@>DbR)A9hTw>8vJ)<0;4Ll}-Hq;hBCD*&I}C-TYq=Dq-LdIvyUba6`s6!KMod
z4b8(3Pfny>JCq<)!5d^V-lE=JLJ5Nz@_=k&aGn3-&S)!0-IzjG?O2gOZUCZa6`YM#
zVeSf$jkm!{L+bqSpr!mG^r96>CNOuAQ_AUZskP-+gO3J>qHGx@Y<G$_T5yF(Kdh^Z
zZV*&d?;Zw;>wPna@Vyx^;oZ|vzx06Mu{?W+7KL%VvCBOCl$cqnxg}HjxBpiOcbtH>
zppZ6W?PUj}?$}!LWk(Qh{%kEAL~CLWMNZQ_I%G^6H~?Kw9}u2@;Yu&XsVVjZIfBtq
zg?cn_e07hxA+tv~MEl?i5s}eCPON5`)swB!8y@uSL~8M6O!7TuaVC?VrYu$=wE&!N
z+Qs2g6~9h*%!8A6(_G%mqijAg+jhf+SRiC@Z+>s!9S<kEUHLf=)_eoKC^WASD`2U%
zCo`Xoc#~?*M|C%dRUzH}_Ya<PsxIX7hZI5OA=q8bE5LAaKAT#lhn`hI^#DDVS~w{3
zxQ8!1sG|Jn3Ir0kT@HV$AHEAcj)wKoSac@3dV{j@T;bUSUE3t`FLR76wGP$tUA*D+
zfd`r$M8eRTdXF)6iE?SFnY8xPWYK!)!sc|w8A&C_=|7-QLmc<64<<>@7d~^&jWK;`
zOnxvrH+CT6%;|MCJl;U%RfA=uPil>~!gOIkaQ#9(ZMxJrW3jI(HtZG<B)0p^4)q&*
zQI@U!oLinLE53buc$jvxhADi-;$eP#<;V5ASwEMewrF-Theo1iP*V+WOarg@o-!%d
z<9z0)*t=L!D5jd1hx}qKnSC>c-g!&zV>8U`dkMG4G|wOYgXNu%fOP7uq9N)(7SPM!
zGf9Z_&_LP<aV)nnMU(LWFc9*h@4RXw6XH~Y_lAa1EvJLK^%th#7wr2urupHo51S&w
z(3Uw~gRyuenh-k0@_g`A>@$4W2j9m*#6fw)3q~Ci60jDqIcb`tUWtFOLV9&%FbhBe
z*W0@;#5CnpZt(tu#AIkmGnGA2w3ceB!k)+W!Is6gwK+Y1J)V;9kR|CN#*yO!ICBfM
z#ASv9l$-anP^j0SGG3UuS!Z>cu;YE4m&D<u*5$rf1=xh+C3D7$IH}mJ4bBJq%{G>_
zt85w6?`SS6iSdg3@v?pRB0`Kp%v`XOCz+9$vLCJ@ME8IWp&k)zs{8Z?ZY&gIBN!Et
zJ88ibAfs^YfR48A_RWVX>ah2;229r)0Oi8fDP$%jyF8NZNgRWcaH(L#67`)`LOZ7}
zjWUB~4=s!?_~;eyjK#+(8fd>~o}Sh0etvc?J;3Sd1`)QwFTa5@&Q~@$g`~DG*+d3-
z@`L?_XT%i=I4cjrStO-Ivb#KYAuo@?g&>!fV?TE@5$9?`FpH7O%uhncW}S3^uhNsa
z><twNVti6oYqktN0Gka_-PO$Bgfd@Cog<T0>T1@<c+3pHdNTGjZ!DafH|j1JI}D44
zq|Wy!q+Fz>2$|TP_j^-1eewEbTyAIf=o5pRS%<7-pJ<{r5iChP>3sW)yvmrqjXuHE
z-o{dPvQ_;(8=t{E{6bNKbJjaJCv_=vP>s`3XK2B)w84`xBkG`z?b~SSokb%5y(dOB
zPaf`k6XJCa-p9SYX(gJFzO8Wh=&nFYj=<DlV-i#>d%(?9sFFY^>a#fhyCo&N1Xol|
z)atwT1>&IGw`;@uDL|R0Aoe56#@V`1v26>!mzlt(C+$M&z|B%V0}luSHjUwMSr|>H
zB9_$k8ftUrkYs>7jhCv(3dX33&95io4=MazN|!rLLDRK34ybe<Jrq|44JDo-!8AzP
zV8e<;DlKq9TMqoDle*TqdZ!~w0>Lea@4d0T`y`^-V_4_{EudzgvE!&t(n!l?SluTv
z<jyeDAT1Q{03h~*c+JJ&d$XKhPm@B5;-ZIZ#e+nHKag#<TSM71D2^d6u<-uO-VnkC
z!d|$p$2aHYMT!7AINL%-zhy{LNXsRKvgB;He+bxg4K2U#f3pP=l5LEwDJfG0f)=8J
zbn>EOhpCbe>E0k_>$o6Rf|lx;@asXr;|L=9y_WY3O4SQ>gL9u<zWKlgT$dI)n=zH-
zk87hgXfX|9rhvh`*~iKfo7xpy{LA4(a$ye_G5Gw?T-VJkjP!3s#g$=1?@KWFUSn1f
zt3F4d4?x(Aq%4LVFJBStZ5p8%lrt=;nN>^{)zB`w!zYznUav&820TEsnLd27Ne_kJ
zT`ZXKe2wqX)lXh?H3K-amBA>iS#;X6m42@JK-k$^$By^>s_ygru4MpUq2rGAP>85u
zpF`hOL`)<daeDHCF6R^uEUBi-+*PSxhawDxBInv?V_iyF^81oH>8CQZkp?IAQnLxl
zQhh40OT=hFk0E4#9x!}AO<$oL%TBpN)lO`h+jnWE)HAr)*;0XO&EV~MmF-PDk@GXQ
z5$;&U+X-4gZxbT>_3oW$jJrZxc3d$EE>%RDB~~Yo$nlS&DLuJGvwKw7kEfum_!lY|
z)dgbz8>5c<E7L|H88FI_Hyx^6oQPaa2r=Z$Cj-^WxiLQ_Zot|Q8A+X3hC%okel#j5
zlf?GK?c|5YbY(YomHc&enhTFHFbt>d{X{};qihq?k}pT*;U=)n;qqGhP-M`wjP2^!
z+?FmKG8TcVHO;kXFLD%O>d^$wX8YiTaUR?|eP%x|TbpziPL!&h^%&)o1XVcNv)!tK
zgN|k{iQRUs%&rb4fm$%OA!OawT~KYNRUO&)I1*gl7EhJj^(RNCFzOKRTs=>995_lJ
zA{%tk>0??3?5$%;;zpyR*fbaA^R3ZtN-QN5rPywwn57K=`%k`Q+!5^1z6ha|o^yUJ
zgdj?_{>!|uF?3`3Ro&e~Vi_%O;8$)7mTT#cdxO!+X$?y{@mHpIF%OQk!Y+GFK_hvf
zdAQjmSwc2x1+J57p<jQ(n>0hOVb_8}ob+j5o)IsiAI5bQ3cPu}_oRQlQ~C?-sYrgN
z*z~mbwalyuCCx#SV3OY3S+6snf`Zf7RsAr-Co<utq=@6a6(=>eE9`Cvrd%g~meGD$
zFqG|={em(PY1keo7a;E?Ulr+0ij?SDr~MgQ5*syf`+ILs2z3)^SWF;z%3_n{?3KA{
z8e$PLNk^ko5rh!7$!HMbBDnlsYEEY5hB~Cu@ToAT18-^<GHKJ-K5WcHBr=|vNOLl^
zfTX+8f%6ybmZP>Gstk-#X)rb_5AlctUSSNQ9`y0;P{^~JUza~b3%`x4+H*MFoDJ5j
z@MIeNz~;fCNkM_Ogwj9N&bP_((_ecE={9ShvS~L_oHCO3fudT^>@(JyxafMhIrRZn
zERhA5Rm_8Cew!JtgL^%$BC8Rcm44ZgM1qPqgCEUSqki^{tW`7xF5on26Qo=25TB>i
zIhw9(=^jR5xy#NtlwwlcrS=QHGVFeKsyHq9+;Et6-@#~=i3uw$aVDDRAw{+s%a0)V
zp3uPILZl9V>EtKy6|u?+I~v3YetW8_@l3A$rYkqQjmjg#R`y0&HzfgKaq}Xjv-wkE
z5`x04+_;7;a#-_ZF?R8th97#|`@sbdM95kk`%}ZZ&Er$f+CSJ%jD0nsEbG^jh=@Ro
zaJa{ropUR^891@TdEk4V``K`|WycidS=8kXAX0Ype@(m}y=h_Qoh!f1sMKHlzza>V
zdmGpNK()7qqsWhgNZ-)l^G4XSyKpg18asxH-{Z~FCYf4DoeNfBtS`27?;|#P=UG-D
z4{?pnxKMcgp!K2IrSjn*C#c#g2<jG|HRiBv7CGWEc7q#ezzRYfEHRZHu3R7GP}nIW
z6A4y<%GsziyH4AV;{54O^A~S+`FJgI_g3C_(ul(9%p7P!$*p$X>$f|txGn1t@eYV7
z<>?2_1Lz(zYtEKje>!l<Ca_`{kvf(_L6x)@zk{ua$UemvPHOJ3^~jSF3&W-2A+B1k
z_S7ygnV(}Jz;3@vAW>jUjlIkr+T;}EFp|;k-010iqUkv|iQSi9^zL&*X!5f9>c{n&
z#1#WIwsry1Rn$%p>6^?Xaevz0rqRJ)V~^bNrmN42H~D=9x%1$o)6?V=lCWC#9`gZH
zU3}3biY`9~VtAD+@D5@lb_@di^>BAC+u4G4zBRdB8uG=JvG<20RybHM@OOV)dRooJ
zv;STezV-@A5x(T*-T83OcIGo><AFiTr26ClM(+F1GG-Wb!+pk54?^L`>}q4pNfZw}
zsK{?=>}j63wQ^jP9;|w=6g+?>Td!JfuloJsN-`c&Yv#=pF(A6brG>!sbazyZvO@TB
zx3k}{@M`ygnttAZzDH84ul(=K{_(Y<D$qgF$gQC_lji!Z);6^B&bX#VG=SWk+W(M1
zLb3$3n=thl9%;h?{f5Z5dh-fO2#6=Z(Rm20e=05nM|b$QnZ`<VyGHS%*_)L6NQVN3
z=~$O!cvk34u|K>ER2p%kE3d;y*JsfAd09QGcwAJW_8Zc`I0q<g6m-~)A0EkS+atGM
zg%#v<w9xI-iir0&_Ax#j7U@^FE8SQ<O|Tj!{%#UZ$o+P<pLTQx$4lwhpM7iUvDjEg
zzI}6TWZuhX98JkU3U<BqN3Gx-z|Kh2890rA_4+BPuv=1C6KB+MteoZa;AtwFoN1@@
zkFu4r1UzR^d>ZjA!_K2CC(l+K)A0Htz;TQZm4PaDB6mmk$}v**r?Nra-g=!utbxR^
z*H+b^$&u)H+H%RsSI%CGhQ8F@RxL}Xn@crbhucI=FaJ3PRi*&P#_#Pl>K|T#GELDV
zi*_}i3kL@!4;+Cbzn3%YpcOhn{z%~!OKw}7LsM)_=|p;wb$}9{2rH)A;??QSrU_By
z$`Ql^RIaXSs$U2_3b_a!l1=$|)5(e7u2vK`p?3<(9MZF4aF$<j9dh$dcvdR`R5R__
zZ2d+q_eOXm>6`pm1H|iCmtC@!Z{iyi+`*B+?O-@8_-(0HQgy}^vYBTG`unXhm#8I6
zOTRj$xK&EzZR6v$hlO;FNoUd887pWt#I5Ko<)UYg{V30U_?D}ujzWoqSMQ~Rb`xbq
zB_~~<NXtV8pPAF-BH*C2ZugH=eNnbtS@U02rt4J7M6et(siM9KIRr%1tamQS#ZMxn
zR=sr=`)6Y*f4o4cpvDo+M*pHU^WzE{T4_G(R)fmoEXom7xIAYDSyaWz<RYY$tOGBq
z*QcL*Yg9ysWyD#+&tHc+vX(4ZxHp1Bgfq?+x2>s+I7zuEemDXSCGNBDeNA22#+9K*
z%R_fsvX>|YSym;<1q=q9oA#{br~J5=e_7bo6ZPJk)C{u$Rv}r*c`$Kcw|nmrZnAt8
z7)j**7@O6HuNjP!qt6R#0qtIELG{b5u}UiM@d?Hx-1X<0bt1OFIoW6UFOVDxj=i#J
zdKC^QbZkH58YJp!C){3XjqlX${76X(P7g0m7rbFNcW(1@nhid#|M<s+^e^!T%y78~
zz}!|rsv&;dL?juo2)!0L@t%B~rz58s2Q%t-y!u>}ZQIas{4*Kz)7e)u!*6mfA~L^^
zxdmtl4iA87@f6c1*&PlnKIE_Y&O~b7Nk|dY#PK7^MGF>`iNB(T{0|+6rHHt(17X}>
z<;Fk;D+kL;^mnz*=Q=MHu~>q7c&?dlgoH5d6Fk+du@z2J#u{LUz4%o8(iYvg;!Z_k
z1rcL5!6VWm8$fPMx8!3r@Z4qGHyL}q-g@;;5N0eJ;gY+d-~wQ{m%%U=Cqf=cyHfMU
z7Z93zy1lmd0wAK^6SVD7ZJvT~&(HLYW#Sa@c_!s}ExN2m&Z99GFEN1G@)a?KuHG#K
zU5S1x_Y}L)ZsacMM*!4vY-XUAUx`c~!|tyyGxWSV>rcVN(rx@OBb0=ae%L}!!JH+~
zrndY56&ykAXLh*tf@=RqbNYw&k;!z=0?k8h#8&F<G?uHE2Pb*_MXEzroP2KItoYOV
zfDwDp6zT=CC^#)(`*w}=$id-At3Qz-kJJiWe*E&<-Hn!@+igHH!T9@vaX2fP@%`}H
zrS)%Z3NEH9Y(>Vj3FVJ*v$1KiT>S`imAE_|$4PBaGr-J6>&p<Dnkm%m)Q+}v6Y?X7
z@o^SRQShi^H=E3I);9e65@K+E81Y+RF>%WZEof%>#$~@Gdf*UZaSV(WZW}ICEg(ha
zpum15+>~ia>+l8@i((f__<VVYE6!>xDE=UVg~uf3F3uyQideps4RTpwNf!z#Wwwa+
z1ZFHGvPndPG;p@PZ{Aa{z@k`sYSbRYPq@0=>FL{oO>;*jzV3Y-gbaU~*QQ40cL`E|
zKk{s=EdHF5oAd;8_+Pl};uabM-^sv427?lhV<@l7S4V)9t2c?_?+OnDmud^8oeTZy
z<v>l;)V+g4e)_wQifW_D1tx-*m$DEor-RR3w8#c0UlPWNzggG$_F1Pn-k0dd`iB?$
zc@>Vg8x#USm2tj4>TU!l_W=;|S2UpO#G^nYIh*_PK*G$jO!Ir0^Z%Yt1A+7v(oY1N
zS1To-guRBU>ac>&-zi8%?W8L>cdGR@MX!uZ3kTdej#`u96DL@qZWBKVRDS+w7WfRw
zi_%PBnLJX}Q$Le%!+pZLNoWQjS|5=XjoCKHFYimy;61{IJu?NXj@}i{vUCLT^5Kbu
zq*zDSar|*0LSNXTkprvE?B{7SQjt9(hr37La>r=PSNOk8Nq|}{@i<yGW(wCXk<vYd
zis-tkkO&rBsVl?kKV@dgXOCN(c9XM(a`OExw*%fSEyI}IRuv)<1Rw%w@?$KKLnO=$
z=OQA+`|P_w2bl^U7bV-65uE}D*7-RoWGGb4Ds*7^GvN5c>y*a`#jlkp9|T8ub_p=s
zuf~U7JcvAR03oP)2>xDW)GM%1F~^SWe|uWyO>dj7;GNe*C;f!t+t-QQ&$Q_@2JOl2
z0`Qbj&0b?07X0AAZWD0d$X>-s|L?K6C~~h@=A*;j0#=#UM{3A3PY{vs<|3^k27mRb
z8^s5!(8f0Ds84q^<$(}8P9RSAYwelZRq{Ukm%UM4qA%sxNB8<iIW*_vD=EvAn{Kg@
zfy+USssB*}IV=-&e;i}UwT(`q6G#dhKVMOi6GnG@$NnN&@Uqvc4^FW#WGDV=3Q0Dw
zSAxzgm4+oqh!uU^Dr^;Oq$<v}x<0yrgZpWrynFKmvm|A&v|(=?lR-e9AIOKA3GIK^
z$U?IvK)!JAJtzE~zEZ=E=PM4jT%C(X5ntwWlBb{N8<5*g1FhL|db5mQr_5c!QQ6M(
z?JG2o9f7=Ul|XFKZ=x4#<Ip<0u}wSbm4ueoF;YuD><GBV(IrTO+`4Ozc*O7y1Of8i
z(*-$qgEL;h8n#jU6+<mrw-Syb9GS;@X<c*i=|(A*(Q>*oSK{&4@AbqIsdj8oP^zi>
z7Z{Cgwc|<79s#rDFZ}@0P#y3E=tK$s*95vAdT`a>EAoJD0vL`cc6X62;H#VxvbH^o
zw>%P@e@Wu%7-pQc@v15wk4g{aC=5!xiC|_N1s_FSxB?cgOWAYrfN%eneE<1FW)AKU
z{zZb@#+WVe3=qot1eV<wzA*eEibrmkO#S|;*1NmyRvc^_IU@Pql9?FDQ1gIu5B-OW
zn4|4#l#axeflUyyAMHM7l^B0i8G+1&(h^~>>oA<0FcYR^o79kNSzwAZ+Z#UKoEx99
z#PLSyJefn(#$O72#-xH<+n}Xd^caVVhcNs)0-#l!L6@f2sS}~OF^YTcb`{!_BU)3M
zq36h2m!6>C@?;_1g$D0CCi~jB=8JiDIzrx?=7O1AxBvQ`@R8RLb<n~dN&L@c4Cv*8
zqRc;DR%hmZV=h8TD@L;{Y^r>;CRtr>_QPPbO;HPlOY=M#jqhGb=Dje)88pT0hW`kk
z|1E6)lQ6#bm&@@#y$=>BAd6n>7n{%ampJ>MfBh#o-Tm){>OZ7<9YnAvC|RN_MF0Cu
z{Kxw{0<YEOIBf9uBl3UVkgo<5Kx5_$-TwbUD*sD1`;T`cnMnZQ@c!@ncxpS?#>U3#
z9&nmfcjOlJ1FZXMzMrCWQVPf0?J<tY%$5KCh;^8K&q``cwl&vrE`ZiB571QWQ2hBE
zJZ0;&tIX;}Kx3ON;Cnj*@I%(lXr7Y4Gz8n<oCejM5IHCMZuhea6YvDG1rEp1Fde9A
z6i5v3K=(5qXr#D3Zj#ftt!@HZ&InY5sc+wkVw)967)D`xHfY(UQpNkk^bU~sR{;96
zWNHqKk3XaUwO0=i0reMKJm*1?1bIp+lM{u@l5a~g<Y5>tRh@#gxpXMaD&*rR44BfB
z^^WF$h{liuZ4X4XE8ap(^&4QJfb3nbq1`)Evg&(NEs=1`IUmG3a}Yg(#g6Gg=RB`;
zB+0sqLkj>!DlLxHd3K(NS^oVwQDK7X&oZA$X&BkSWf{d8zykP>SGGrT!c&0eFc0+J
z^%qdHUDJ>wNX#_eCx7zt`z@eR&jDM<F3@7?cUx5LpxVurD7tl;uUs|#=|ZzmN6J13
z96BorU%s3!x`jFpUBxp?xt<9`Ju0D;o#ixY3-B{JuHQ^|1VU{Ql=_2uG~ZYMh{;_g
z=_D@7DTl0``yhSl1d3@I*`s!%g6Sbhz!mUQ+=K@+f#%&LZy(&!SX36+MLeHZ`=0e~
z?P*owYp<JA&5XI92{9^jp8m(3Q=ptzrZ&mpsuHz^`hPn^sgO*eN@ICL^<6R~7$E>a
zXD?O&>UAB2fV6EzAv(!88Uu)cygBed+7k*i8%W(!6uw-uHWOrnf<&KzWy9<|-pT|x
z`W%BP+$<JWaBNd%7?jd?p>q4l7SB<=Qi1)tk622|kCa+ek6!ihJyIu4E=skjo$i7N
zz^8>WP2xkw)=fY@Ur-~l8v>hvT7!m!Geo;x&H=u5UiZ~VUMEyCdmo5tUODdqmKCz}
zT$oiinPWub*^E`QrUj1}>2&5|fDepx7uv`)`+{LZ-y!vv<|*TF?kUYe@_4t!hw8V_
zqy92Pd}kQq3X}V8_GyAQ<|{bc96Xx>nNSi~MB0<LfLR^R){sxc2?gHoVv?*uR<Yw^
zV3Uc`R3oM+kngtJaRN3&O*+4>h>u~r$~ck6kDN;@>VYBPSMXhzZjR<Udh|?@>qGyH
zhWp81uGiW3!93JqeEd{Gy{w{(WZ^hAm_R=A3y`BAkh}(R|HS(@0AX)4gwtUZarLV7
zXoN~F)am)Q=SRtTHTYkTM%Gji>?ia>(iJ~u4KkX;a<aOGtTew*@<i$(Th0+}`XDid
zt-QZXsvfW@v$~$8ZN-7Dh@(8gK5)hU=_>}lLp3`RPacN{yorTZK(yyydBawL?yX?(
zFBU)t?t+O8$ijZtHHoQ>1>lU9qxG?gi3V_oj@XUL1Z4f#v=4Ndj3F9S$;C=}p9iXP
z7M@`&fHzPp@jQE;Sn!C2Hrh39#&wCzo()3D311&d$Xp<_;T@s{vz0gOlG_Qbw9xbw
z1E<b?rRqhmtA41<x!W?;`}ZZQ<-j@kXzjDspJ<K^Ybh=}G0dBhVVmF}##w$#b><6>
z^o&q2aKXoui3Mo-$>+&%g3|A`%Y7Fmg@v!@8{SsUyDozLyuqSy={oQ^HTzyzCOz=)
zAg9N?0rM%TNT$mZRJN1&3SOX_%EX$AaPAv*)W{nMDlK=M#tg>jeP8>0BxbGuD`e>k
zKm=T)t+pST=SS=9DfNaV7K?z2#LcHq{eTl;-a%=5x+>4xkidEIo4R@F&D>6X+1FTi
z<W)1UiyDmAyZWyO4jepiLRFZIbdZ=!s$+oQI63!64rlA3XMk*DUhiYeyNSXx%ds&G
zvH|g2we}tim1X1@nAKmNPu~@%usot-{!C5rcs;JR%0`;<4Ah%Umc95=9AeQOW_fkE
z+&&M)`5#rP!luU1L*xeO@!2Kd0a#HXcoKAIO40eD9^v*Us>;LspiIGn<3R3k<zqDT
zcrrl56~6n>I8GtS6P#842;<Id;lt&T7e+x^auJjHlMlqVfa9$B+bc^~^9Gi-5dd^s
zpM~X#Pb4}>POl4Vj=fL7b0sx#N>3ylWM=VsJ>NRLa-bswi-oEi+8x<bh5HjLIoc1v
zQ>Iwp_WHP)PdYJNyvv1#@Ue7$gcPARom~RJ*^NPV2pop?fM51^DZ(C~g*aPEXK2f_
z+KO=psOCoEy`r`bAP}{PjItTQcVvxS=<0?jfq@gCb;D;~+r*c~I3pR=<4iP(C(P&s
zSZW}~{s1_hJG4*&?60pK0Tw&dHZ*l54e#F&R^lJ7`vg%FZf*n>i`jEMa{jsR@qLY(
z$lcgAe>vj3FGBfy+bd&!xaL4Gx+v)~xTqv)&rI>EQ{ifvnb&hgv$VEwMjd6*^_72c
z>9}t!(xSJSV#=<WQpWBpGkM3V+eF;8lzp>7w4Vd_SB_9Gs}}34X1UoOQvXnlzm0a5
zsh^hRY&DP4K*6#vl5MY8GCUM|7+ad=Evm0h{uFU<1{EvSUo$wWRDWoL%D?OjL%xW`
zOJ(%S9*w;xC|IJC`kwT>Kfb2TsfKGVh!>n_|LIG+1O%Uc8JTW5afGZDV6!-(@CjiK
zVY|+cEY(Tl+Mru7fD}T57a^%}fikmD<XlS-rL0>`Nof2w@jki${g(VE>;Rkq3Q2^H
zc1UhEx^slBg0F%i)?PeS3-!brz>IO|B-M4-ax0KROMWWyyQD8?)dQ4<E~jG)t@2PF
zYwc7UdLJI6;)yYffn$gQ6ndD@jbgV7$#zLxxoHG`N!PN^5MW*CWJscP%B3)pTn1c6
zMi!m+Xq(R(`=`2L^f+hs&;oIT@6hz#dCCa&U@M%Fa`N70U|`K*`7D6GjA-ardbrSX
zrk8sxtY}bUss8M|E1kdEK<D&mVI3+FhL|KmSQH3UZ7IudztwO9eh)a@rd$#kZvmg)
zu4eXT)>jamUd3MS4{J_fr92SiYE5|!k|vH3R)SgD-@>d`!EU9$OaA^QQ@0a5qS%Wf
zH-b`R0jjNbCvE>!aZdYNSeXsW>TJQ??sf^KOy4;9=er$%n|6*^ukLbPZ&lFcG=7yb
ze0wPIWELFtLvYK3vjx9ehfgZR1q%}~V7E*@UhU{oUQR7x?xer{-4&iOQRJkb+Zd%!
z<iPElxmCE~T-Cs~nfqI1&+BH5e$uwoFkz}V$$+e6f2b!sqRLN1ttzkob(Q&8o%U!h
zDh-{n#i0SCq-rUV_?_kITEe_={wNq(Q7qG)n@YtWFU<`@GClIdF4nZ(a#!j2jrksK
zMWY_TvefsUAeT9{^yBMu(q1&;z$$#vQhHKk_gc~q{YkHS8*RtN{pYXs9R~}KdXBG6
zRwd7q1>e2Nr|>;&4*H_R6SPe~63ioJ2+UnaGf$HT<Pa1j6kNk-e4aW01<y?-Ceahv
zL<0w-CrOTklEmwWrAfID4g1u4p#F<aCCM1@Z3r^go$k-iDt(Z#FezkOL0J)3ZU49h
znl!#AksTBRPpWD3my?la2G^4E%iZ;?yOMs8Chg86eB8(R8&!~oRl|CA(!kwDgmnUW
z=!ZC8%Mku8)dH0V^%MI$p(#A%nk*}ukWmIiQm%H5j1_5~u3`q~v4_hVK@lSHS@4+M
z6~a#938w}}>R?S{IC!>=;cq^g`(Vr$cudgPPlVWnc7)$`7{{}kN@H&f(+}A)D<+N}
z3PPn>=vW7d+cWI$Hi~%-?>T|d^!^6a6&(}IBRnjXW3siGR)3r$oM=lt(Mx!%C-QrY
zgCFJ253^7b6Qa5~EwRP^dG&>4xcCN@ZDR7y4NWT53r=0mF8(GQP)*qT6www>>lHLR
zz-LvQw8<X6uH9!&@y*_DxfPfC-r*W1dyu+(noIo8J4LLdt0@2o4#+VVNpW!_f%mG%
z-TP0Hxc3Xb9&=h(-6<O=yd>-o3HHql8Jf;|ra1|p;&uF~v@lzF7H#SKm7jBJe<gbI
z8a}R0D@pSd_+>(3Yq+cYG?hgUJv&K_&1??CV3}m5c4M#Zuqz(ilXGLHo!YpHCX&7p
zjzU3ap3p$e-g!PLY{7_RIlk-U`#W^%l{9P3e%=dj=V+0uhFPQIbSJ9Q#(nOB;hL?8
z>$_;^*)}B7q`QJqj<Vh+(np|~3U_$35?d7EWd*L_>S3ZJ3#}KO+MFvq@GW4+9IyD_
znxy6t$V&<R$n<ZrCrML`hSdX>qPr5iWQ#Gzpi21X2OpvXE!vQNfIa0#>~!2udkVpQ
zsAH*<P*NIC^!qp@bAJX8h8OH!F~S9TEEGB{W&}KE!|Y8vhbx?`f+kRR36_oFmFprf
zaNdl03Ce_5RIlb2464IZV@@blFvLWx2_?px8n8+0gZeJSd#30%%(bqoS7C9m%uWNw
zgfBw^aQ@Ym*BiYu-j7SY@mL*m6e=lIgmvyOq`@+QNUvWT0w%wsM~D8^JU1flkEgym
zCZJ+>Y(wV>PcoG#-VGfykEMH*Jip&L{)Ul?>W$`Ni^g)GeU81zaVbrvZ+6!DRp(6F
z-hso7824tZa}kqb#Qruu^dB>KQF>epB6&_bN<NJN46DGFEQaSaI(KhTjcL^un|T>@
zvXd4ELSkyeF|IdRQ=Jt30(2cJEgo8(JVvsbSa~!+b}+|OWoW}wzynPnGEty|F_NOH
zl_d>2K8|w?k;ofuN&kXId54#BJO**{ox(sk_VD<wcVp&Dr%M;gb?qYU(%y*g+a@W0
zrIJ>Z`ol{Q<72iYpTc>FCJCGeo-Rz4Eg4C9K<vr=N8U7E_(%<lS)-pYPZf>iWk<>v
z3iTH<&?QD^k-}UiVgGE5qTDOX8I}A?kA5A0J!Xe<bi!3CvY`9WaBgLoI4Up{MoB&x
zlBUPV^0br7rFd$GZYD_8(@cdbf)Z7bOpMMtqC#0~tnZls8P<5$-O)80f>R6OQqO$_
zz3(ia0=PGH80;?ev^|B~P7YJ#Z*9?@^*)?S&@&N8&7HF8aM>q<qGRU+oma6$3#((3
zTB}bi!;u~>)bmW-<2N?Ah+HqYE=fV47`-l8Le!Y^EPQ)z(ELmSa40cq?(ih72Sui>
z!iJ0%0+9}SsPJVosX0tNCnA_^yUIlBM7*wA-(p0KBahC)jqPv`su&LM(l0QM4)n-e
zE8U`<xx(?46xcI=@?)#eZsdKP#AD~Yx8Y%jrK<IAXleccYsCBvC7Yq!Z|i99$pE=a
z&H!xYedr4kvP3mC{a+N!PL7Qy^WWZzo+?;^rex&FpsFo{(I6$SPdh-XW|Bs-K+YBB
z!9?_OIi4~7pnrX0wXz5mrWwQ;$=N=y3#~=NVp~z5$)@=cKr|5I7ASqZ@hf5}W^5;n
z5>>vcJ8>an@X*Or?p<dOv&`)Ac*c$viEoeOyhYM5rEDtRuReU;9cQj+);YZS7k!&_
ztNHu`>Z*#eSG~^-b4!9TYXc?8I~3X;mt*6GUzPXP0=#?2&kw(Zc1vVozFe!EbGda#
zrXon#+WqnYGGEPKISZQ%XO`zQCpeZ^+uCeU-)~bNsr$CfLr`UN8hG3RN29dcb-nk&
zbk-$%<VDkZXm@Wos&Rxvlhdl8I;&lb^S;us`|}tNRw0FpWoghf1cq2Acr%iFn_>*!
zyqB|SI|t5q+*0G!*htHK9&=W++wBQAy!taG6@AzHThZQ*U;1{J@+xUxPW0_#cbr`J
zE8JSfq&%>QP@kVT9-ZNDPGZ`x@^?v5^PkG_g^`40aEm=w?&wYyMZio~Mg3)&Hl@wz
z<0dsOD1pPc#~c}&@q7MzmYC_=HnYQY91Qu3mnxqM`~2iPV3+}eg`y!giw>okHe&8b
zh;aqyMypq1c~NQeK~`nBAsM7~QRF0yv_A~+a*Ouppk$G_tR)(b#avpE^461ujeAR$
zXJA%v#S{^e=Fo*Nms@>of{7CsDEjkWri(GT>)QJTt2lNY`I}C)5G~J+KDt^uwWle)
z_rXT2L!>mNa&IFtDO@6%l1Db9HgU>uLB_%6>ke;0{3|l<@p`#lCd*qKzXln>cj)U<
zYbJ<n;bq?nx8(yjpDeo>ReWl{FHo!uKaC<HQ9H#Yx<tg1msB?PU5S^~TN0NR)aj7S
zE#RaVvhGa3lWRa!7&}#CwqBROdY^F8j43xkSJ$zb=zGZPkvZPi<91h<D%?OUIzRt%
zJCA<Whi2%6U@*v%2$g@-1Iszf@QRj3?!rG$S(yd*u1%9lqckOSIi_ml+UI9z|I9pw
zK7k1$57U0F<^#3S(2VYAvnJ+aSFY)kL3^?uK4YNDloE+(X9h@dW{dl-V~Nm*G3UXr
z5=px7O`5>&+FAcgzEaFjAVDKz{CaTaQ{$xd5aDVr?jS_lg?I%eY%)th=>F%{0X3Hw
zrF-DmD#x!`|4XCvro&L`TJi^7DIGOx8wkV`HY+Au$y58;K+P9CEThNg;M{HRNNNqi
zSbiJbL&X*yO;)JbSinE6?AJ6iDK*W#=uX-*ugH))9*VTGD>$bf!Fv%$bRGr9Qsep{
zhG2jXr5!yl;#4OKe*QHmp1JG$73T#5(mJhZlVFFES6?N9s&M-bYL$A&>PV$B+Jes%
z7?*)d)P-OsZsPFdOI;;N<DlSA`KMtC-B?bNWKD<V)i%8BC5P4eMd6)dk2T_>k{!bK
z?ZO6Ke3F;CPImIC!w6S%W~Us+o^L*y8Qi$O23W_-ZEgLz0jWA6LURtcG_K7=@eA<_
zT0+L~kLw(QD%$bnv`%(oSDczHbvHab3*b33#aHBF3I^}3UU2p%+=}YS%JIQww*9P5
zfPnYJZJL=4qy9w!ve&j{&1XN+owUDPHu}o2UBSbdth@hdZT^o@aVJ=&bTC+RSd$Ud
zoT1}J#sO642nI*#FP7-^u#wt<zzort!F_(r`eAv`GF^p1T1y9sb`|H<=o=jU@0P{o
zk@JA}IncpS6g}l5-^obGu<%Z6rZy#*S$w=B6t4vB))xH%4m-OwORr?a*ajE!s|bQA
zn^tmR&XQ9og<A3AGPQ$5AS!uVkACZ!S`OUKDxf)kN;>ggn}Xr+&Zp-lXjp4dd>XLU
z>GnC``u*s5SkJoNa_uv$r@tAI^=jpmtQ2<;Kvy3ab4Fh2iDhhr#!;<xO<?5ZDKEDY
zll(a8OsBSXX;^-4wH%)F8{)@fML_WP!z_*FW)ML8<tPZ;`p)3@=J~M$Qc-K?I17@0
zj)ng@OtOS@7oFF=9}*E9h(PGmdg}3)wgWJY+{lY|24pKG8_A)sjX3aCV4-*5ArLM_
z3J3-T(e6el!}!bZ)w{Hp0=HsV_yV5#CSntyuL4a=H~>#$fkB#_MjrAYdq@fHSfDTD
zyW~M_T`kGgHo@rO+wx${L%mjEe@`tr^uyEQF{p4P5>v1nc<eU=$+XI`2EUXaP9?VF
z_I!7R@EY@AC%km`Uy^WsPhN)Cwv_dNNCgF}fkoC660j-Q)+5-KP@`A`s&y0JAh*Le
z>fkyBp6QMXY_v&?@L+g<az)!6Sck~*HOqLScz}5-hgZE>EQw8y&`0UW*Szy@V!bO4
z?@w~KQK$RB8U$hdkPEJ;l95$}4Pq(oG-KZ*9TL%chdL<HMQ_j>p_03KY#(Tkq%~tT
zz0Dx9!iLUH-7>=7VccOsaX)eB)t2tDfHl;{w|D>|*p1*F&nz~&WCrH3An~0yeL^3r
zG1J&i`D-NU1`IW$RcZq?u*U>st>{>cf(6*ehw|+qIXzfjgyO~y3F06}*X>HZ38vrY
z*@$h3#!m;+g_==`Mq9O&{R26*SVAeM)+<3iz8>O$>4Ta3?_dFu1-K+-?lFy))7A*R
zuIEfG#C`R8Bs6!JDH$h*Fb_B$O~e7jeR0kPW5JKZABMvwQiC3TKY7`Ph+Ty<I0aC8
zUo&=aXr(W_67!5>oOsOsECD0YxlgNhE5HAOw?B7~bWQmg#mnl47al`>O&vsYoiu6W
z7|kRuOa}Oh!nJJY==*qrbjyJbC3v?)4&*A$_+4BsU8|rBa$8DCgO*p~p&8mWP$}#h
zuia;?lKQGMrar>Fz#X6bMFwZXcPAmRa3=W_7)PBSw33#&8W4-ajk&-sUE}}dl75bo
z?p$NZ$(#ZEME`X~3>tj0xEZN+gtP8IeRB(9FdF$ZsW;Y&AC<1{yJzEtk5+{~<k$!x
zJxsR@=~8J|3saxOX)#pY3IeI^Lh#i!a3X%1;>jbIQE(^?+j~|&g7?IiFTA)kijls}
zRkbJp-kF$t1G}wGvtllov>)*b8eP`!ePi9ET!4PC<Dt=4(}!<Ip#hQH7nqYB{fXE{
zztSJ_i{SY<^j>z+CfQ>U-r-;5AF6Ocf)Fo9ulQ$btR{<59X(#{5%^*tw!s|8UmsZQ
zAE-ZJWc5ez&bu@SJR+Ymi>r>9D2nfw0(;C2n#I>UK!jXRZ<Kpg52U~?>*}mpswMc_
zW_Z_!VmgYCkp&+Un@*S(Pj2kBEoak}BADWN_3F|hr@=r!*7GS+lS7gp1rPH5@D#a3
zpXvlBGVacLx!F9`O$iK{i#7FP^|r(|i7CQIk7{o$M`%k|O47$h48F$=)@uV<C{|;9
zdvv*EPy%ki@wZ0zkluG&l`yvwTAjYasOoUS-vVV<c-ME25gDn1+BNJmaQ4JmlT^k3
zw*}%K`W_7L9jZm*AG?d3bf$NCB8J*rs8BtCS6pqqk(|2#YK#ZD3qsY%=l>$7{$ok(
zl?B=$|5l+jjzv3i2x2GNVev-)tC`|A{f}nA*sJ12eWPZX!6{+LtM>e{Qyy`jz-X<t
z{1|cPYMhhp+`n#8QpQZFcQY3V%IE){?CMY=vignv8n9G5d5Rv(g|a@O9!$t!slQoO
z;0EFbYVU1@PJs&RHel*_CJ7f!iEY{byI~eY+W+<EAu|{dn|2np{rms_Yhed}O^yW$
zz{aPthX0R$(+f32Y9AHa|IdH@@7MU}555~voelq15m3kf_40rJ;9Chn<29Nd>HUw~
z{P!Em<OeBr8^vFRe>I)_FJ-kDY{3!;%FzC<{{4>^won95!2k0fkFOm4V-N^F2a?|N
zSI2TQrOC-8ya3%DlTgt1_pb^ag>$ei7F9_KopKrf@Wituf*V{n(<b!+6?O~#mE-`B
z&%E~m&C{_NH%jp5Q(oY#b_Q}<T_B-O{z65UmkZ!T&V#h8{&RRZOJxX{jURiYzpjl$
zgM?3KAX_v*Gw?ki{7KbR#f8d1DyB{SdV($RIaog*UN@V_?;1Dt0@~Od=xyEw`YJ0_
z`>v_0yTf+?i3+n|;{!d`HNsx0#yd<I1=oWTg$D2IRZ3fcY3@yc1LpZT*aJaiW12?x
zo-Uwd>;dhH?Ht@o4G;*N`r<Vv8z3;-wqx?X0btx7ZTXoej)Yo%ckZCxXVEFAcx>qN
zCrAvmLOW-uix^sMPj@EEkV}|aLf+HLG(el1UI0*FhRSHERc2J_!y1VqfWRD0-9@`v
zS+F-OL-`hhqnDilvbsHhr8%qdik<#*c=0dWU^OndBy@X1`z9lyeGn9k?+5zJWla1!
zv3M__{~YWCBVq8v2em+J?NCU2GLFu<e3LWv8+s~agfl^p{6PkG^Vxjlqew@;!->>0
zkbVB;1T20Zhx;G$R&{aN16R$Tn7T=KR6n8rp|-yMp4Y`PDzC$e<T*G(H3E2WD+Cc%
zS64S3<sQF#eK>Lryj_**USQweGOuf05eUNnH<efDHci0du@CJ5Cy6W@-)yZ=`gj0#
zc+@0NS}56(Ky&GEDO6z+_Dt8jS-kfUk@^EFL+2Fmt!~ExIb_f)6=Gm@zGS!5k+d$o
zq?@RB+@$M~OJ>+NFgzay4XmY~-qN%7%}BcaPWUgg3l%FQW%n?>pg{^LgNQv=i!IB_
z5W2p3iE~oVk^WIpq9iaDz6I?LivBDz<Zt67eyn+x)JzmQ3$h@`gA??l@!(3luSK{A
z&ylOa$bFF9{wty_z{)x>a8O6BM7l%dG&`-1Hi!bfyC<jG`X@(zd+ivEkUaL(8!+*I
ztwCPWtE<J^eL_8ZvD+l3TZr@22xfDhFN_?)l|RY+L?P`I?cr!J7~N06<3+4po<|YG
z+gO`VM)m>X8H4iYEU2;8^7rg-{vT=C?D1XZ)<a2PzSM`<={ooNRY?E7H>tRNg9@K-
zXUC}^htb-DgAvh(tdKF_5b#+gsBzkxxzzXmHJWk#eCdb7mEJ(DHo2P}2@*Knr4fk~
z?A`u8xFvzD$`_gHa8Zl^NumNP%?E)3b6R<SLRDlT-(j^jvEk@tCsZSUS6nwwF|FwW
zB0V%f9zpjG;A94+*v5qx3Dw|KYVGf~=(<V%CE5#K^M#%`>#2^MuIp2LPvtRO;-bj^
z+<`+sW77kEhukMjPcxl|qc)9DGTA?tDMenzyzmf+*=o@kYP3L0mgvTpYAOb~k_|%8
zYvpFich=9lc;VTh1H6;$H)lZ6UCpD2CgmM)e?!fc6#Q`1F)1NJhR8L<<f0Xo@`8s$
z-vdnHOEqQch3PCwZ2}QJ^K2MOM;NAX+zE{C9u$bDTR_1fP`7@5{bmg`CUjCOuX8qO
ziCpb+%2=x#OK{W>9dz;e3)uK?fy-+L9Nugoeg$0^Q)oBU+rzdL$U1ehj_*j3*wE|R
zf$g2`8oq_Tu1rB=6!Y$WT%<E&5bw)W&2kaqYyTgv;I%Ft<m&FO$Vf-R`==5g8~IVj
zLctUh!QkfLp7t@TQk-9M8AlwZ#gEw%75kWz^W4w(4G`87;h>~v(F5;c`0+P>9=idH
z*<H>fg$lmbp^>k?!3Ytkl9IwiYK&yW-muz;doyIt!V!1iUiE|T(j5k#dv~#1Cb!oe
zz{fM(sU}G`AHSILH}10FzlzNz2X0-;-D|0`*{->~`sl~f4@M(go8;@2W@`F{Igdv+
zI4%F>egF4?^IZa1h^yCh#Doby)PXx@9^y*>27VEqAH&e*_BXf@8q^f|J8$;vyuVXW
za=Z*73e$Q{Z~ceT4uZ~c1a(1Jl>3nJECbo0lkmz&3^%tKDkn8Cl*;7redj5E+XOfS
zVBb&vfk2^zVugxEKCn|-K%@6u&!%1X9ZF6VFyONLX{Ut<fSWw>pc_2-_UotNO*@IN
zqrVbz=bp61cD`!s%9dpD!%ag&m&v_e05bEb@d`CbJCTw!UTleAt_N+hLzEI|pxWg!
zCXDYf2VGJjjlkbg=|IQIDn9%tq#eUbCpYw^YEXE#J5@5j`DZV43Gw1@<14>O8X6L{
z5F9S_X&|-lne=fi01Cuc@t%1IB$Xs1rL+CAq7-sGi#<3Acen*D?)2^vmTM7sP0V>p
z8T+4cPWrKLLWdumAxl1$AzRZCO|04p6x5?_f{NMvMvrk;U%2macVEte@#Jd*<_KzJ
zhOYZTQ#$#I9bCr0gn@E6?I3B;Yly>I+F#CF;9z)7n?38tv(~DW8FGX(5X-Nc?#+)1
z<lkUWv7yzx6YThnMqoIY&$9IM>uNy-fswV)+O*D@IuRTWKe%nenXrN@OGmdqH)qEY
zgw0<(z?SI_S~+?BqLY|1aGzTv{omuJ9Gr{G5AuDltv=uOYQ<kZ=ic1;==k=EgOtor
zW}~^V;gLN>!86?5Qu<f#22=avnBIAp^erjp2J}g*A2J=k9T&fE8l@v0b%o(2W23Jl
zj4bnXyM#@tyj>ITnp1OAu~bc_=f3pQ1MraOml1W&-Y{9Dv1$Rd{bsFT!*XzJcCQHP
z)D7tj=6eGKdv}%i0m!hi;Evyl?Ax*Oh!9sm>qh`V1r9Uz^FtPsQ`-&}WKO|JNcl2b
z4rZhT@jyCb7IBuMu&R6<akw}wvu%|z$DY^OwD~-s>+g3GAREtsLjg@d7!Xk=N<iSc
zg^alBDds{|hY*H)E_JCYgDEW^buu*zbriM5GJQxUK6fI`QY*!v0HH}%i(Zi`Z3G@3
zUbF<^-8K{aGJa$S!Kd0|XcZRVU>Bwle-m*LS}yhi8#LP<GVnZ`CHkJRHlO>no#_!q
z_j+{7S=5h7(BwfWK=?lPIAoI61eYQ;4^pQ^@I%75HdN#=lzvuD;bI~1nL{o2qk9T{
zGEEE%G(m>p79qrvl;G$OCU*y{M3ei6Ezph%9z+}rOAZ;Yet@-9E@g5+SQxwW-fP58
zvIuO(<});#5Veic)CZ0xq}T38Yj3YLNR6!<q3^A_)dR%AJU~AGa2Ya&KWpD(?&c{<
zUiEw>$dBbza^&L^%wts6o=f~u_E!@06rr;W&tW<fz5yPCU+&oFC|#|P5mqSspykSz
z^5cgq!!db;&O7S=I=YrbenqA9-Z&F@`R>Yi?xLYoee(8xyb3L;Q%dlp@7ECZnjLKG
zyO(~R>f?LY%u`8B!Ccm;SIyhnczk8Z89ln3RzZt03m_=1-x%fC=#;j3KmFBbNAA(Q
zx@m8DpzoT$>}AZ+eBLAHDt05*FwL0<MgRRgj_;bjc|A`4<Zg12ra_w3<iguTcP~u=
z_XMjK`m1-briRJOkIlqn{~YKnAVfq;v}btng9-$R^71Krm?g{w_F6vNwBmteEyy^{
zDmic?;<wU;9<`HT2_p01(L$~A>~TwCLFT%S_cE*dn}lq7*=@w?#9_v-4#D-501(D5
zA)D0bqf}(AOxD*w%zY^c3*izh5*JeP;s_E?y9_|N3%eIhC5>`Wp12)wnYE+?17R2X
zFwDS6qC=v*V>HhGuDnfj*xQ736Id#-L7j$3#*wIG`HCRb+B&W%&UdZ1X<YM2M>t2E
za|_j;S+<1xVm{A|VK&4zPpzP`5+ac#RWD$HG#7dV)|S1zBMzRVN6>QW?!c~x<;RuY
zw0zV{EUMgD1QZy;&Q@$!`5()3_RWy^kDduhMH%g;n==e+Z4lF^Q!baFoua)e&Z-vF
zS^e2+PnfRb7vjM`!28;<0$jt@!!mm=+L_-K^h2o$e}|}x&`i=6RO`CsATO~;7c}>S
zoBK)>a=c`&*W4tndD<nC-HF6`_BOqOa>zew*p@n%?tr_WebR`1;0%<Qx&L{1^=Ks8
ze@wjMI;ZS#G{aoFBx}jDUw9yJ_&FGH0=bc5cm9m%N>P_JRkFfW_qE%8yIJZTL2FYQ
z<q)4p(zfT4lx76IMXNTkK9LXH20ibZBWiOX-4gTh(Rsz>F<_J2p&qT&MA6N1uirbn
z;v>}B+fDMFnu34RVH8}64CQ?&o-_FUj&#8ID@|SU>yvXD3wL(631lR*rS7dx<$(UL
zCY?1{l0!~u8)${F-$8w_D3q<Dhm`0Y{WCL}O*5HM$<tyrVhy5COOoI<rPrAhge^I&
ztw%{}UQXnVY#YF0)1M}^i}<-O`?z8J!wU|z=lWmXlpXP%C?O#=03`dEbWA1+or7&y
zJ2fI#RM^=%ouR>#4lypoxDE_)<Yyh+c+#&4W0nxotw2%t>j1(4-W5WgykQ<Z5lhsg
zC#>l-=z{^B{A_)os$v>Fv*R*?gd%w3)V%v4>48oM&o3)u#I{F4hIoW{)4_RWpijOj
zkDSIFgw^|jeK93+J^9%+9l2U<fj@hN2Jcn>uHd2nlYzs-wpsiGja(Pbb?&}!LCY)g
z;x1zhBit%%^v<=K>XYS6|NB4NjGkvM3l_Kh+P;5K%SRCs7!g`t5zWz~R$9?_x#=cZ
z-C;sp5r6SfUTmYYuEYO<()uS6`=qldYAWe62SN7-oUk&N4Cz3xUlEM2rVjaXFk$QL
zO^4#KuuMU%d*K<?!4;vrGz(3SIi3kfvwD|02M8O~W7HZUazX1|&v?4WRqR3@D9`b{
zEk~dhaZ_4e=$r~Y*4reW&_ax3@0<HDQ`t&&-JZo-rMwu6Gn6RXC0nqKH&j<JNV}L<
zFq+$_Buq(^@kwzmLWFp?1P|+ux4dvbOojeAO(c=>G{H4yG0mEaG}r0n<?gv@rMnU5
zg|0SfQz6oU!ohmaQSm0DYZ8s!TjUIX6PM@}*6q>DFEv?Do|3|HT{2aFH)bf<7qRJj
z^Q3mk2Yva?8s*ca;LsH3b|0K@9JMT%5fMAsn*^cu%Y&`J&-g&|uiI+B7Z#y_n^yXa
zMwj>@i91)V568Y_UnoiVGL=~;{BZ@)Rr8nWNeRZE`cwH+bEPaBh2bdrpd@6QwOR2g
z@hOmv9y-iQzX=Ib4{>+F<0v7?f^ov<eohOtmE^Sgb3O8kDXs9VXh3UD@rhBgfAst=
zI&vV9Qubo1l?&Cb*0h5lF_r@otse`>iWaYPbkU7AxWXA=oj$0HS&t4w+QSo-VGx%z
z<8?+Ext^muQTIy6>7iGXh7k_M<YBnGTrgD-j`T++mJdF*lNNkDruAJ9DH?`%nCF0D
zx-nQJw?U(K{LBp#UASXjZ3WP`z4lrNSy9W@_?Xx6FY0z4mkvx+vmevGRg9jY@#k-z
z)PN%s3y`pPVq48xzAck={=J!A6s~JDZ913tqVKXagBAFOTJ!Y8Eu2J*Uru3D?ieI=
zc}re*Y>!--gv2sLsH%6}8AH!_D{isDhk`8IoM?G!dPbc&(U-i1qQ;(k_lwH*xR{U_
z347^X84X?AP`7i!c`GXBXaDRRXQKq))#Z;LT6d7-=cVYret)N4;!6FA>NPO`nt8pl
zdxMyCH;``1$k>3dpd^{R`7&^ntJ8@qit+6X7K)(ZN3-r-ArHZZA7SL2(NejAR2tNx
zNxZZMPL~*VA{`SJ#p8U-JevG4E%`jgIfT)3%V2G-sY=gf=ME)7`6RhP8abemy%mCE
zz-uZz?@fLHu=w})#4krS1cJ`lD$sCnaD>}wJcKDUysLNl;E$7Y<1%o5Ljmf|z>5^<
z@Jv*APwk!3yN~1{O%eBu7nAeJY$y#f%Y87@<Cc#!EK=2k&*LrhMl8r9Z>%TPtH~O9
z_7+qJ;AkH{joyq_`ybDaO=>2Z&v_=7KTHh$xO`&mnAV{$vvFy8Fzg{K!dcoYR2x&9
z!!!!V9MAnt%TNW1Jdx3GmRiY{JzO5+-%zMLEQ|WAfOLIR`$b@`Y_$dbvjRt;yXWx`
z`dmve*=EZxH*tOA+``Gw@KfaOr`ib&Q)!`yy)=?y3!*+;g9<Ogh^X!ujk?=%ur{;Z
zwJ7iIM!<|xqclhbD?X~)Od~R`B{t<rG?Q?xxx(Y-^q<Yx$*0IQ7umiS59B?|Fnj6l
z=G)0E@h7K+QM!2I-iag5+M88Jg^N1dd-9|oe^onA{>DGv?FV#crjV!iaG^p$1l4D9
zxaV~JUSsCBnGCzism>&(<!*d%X4VBd15=iF?fBDB{7uP~!NAz${~_%yfU3+JxKTmW
zgGhO#yOHjY6c9uK3F#7$l9oD134(NYHzEi~H%Lp0fOMzQ2c+x%ANSpTcYVKa@7%dF
zJL6dgdCpV+_+_tJuC*M|ayQB0n@@zR1>9`%Mdk?|<eVs+oGjHPsIQ1oSc|6ZHgi(s
zWGWEFFl{+)T~OOo|BhWOQ!|?>ZvLB}`aG2%>B3?>7u~(lz)kbh$bzl!9{4sAbCOql
z9W(6O9gs>{(#O~Q{G9{m>131De9WrLy6=&H+4=N_AR_MWl$}~$W3_O|xSt<Xins0Q
zM(NZ+v>0&r94ryjS)`smyi$RbChw4$YVa|t{|HW_h}B2?6JgSE3s>8+d>MP+G2<bj
z{6rYf=0jkAa(x&V{Vs%}vATl@+mcY|tDU5%)M3ZXbUVivY5zfxdHB7)sp-u$a30)0
z9Qx8^BFxg@XCi&2eYBJqg?r2oKe42j&(m}LylCe~xM1HU*O=k8BT6*}SFYJfjhGbo
zvoE4p*A`51*oQH)?{DQ(KC!}#IgW1hSVQhdu{>OKmwnUA3(KU%Jh&k!_&ua{Y!my)
z(_(tc-Vsa93)|=EuQXPWCW4;c=Z5-7RYb*3b3g9edre<K%Y&zUYv%@HeDdY6RfEag
z1L?y(BBkIMHy7k00I&2NAu!~S_?fCh?sWG}_%pkBfigq}B59ZI2FjH$LjA^r-##sN
z%a>x~;vl1t5^mD?dFUi9{IQSgjZ5IdoADPO0QEW5h-RrB0zGeO#dl978VGb9A`V|k
z<zYWDJ`4#sYxNdpG4Y`^MP+0jr||$cI@SYOoTq4NCv1=Of2uSRR_^<t(@ae7?Q#8#
z<w;G2n0~83+7x^B2=aR9o+W1Im*BS|Fba%Oqf<?uzAc&g#lDK^Xd@nd3h*Y{&ptg2
zbhE;yC>x-x=}4U3ZgH`6g<4xNSq=Y7AhRZ1@t&8!=%0<OORt3dY2rat&rt4b$-VhD
zfIvlbaqK6kzRU8a2fVYrDc#ten4O;{LT~5b=SqPZfSi^&ifL*s=-PcllF_Ax#vq-S
z-~#*R`pEpo{#JXXz%2kKBkO?lv3+A6>O`}GQ<mLW<HJcsmIrIg(I)cEnyR&QXygci
znE2U+;|HBYogDB|zVYYqZNWgZxBi3*vrpac=Rqx!HlyKGl4m>?msBeh^>M6DKOi@&
z3Wxih;zyRBUZY^v->Fex;WMfB6qQ83C|PHCUmd8?whyz%7w<N0Lv6Rjd-a$^ondtt
zZ39$JcSnx6R-J<&o<~;|hVc#zD}w<9x}Svjrk2S2jgWPC8+x|iDGCwWPmv4T>Idzi
zF3-gDXh^Zf@!J;(1B?G}NR|@VC1goyeK%d`p5JVL>Y#Ro9*guX0RwY@krHPQ{>NwC
zAKJ}m$Rw~-;<LQ5kdyRs@pBqWKIVN&uAu1aj4k(CQbHA}v3#<?SEu!c&?^f><gNy-
zNrwQw`YB;}Nx70w#7t)dZREKl21BXTxbc#XqU|^zvFb)T?-9EJSct@5O)k!4g6HjL
zwa-v!<aWcO4`Vy=vXcZIYXI>*k3NuaoV3eDS?qhnqSC@o*)hI<pSbOTVaf%lcG%T?
zwZ9gl``r&|;mt;#ZIBfV%5ORA!j*25&!5RwZk_~)27itr_|gMy7bzM6U_J$Ke{~yU
zdn}Gkx9<ZPkmu*{hyOH5rH<o-APH!U(}%GpuL5=J?qZ}n!dC|Ki`YWw+mHcNuy-#e
zu!Ka&I~jpvl^3z6ub79&|GM1Cigyr2@S*4%JO;SY>*E*F57y_l2|DF<!pKKWf9Gl*
zJ`z`%i;B$OaYzwzKA7?JH|L+SeWRc4|6>QSPPN0)?f?+IqEU?cm+^`IFh33-2Eqk`
zp>gES68Df-Uwoz=KuYV%`tS8%f(TlX^PfXAk8!AAO6Ydc?@Q;eiS9_ZV%Q93%Z5@c
zlbUKd-o<}Z>Qu4hd_;2#Q2MUc06yL=2-<s-Mwkw&v-U7XH7*50O{q)N!tSjx`YFng
zKI-ROy=5b%79W~w+LQLv9bmgVhWM$iZOFE$D2;6YA{Nv(4GB4^hepZq40bxl2r{@B
z9D4xD)(Ru29~wNkX2UVzn`cfQV7LGFkaMJ{df_HN@Kdu-aq_0wbp;M1gcjz_7leme
z5?ay%VwUO|jJ9VQ8enhCU{~8nPq6JLybq}jQ)6(8>PjXn!*Nvha3lC<5Jt|xjlRN)
z&nK5c4vJSn$Y<wS(A}>)h6Wf?bi$<uwoprG(oGxRm3(^ud+(Ds=o|)Pbf7QB;jQ52
z?d&|)u?B;P+S3CvrI_|XjpNiCoUqJbdnuzWhM4atiB7>LLb{nDzwW41U!Uf!(OD?<
zcTs8^xW6IS-Qno11<}JoY|$gi(~r$J=RI08ZG?TE)pz|5s{IG7ulh_>+iV3m&XB2m
z2xBT?$}!3{DWliayK}t$afGNVW3@l>V019G4GfHQhngp8qK&JNtl^NaZ4oDnJhyof
zmmuU)W*Tu$>AjF~qWY0C1WhrlZ}Vd-mdDl96$6-<53Ny(Y{D@CF3S8Cz$=ibJcf7P
z#`~_4QFG7Z5F4>PYd@|9TjPf67b-L61gHW-O3)sb{q1e`^6<kMp#0fvXG4?h_y)j$
z^=l!y--GPLpS?kX5YNR15u{X>h{*@#&??r21{b#pPBay?Il#6}9bTOeg`#ACA|C`S
zc3(xy9iTZ#M7%gA*~5ZF+<kE&-(Sqs_;xSpeR2%;qch;FKG^x*jw3nWEb)+K<xYj0
zBemK>N^m?TcChb@_=84}JA9&EE-VTc!z+ot!SM5jwA`r7fT>sEql#S8tYgA#a~WX0
zo){Q2g01`S29x>NO5|`zYre(+<w{Jloz6y+nP<O^X;xw4s&nYknt6qr`wia?2)FOo
z-$NF)^mM4r^^F?j(o#py@e<>4Ebekz?$Fb)F_OQQdsj8)W%py;4<@ON^X&OPIu7-f
zCsI-`_C6J)WzI9;OG&)$m<%)D=5wqu=yq^5gRjQz$t88))rzZQH$eRvRANS+1hWTr
zDoloeX=$GJ9wMZDf^jS|)84X$Yhx)fw-2F0(}8cX)2<TL-V^!uA>-3`x|ay5MK5ww
zACXV?PM~km9zw(ncyV5$zl&HBg{zy`Z(4M7CKiX8Pk_+2M!x71Lh@k*pFL|RsbJ}F
zd)4?Z<k0dr0<oAuqG(oS&xVopXw`@Io~zrK*y<2iaHWarNazhRg=e-mUBdVWVv-7E
zn@tLi=B&L39qZKNl|2zR%$S8QVYCP&o)KbZE+S@lXHG(HK4$oci#3ShbzaBtuHh{z
z!F%pC@n^n+gY-UMvb`-cc6Z6xo%LHBhuF~s^x6sa*A|88gYME;X%(5*KOG`=L!z;F
z-24_msrP{q>vX3@uf6<T%@E%*Vc+QJua%cRNubC9z^4(#FY8Cq>u@R<(%c5OTcR}n
zgxCJMpCF?}H&jCPZ-l=1>*}Vq9(O=noh@ir!ftQ-r}^ZHlQIN*|BsTO`##A1T29aB
z`ipTFMyY=XFeEdG^bo-3ssng<8O9E7Bq|a(5xzJ8a;wt+H!{;-ToZV5STWS_n0`&2
z;&L+hf-i~&Q0`rgk$fF9i|+ZliUFA$KmchXsv#RV@HV{ipepP(FVy<2tb1SrW`>(R
z;w1fz;?D`ei|>he5M{+NRrt-<`1G>Ah4Q9@EwolAKi_TD|5DnI2iK04kE`ephB99_
zleNp|`14=>jf4MF8t@46hDP?ko%rpze^DRcgQ7B+@E-@N`2K#Lf3mm04F2>sSaQZV
z@D1bt%fX<ATP@{ervJI@{~Zj|Bf+IB%r48Q{JX>ax6jehfR2e=cxv$1P5<8yF;YMm
zL6?GA@!($s4Uk?m_RhPTe?*i1u8H_{x$<bi$pfKH9_#-e<FCJ=1Jqu5wt#B<KUE}u
zyFUKnbf6SROV(7<>VG>A{yH$@U2qlY+JPP+y%g39TDl<vbj&x#3Xo-y{v6FAB<@zu
zO{f^*Z~vLKqfJyTG`_c>4)^~ArATi9%u33f^(%D{3*-kAzO<YB<fZq#9?Gjn#b9|R
zo49!nf<kP8ZtbJ%dSTU6wrrG+l~rL@!{0>(Bk5L<t#BF|*`XyW<UIsV@d_hAsc(Vm
z+?OO)<q@ZyVi!vJhT2ksASXQ=^eQLT&3R)?Nno%JQ%@up0DVBH16UMyIabeHfa0UC
zDYh->`v<(={QlBTsHl!U-_Ne;_^0}zv@)OiA14V;I1dfF5%Vm0n0=4tfL!Y(NYAiS
zvN{3+hq51H!12y*IUn#Xna{ZUjx&f_uLC@-Ns`z93(bEH*!a|D5ZV(ytsaVL)Bv2B
zF<D1}Y<u8B+Nr!#aO@LNc!cr;bbmZ?>vPzes=1?zuiCLOUIsN1{n8|KUJV^F45{e=
z0yr7Zeh|_4Rm(Py;;$#;KOe&5!4un-7yYr~_II$n*EUUYBzD-&%L;u7OR0x~`S(Me
z!OC)Gb>#w-e#rUxQ5kN4O#4+(nR}(Gk8e@{o`$2a1)Q$V5I4{MkxjR@8e-?Wl0x|B
zz};9=({=ntU#yK6oB9bV7KilR_1CyZ694YgUO$G9Di<MhMMv|0Yitrni^MS#X~K&7
zsb(%k`S!}t$YO3$mBncPyNjW;izTS&K0Z07Y#nH9_ZApfPe4y|({=AHAkN-=SNR19
zHMT&u_kqd<ShIK1r;!JTlj21&`(<h=vv`4}bJap|h~eqe_dm3|dG+3(4YU38)_PS!
z8SRszmuv0;=p6e8z#Zlw=L~H6zV7w%ZJj!5Dk_WqDp2CM;4arScd&(UqHkR3CYT9}
zW12}B1C^*l;%B$sJ@{D<m=`ya1Y+T5pI-0GZyZ1&15o5<=kO6kbF-H9a_#5#n!-n{
zI2xR$s_m>>e9((5MvJW9=fv*1WTut^=BvPnSh$7Atrq~c%wno(jqtLcfFpB){G{wb
zV>$EXv3|~jkyvjvs;sBo8kCA93*uhpkzh9iENQYYp0o9Qnced}!|Sqb@K`!S1&!xW
zqOSX-+qDSK_<rzQrn#mjbNjqBRKQ<!`RwwpNh2ELG?0iUSY!gF^H<q|T?p@p?PHb_
zv<EJayL$)Y)>G_ys#mAJigt+3^c=Hv3D5vNfZJkFMlnY-VXMy@e~*J3>lX&i9J?<s
zx`~9M_xfB`fvS}5rlw#K(k!JtbeU(7@rU5*MQelKIv)E0WM|f2hl2e<4}CD7ZrWc{
z-G4^mbV@J?UBkWW1*4^TeZ-I0nft&%U*liaWhiEcaY5OZ?6Q)eCyp^X(BSIq3@T?C
z4nDYoKiqY?YE}|^$Ux4+e5PU57r|9hKXF0^Ch8XusSR~x=%tsS;^aXK`zlavFv)hQ
zV5x+^58`Uo*EUzLx<Ex4>>Lx#AfccI6SPD<`LY^64)%l_lMX<h1l4D-F)NQFL5l8|
zpb7&6LgWT$0RJ557aQA11{61qSd)jqidhGAS|{xYvpq1UuK7L?Y6LosYDl~RbSIBf
zMS}1WsIClk=EZTC7uWB0e>i&a1DE27;BvIo&>1d2Ff3ahrh%@=z0CTLctX-J_m5ya
ztAz~ITVO%oa}eQMt8tV|Xfp2<{gPzfrp+IB3#`@O^LOAd00olg_Ipyi@E{`7RzGZK
zu;fp-;c`@>?~u^fq50*_-t#UK(>TPI5FW)bifga!HV)=Jkvh6Uu63tWS#QzK5l<*K
zh)kwE)gIrjbuq>jWt0tDkfY6zeC{Dyh)~c3!yhIGHCX+@$my0>XGisRW7_u@7T&xt
z@9pig&dYwkBkYSV>pfT6vLs!%(c*mL8|ab&uTAUn3%vW}hzC_fe-!9cJW<ishSL@~
zsWv<Fd#t}Imi+i<7Qj>0Dtpz<iPOWmwBE*ir$z_$9%A(SqJwCs`EQ&9i_p8<exw_{
zQz+7P4CiNjH!?b}<A=35tv{;MygB7cHu~Pk*q7{T6qw!9m%&Uw@$yTGmYxCHQAt1c
zPiQaAX+{2Z0eslE<Hya}Y2<;rwB2thT<q8YtoWjx;sVr<j8omFhUtM<#$(y3Cu;p1
zgy<ptppezidmL$|Bn9O|xsXt~5VR=iO-UxVGGi%W+I1}iF7$naNMr^)_U@*O{bFmd
zT66fIcs_f1F}f<aKdjR5?Cag5mnxre$gqVrI4;|7eR-;(j8BT&`>Yl`P^X3(a{j)_
z8q28<Qv4st*5MldXaxh;YCEf11vbe7n*?t!GT4Wo9lxjfDwa}k;}8M<i3`|RlCIt~
zH~<gb7yNyYoM0jbF#Jg?!16UtL)Ldt;<xH1##XZwSwB<lpZ16SM34d2%c5l*>l8fp
zXwP2?p<4QB6eyy%37$P_?VV7`cW!eRu=Bt*CaOYsEr7Cy!0$nfmTDUUlVCa!k-@(&
z@w7stidB<`d70N*XR~3fBza|m`5;+)5+?)Mwl&jG*D6F{v%|BBixNwiZ`Id=QS|ll
z$I!bfUbHovo2Y!&M@ikYcN1WDsBs#bxI6%^EA_#y)287R*Lk-_veb8Yq}K^VC7bvR
zY|K0-(HpqMfawv$!*PJzl`yj{SNQp}#cZb}-@Wt0oMxXBN6__*p`a|_CZ7D(=-#Mv
zaDzJrr0*hKxK7u)-+iccYmRlDd#^rSh%K_dRXt}LkWL<b@fAd^euC||zMge4mqRrp
ze$h>X%87>Ku#?R5<&lqKeLT)K&WPNVWsFuRSnZmS@IlK=21sDZG4ge&$CG^n)B22q
zQUZ+X$A=wft=QzC^=5xPj`b}c^;0VEU)Ak$i*vu6#$SMtWTfg_JvS~|!b&65nU<SM
z!vW>%kMF*;qMT|D^-Yv}o)Zjw6T=u0KnwFw?Me>WPm+0F7P*$NyGJkvj<B2QFe|9~
z4Kkp+af8@3erR!V+6V<bxi0z=>>GTgZzB<2>IEV4Y@B-^%8fivCuiMGfp6Xf^ngDj
zWf<eQo<~qD)lag(KwYD0)B*FRO0DIH2NZX`t<RMZ#+0`S6?eDDz6iIi=)8@eDo`)p
zFc6^j7*r?pJ*@|dgwv8ku+czJk>C<DJ0%x3986pgbD>19e;f?v?2Cicri@BMk6c3h
znB}-*gBlJ7lkF{N78rsC#s#I&<c*dV^@d|ZPMO&gV6=3fW^aRA{}@Qg4FmAvQn&q9
zbv%st;Ast=#&fDaMDCZ#L|V=9u8D})(-r5#Mc+In;<izQlY8GrMZw>GQZHdHsVZ;#
zL6`E?VC8u8s`!Qj7W=>)ck#2PCj|akGj0@I%cCRxSfMvq^9v5M*c<H*g^`()PEq6u
z<tOTBYoMTPsL~HrLDKx1Vx4P+mG@c$Ec*48nQ*|_+98Pl$g_Tl(4iuY7|C5U!M;v{
z;Gs^0;3rXh<0Fl!%JYx<mL=^ASOgVee7@~ZxM((R-+;F=4+bORZjd>Ir>R*lyd#qw
z(*k8R=beTgQq8SxPndh!R1CEe=fC>mG7p^CbmCar)NS%n=zTgE7xIn@NUOxg{l@Z2
zR88B8*VAmFEnr468XH!d{3N$FiNomF<=w4%Ra_frF#^J7LS!&i))FQ#S__J5^HXsi
zW_E-|fKHClptNY?nJV)HC22OvJ2Gbd`_!S9V*>1itBfojrC~>DKv`>0HuH1uWiYqC
zEcb<^u<{8WW<hv?EMLcCyAZ$ihv3GM$(Z-oR79-D_}ww#i`^{yOMJwN8&BrumweF~
zY#A4U*`V?LF#chJh)yjgnyjAht&pKx21D9AGnB<-Wo<RH?-bI8W~bgeNI{#PNL-+(
z=Tdk?hk6a2@`Y5N6uEF=uRpgxRpj`vS`*FBCt$cbA2!Cv=fY)ksGiAXC4EEqO(SUi
zs&U7@QHtZpd;WG31GI?9YkC?|pC^pGWg{!8n53;h_~w3{r4~8G&aRQ&T@7lp1@q55
zi2CcemiGH^#zK{hn(1$q_!34KlzHskN%Fu+8tkVy54+j4M2E>d-RiwZz$@&E>}X@L
zl;8>Z7U2}!W+&IfGEup+3SvgSZGs4ZdENa=qE7J7k!@HJ{4Z8;iB{A??eGh$2F2ob
ziLnpA^~-P?)r{Facf4cYUl`)^Hh!<Q{{a88pEtpVvv%RJC6V9`S4RQGVuqH3Qf=&4
zjEd6t=sV$9Ux>-2JI+Ha)(-L-yBPqkF_fAm;<T|SJ##6QJ1DO4Gc`_>-t?ssr&16u
zYlDBl4pDRp9?y7ok(|4nW9N~pJd2{RGPjUFF0osO+vymrd-sRLZM6MiB9+}%&%@P}
zcR`ayqgqzGHr1S8Pf1bdT`_0P7fi<cWD5&N<R}Xd8RR6K(?)_Hy2HJ8;r@I5rD6*S
z>^Bqm#sZhyZWzBZ6{XQLrsDqk+2^gqk3F)m^{YP?Pntc1sUSg7(<AN{Q;lr7z1FT1
ziE)#IERkLu&P5Qbc@V@xRbHAY!afnOdN(uK1+FDnPs1qPchcd2%o{J+@qn)`beB5N
zmoHo+<^`<jczkToA;Y<U>5C1*l9N`)cgFZ^G_~>~a;HxW5vWNFU`*4C=f}W6qna8c
zNKy_ftB+BiyI;rl2t25DaV9JT+@Zww{pO}q0$5dIKlm0ucPI}E7g9|}-&H9^Wi(XN
zzT>6c>SDPjx^oqd5{(?hFHXu`nI>Laq-qkeoIjPHI($De@a55u)*MsJm&~lQebf7M
zQCpp|Vl|25+-E#L#|{$c1GfliKfA?M%5Hlhh&TGahdEL|$0M>7LSXy!_tGBsUIdTH
zHYQT4QmS1?3K!rAU&Sr2)y28wj=!0f5J};Nh0kXw#Ler;me${SqY!Jsn>z9OgHjx2
zN5$*Xf<(>bOd5oDZRs^?>VzsUD~nDe?H3kLeNNu>-G<Ft9O(foM-ti1khN;G0qJ=a
z6Wn;o^f3lp1iE$kS84m3{^Qiwjj4<{DGX()L+5Ig^2pkr(az8Xk!}M2*nS{;G=Fk@
zyCdWWxRgZ4X2xgK1nXP(vj_EUIEM?lul;)btqD}bN@Qe0Nq@ADLXjv><Yfe;G*%wK
zP}C-|w5P&qK7J%tNU}(3G&CsTPaHACllGM#-2eWywOG#+_}z{aVIZ;0v_Iz4P-sN8
zhq5!V$r5t~uIV=ce5N|mMHcmxvz=lousR^$N=tcT*!eQi?rE=J72gVRr;`pOF@*jb
z7=i6nxz3@uo~S?G5}K{?0Gn5_tB*amwRaPqPCiICh6;R3`{T4GZ^FiyHsq>8Z-$%K
z`uV>C@4iczgqpnRJquAIB_hL1q;VGKei;sq9yN=w13=hJjO4YZ@yp7R3qDhGaBWWG
z_coK<QTmzTN4|@}Q1a=H%EDuFKI54a0`<yvw~!zt0tAoU6ShkDgFSo@CEy2QoqF(9
z*El0t1;)&<OoqjPZ{kD%WMxn{GT<I|sju0+{oEfP3_oYFyt=Oz8M97o)2pFyFEqYU
ze?zdC_TWX(KS3Drbr9N|o#p5t-GJj2E1nJ%V2ue-e3gD1!^8?$8r8i4B}emA{^^>Z
z<H;whZ1mGCfpv+E)P^VV5Jb22bflH5{>U<qbh`k0<SB5==A*C(Inuq30ytg_RR;y2
zBxYL=e*D-COs=aSn@ouvL&XF_4B5-)#{mPNKO&?Th>d76KvX5ZGp(Ki{svR^5N!63
z&&_+8qw&_DP+&2uTM_KN?}v?iYLlW8_NPzUA8p$mfaS5P%^0M_rBD#+?~M}ZuN%$n
zIRlUdX5urW+Ga148Koi%;XVyECsI7u{v@J`{NG$RW;p=hsvLwDSgW$3<nq!4Q$;7N
znp6Yq-n9@~eK#OF%Xeuzekw0Gx`SnnH2{jXfJEv&L;pPT;4@8FHKbo|tcK!P8t9{<
zDdZ)C8Q{B_)U275<QaS^V+#@wddzxPK~U+*-AjZkb);`dOeazxD}-fG<;CRR^N`<S
zk!d}tZY$UU!7C1JqHS4a;rwidtf>*nHn}ZHF-E~W4BYDk1x>$bKM0kwtn%!%i2aD(
z7FltWZ{A6C%ptW5m9$!&cH;Zj0u8oRoTsT0{la($(M)5OOqdh!A3S~23Y#Yl3fw&2
z6$U;F(J$)v#{#p}LjQY>qIMPU{S$pa!DW2=I`<QS<YV&(=~8Hr%Gwzay#fn%2ivo=
zov3uV*oHSVVL{uJ)bzULhCo{H1clp>@R)@I3&fYRpD%K&i^0ZMl%6rwiXIL3O{Dpl
z9mW~$CON7KL{Ry^f%(TjNR~hwgWTLD{+J7_NRq;tY-0x5Dk$O>>Q;6K+h=PK^@o(8
z*~dK`@~Hr8o;okn7m=kHIn8z9Ki1m_+`(?+Eb#Dyc-ZGN5K`oF-QO}9JE$bSKu#@z
zA(V+c9dKP)leVpxDK5P+uYO%ku@<(}P9Mf+C>%aVz*Exi4)}SGsX4)5v%Z?v!BFlF
zQN1YtVVJECsQ|cGWTWU9&p8#KC_KoZOCil+J$}Oz|9v3w%*1N4eJH_s6^Pn<<tGTj
zH^5>`@n9-~Qa=@Ju6y-q{1aizvL8rC5s4)_8v;xR$qWV=JikL=iYx$DXuqxU3YQTD
zMT#xgM*r219KE0QAhfGAXF>P#og)CLi2<9^b<{Z^<6Q!qC-Uvh00L^m>~VlDZMCbZ
zZSrl)3&pOG;fCW)K<FR4&%!mC8i-g*EtKQzAc-+Q^mhxv)ydj4S{75g-2|xHv$btl
zXFbPBSNU0XL6l67!mbP8^M1uKZm&Y>=WD{dP*a^KDM}0eg30YkfS!qn_2Wq}l7F&g
z*(Wr&Ht9f(AUC)GT$vpdGUaJ^>noHQwdZXd$EM3tLy$BS%q``SJ4oYw0TiVd&8XDW
zsN|`&utQ<Oxd-mBGXGK~FYd|kn`4Zfym}iN3f!TABCB|>wdDT};)2g+#%TH8gyt26
zAD*m*!pa#@{W&50NM#}6BxqI3DCB(;Ol-xyEAyV5E~_6^xk5ZgbC6RHOIa{MX}8jo
zA}%EH^Any;ah~UoTvUiv&N@R#Rv++aKg}K9j|0yGKlh?!9%^pAeLvwTpW}3>-|SE*
zchOa0J53-g4AwSWxs6eF2m4y)a@?Lhz>WT#H`p1PDisf)xt&U)`jR0<X#jsLFAU`P
z8m~-QI?$$G@sz2lGCC8f(wIcPuTV$h0!pE;4;f=Y*#fxLCjtPnUCcB4QyK$c{PJhu
zue5i%EIM)@N1I6>v6jCGxXt>rjn}@+vDkbCK7d*<+K_uKwfv@KDgOQZPf?ll|H3*%
zzXh=BpzYJUzdvk$qc89t0(9|O>K$I$e<1>RAR^#x)}w#H=JUiM{COukPxfC9mO}-I
z03K4+z<<s`{~!XcK`e+D!vp!>fc(Ed4~$S=Y0i9}@_(6;zq>4deP#*=Cx27;?u*EO
zIhZ-bL}ALr{2RUF1x{XVPF5-E|8TIm%W)y>D|Z7oE=SoY?@OU@rw0WJuA)(Q6A84x
zJ`R3{+AcJlrip#YL>(x2#pH`h^i^&XrB7(rdksIm_If7KWpH`!QpfxA;GivK{K);D
zkjS*^w{K}{)+UYTGtB=Pi9ch5lRci#Q~N(0|If32K?Q!M!d7vZ?)TyQzn9JH4AQ4j
zgueJU$_F31`WQDJ=>E&q|Nr`hyw2bn)wTu~B&tEG)E_O!xvI>*W-?WL4QFb#n5lOW
zB-=eZ-m}kt2gtU6YH1WHU_3?GH!_UCN#tpwLVbcj0I$}OA^`Fb$cp7Q0IK8!+~Rg%
z2WbZkr1tcxALPu<-;;ng>m(lY$Wt)lOa~KjPeE~JBl@21pnRCYc_~7s7VM=elwE;T
z!`?qf^6f3VLhY)ea$V+p={b;yXb<r+z+C2g3MB?d(me(^9@(!%T_+IuItnuO9|<AA
z_&Wu`DEFMVHbOxr?W4){y+AMm4vPRQxf51vkiC6Qj9TelyhR?s(2HcF=u2SFAnRZa
z6m|vlB~Pq<@ziVq+}tn;j7ZEeO#s220)@c3L%H(Pfc3OG&3npNZU{j5Ef8$s_~5|<
zh(&M$^MSQl0sjA>(oPV6R@Wiffa`Tn_KMvyQ1gAM-Uc1c{nZ<w*0FFC&=A~<_n_`%
zXo8l^ZVn6zlJ9+f4o=vj7+PuTf~B}M=rAo9U~3_rdFd}xdUiWrz%H{6=pEpC!F@^m
z{27}d0M)LZFbbf#v*&whmmetC06u8rghi>te`oQm4%pE*D;$0(3vEDrR)y8%NSm7g
zGP5`R{u>=oeaf6cB<fZTwPX~&BbP7-CO@VaqjmsdTN)!dmNQzx=88}|GzRp#m)t;u
z=!W%g1#xX(va`$Y|8sOouY?j#w527^alBrMr-QoDE#Q)pccrzOD3{gFWxBTWO!AE7
z?p-?&?&u7>iiH-#c@ezFV1-%>o|B;v;12|OKtvE1Xb%}dRh{0|65$+y4dWExAZ+Nu
z1Xl>1fg(uAlYH*m5m-^wfM3!l0&1bHkqhm2uT|f2QeO=^o4q_wb77~MSu6||uNl#@
zNa&<n_zpsx7j`D31VopkRi?lZ^<@2_X5f%kka4>L$*~HOZuJsdd@vYoh2j>WB?U+{
zVXp!RTswYe!5w7O`rAqnO8ncqFp181!_10;YX1q^8V-s?FS@b9;ckmM6&5CENVvX)
zJ#Z6gRe{JM+O{VLGI?8|c5WKTQLOY*iyKc~P}OGe(V*4yZvn*D8Q#~jdc|NAuy#$S
zAW+Dv0_PpcOX~pwJw$Wc6NwRAU)i3I;2SZZ^>&1iXy?C%aziUD#|P)hLio!L@9cj!
z?MpNc%Ep>|_0SZGCDvt+rnx4i)2Jk+s(gD#d>(XkmxLcXY`&=bedHbH0qyl8KVfE<
z;PlW@VQfoOZJ$-DL7_BlrEx8X<J!}zXU(Xi`KkqDq-o`E)OA4=H;YgFR8CjD8g6v1
zeA0?plv{>R5BzO%Z+!L&ID;y;kYT##B<zCQ?!@ZsN%wyGR+>eXUtN;+g)eA~;duSH
z0c^!5JuRv4;c<I+uq6;5Q|jL@eLn+X)WNf$SNe|jjph)Z>;5n3yyC3rrK|ZWfsve!
zwGlkkNpUot2jJNh5bYcX<riCkN#m9=JEKpsG|Y5j@?uvfM`T7$cughwgfGg<jd^hO
zn6N}uH1pbm<@tn(uyNL1;+FkqPxAesElLFQPq5eFGx!MvKjk_Zq9Ipj5|hLH`-QU~
z8bh%>AR&5@rI#Fg2$ZqV2HLYx1sg6IPJ=B1d@>UwnPO=sMbJJs#dB#Uq%M_6vLk98
zuS1E(nOzP7=HRYbllz+gd_ly~xWHV?1AoK>vY!+cQDLNrOHP9b0;@3tw)qN}rN7V8
zT{kFoH`R5EXz~!Os3#pU>BG%OK4@+nP<R|QQ(cdM50g+&1~8<Wu{ZTM(I_lvKFV_*
zgF>L$8h$G81!5MP3Qpbw*}Rbf8i25kb(y8AHbB5`(}#!vMj^HpH`gF*+ehhebF#{N
zlze{3|5f%i@rPMY3?xZ`{L}dwE<sT?(I?4>fql7KoOpIg7i2F#Q{hcujG7vfb)YHn
zy?~X20-|upt$58w25k<~!WDw8#8)Ok<fqn14jTj|`75Jcs+UI4YwV%AATn7Z_cWO0
z(*Yn#I7@I)cv67D%85^q@L07gqnlPa+0TD$2CCH6UV8Otls;o8nnD=yaTzVA0ZM3H
zgPVkXtN~fn^zLS_1^TnifXlvmZ<Qg&ol$J{_G$?K%jg+kgp7S0y?5^gu&+BJQuG|Y
zKClK3-Vp1Nzx0S+!tkU0s@dt|tHq9Anz%SW2Q^WeHLwI`568G~vrLk$KZ3ii2obA2
z)f<WR-cD7E+jeNIE-cGD5b@f6XYA;Gvb<av`)$7N*-ii+OG!g@3A6QZwq-)?gL2*H
zB6rzt4(i6JVs-O4J&4BWj$_lly}|r>lS23wn{MSWt9I!?%=Mmj`NZqXJ!=~5W2zr#
z@w9oI{WfRYQ=CtfiZq9oXEnQDWi)}e&oKckk7|Wqt<}MXucEpcv;=Je)x*bxLfju(
z<!198VH7rkSW$hpDH{XGf#;P7G%6p;mU-O!=q}hnDgHY3kwgmw<Q9HpPVp{uk@iRs
z!QcS$7ODcm!G1&TY5QrYTc(fIHb8w279JDN<UrIRwG0h)dX>(aVjOxR&O4k4Aj*`C
zSZv3M{7#|g(Tfx6%sAr+dx!`M1@Q<{Em@)Wy1P%s%3Z+jFX;v&Dqd&bb0yL}(2?}4
z^9jO1^~2*hwk=TP!e{eH_!iMB2sZvI{G@f0>!2%mz?N_ugyiGr1Jpm0f;`OMbjcnw
z%&tgyhf*5Na=Nwxy~E!-InBRVZlP^Sf{u<Zz<T9gY`}F13{to>F}m{A=Y~)lK+3kT
z_}nY%CZ-nF)NF*u0SMkHdS5+0yEwrh-&_qc^H<qXX%_^@tCmc2p7=-DjbfdeNF(~m
z>IG|@?m2v)+u?$u5vDyaPxav@xN(~@a#Q^nAF;5X(TiJ{b&R<#9?et2y&TZ${hKA*
zs&6R$ShH|#r2#3pJ3AFr?>@d~&+^on-J5Q!nklpucu*vOLxd|B7nF*3llEMZ%{DPT
z;NFjrWn1leD`R6E{-8iyM#;&*ShU3D$d|a~a%W3Iz7<bv8F3!s%Dh+BD?;8}rZyY6
z6h~vl<e5?){mjaYDfP9<`&6eacDauxH+IwXYbKrbk>r<ZLC=0if_-Ltj-7@P*FM~V
zhTN60(Mpj$ZGKfB<NOQy%l)+c*h*;IA$|)54Ze$-%O#a#tFiexf<}Hrs_p(qBr1NY
zDdOaORqRsso6ZKG&l5fWJg6<T0Y@2~Va6oMN{$2M5XDfTmU>d~g~q?x@uhs21U&7X
z3n1g_nf?)SpiOd^L@(Fp;-yY>^Bb(5oKPD$wvaH{y!3^P6sFYO8BinJHu-UN9n=LF
zNZ`*nE!hOa#RV1GRK8pSTiK)GW(#hX*VHD|D<*r3gfns<--t`ci;yP~E>IhJGb{po
zx=C5f<lA_$vpk7jnzIR#UZlqZG`ilIpk~$i``?AAMuWxl(g;!SMd3s^!`T)4uU^a-
zc59p%!J1rAqTP*O=m-&>OYfMV()7A^b2-f?7gTcgQP3Mm<k1Ja59cZIao$$lp`_Q9
zNQNpRxRLNMotD*_5zhWNh@K}$<B--7%TCW@Tg0%JB)EKrb1|=|yHSx=`}p-~fE-#W
z9RS}e+g|7&3wmCt%ARG%`KLWhFs3%cI+xD4^&>u^3g<nc+cwsepl-TBtme-6BQ!&E
zO^TlGz4H%@ib9Vc$6c?pdY+onY)JVMV7o2(E*{s(ZVF4hVRwkwimpIT9XU@4x?_b}
z4s}k51F~y^<0kLujiPNG%%+#W3=&JeSE%o#8ok#;4Eka|k-w$X6J+BdG(W2MdUu`6
zMFyu07(sb=`4$R0!UWq0>QNzVX_A7wg`o@zIid16;X*hYg7s#Z{xVp)itN3R6qn)%
zBu<S-SU_*Y7Rbw}?5i0@uR0C$Jfn%g{_2XpP~lJb@S&^5+0T8P3=Xtmfvjz%i$*L#
z!PbuWxr}(GG}G#O{f<R=L<I;bvId5Xs&7k(w#TuuZqM2eo?qrQ3$S<$UXj=b^I?ox
z55BjU>T<J3<|?@abqg&u9W|TUw~}&c2~W4WaacMKcF2{#zLJP-<mUGM3{+|uz9%CT
z8zi9RZUG@-FW5a!N@3l}X@set{>pLknvem3NBHJBv3AE>xTG%$$hxOl_Uu31^&Kt^
zG44-hi6g+OwU&AP>_xG-)B{@9!;|3kG2#LfF@pO#LRvtXpy&$we6{=_*!LARs$=Sv
zjIJ4OuJLB62YkZ4AFFqCS~9xC9>Sa}-#SwZUUk7;2sx59V>4)3`tjT_323dxckMft
z9iCSn<+-re+Iq;hxN@Q~({BZEqcKT*!_`4b^yMg2PYYlROQ67*u}KUWyoo#4i0JoF
zgDi%EqUY;B%pcfmTF*8%_!huRvBCgC-r2DXmZp_<66tTh0Up@IU~=BQEd+gXu_Gs6
z7?_sd?G#xNg+1xOz8@b@LO7ZqYV5!ux-5A_8WbDgffbaBxVGA}Gf0bxnmX;@OaL-Y
z26F{v>JOPD%#)AzL5m8%O`DFS`oe@lMX0NSz_wX_>ApG1p$L6Z%R|garUohZd)8!+
zzrhesUXrsNpDKORdb)X{0G`DY{uasN`cEmMb6i!l5s$9Te&J#t*a(e%bzeS6Hf%(W
zHJ4g0HXv#lHMh9`gtE$1Vyyh-^8E42wC0G)B>5RW<qgJ4OLH9Zrd!A;;q?2AXM>bd
zjdlZAR0lCW)6`XldnQ}m_O9Ae;k@Z$ywCxODZI6Yyz<Z9g8hWWTO~4Fui7WC2P|9S
zxq0g^kGw(!n!^eOTJ1mNchHLYm(_E;<l_p!ZlN?7|0#Ha7C*3&cg$sg{eb->oTA+G
z^5QEKo+x+8?bydDwvt~pXV3ZC`XF4B{^I@;<`Y2VPy|*JRiB5=?jTRxmlw(IM-k}x
z`Y-|m>iHH}Y4ViBDxT-1lkn63q_Nh~NMpwCmtoHzR2Zvg47P0oQ|MC{mR74WxznY|
zpQ{fYT&GMu^{7r5)}3`4<VNQ2hXD1cvT_&AJspZ|8-#r!m1rl?DAy{L;oVV>(mjD9
zYt@lvx+Dio)p2#Jd=gQU>ds}WLiux5{TDY-Jex)ZRfIaS*nU!<oqL~#Zt*%A3@aa&
zjC!5crc_Q&gm~I3bG-kq@AZt(KgMg|v#SzrkJns0>^|!xz9nCY{qe~&ikuHp;Y)>#
z0~B{HE+^3*o?V}7wpyH<W#o&wR#w(fPkwr9BuSiz1>UHjFi=YWMto%VfN^1cr{DUA
z5e>cg%?sw6-Cv8934H2q%<tvYTT=M6Xs^57%pb_LFR0LMcO19F7&VDeYu3`Vf>hU5
zL;<za*y!nwu4lupxOqcp_3HR7=-kqNo*z%I_8bK9z$<Vv8^MNDs2+3CGD@DWq(0e4
z=MwBzSL~{h9VF_E#G`RHuTfKcHcC)9+pi&Z{|1&OqxO57JN*PgEf)!QV3L+#7$nUm
zB_V$%=MddnDl+qYv#n6ZEES}*xOd6pOUN^dk=G<+_@`$-^1qor{<DNhn90p0j1(th
zU!y<`uj$U}z1WBq*5tH=Ib4k&egw{0%Pnch_;!KfR(-5i<C_f|#VQ@_noZ*!7(L}_
z-I5yLBUDcRv-49-5chMksLjGIoIs0@(bjEz>p2Sy^;S{z9qz@tddiQXp9>cQsM}tr
zGAo#lpUb`$_CnP6B?}HUqPc9>+H%jkCC%@5#db+PL~ym>dTVhl25<|q;D*7>LvN;c
ztvwj!?%|b2+>OFCHmn(z9c9ZTT40c1m?@N-fb$7Ua{8>S@f|^-L_XTex3h1y(=DUm
z_X#Chr4_WiVPA$V+Ra2G*bL=ReV1KX5<dCMmhG(dJ};#_*#lc~P5q05s-|&Q`Lkvr
z0;Q+`kl_=$$9S;uJ?-%5Kq8MoI?KJY?CJlg(^ctkj$C>mpeua^hM12Ch{}{Emy+ly
z<CJhJSxxhtk^1)B5x2VWXaNIm1x0u~Zyzk$9xKU;N#HUL?E6Gk__~u44(CMP7JOKu
zvnDke;z8ee7wYL&yDg>}#u!WoPqZ}hx(MfwdUxt?)O#nj`}4KH_T@P_RrcD#d||i#
zVcR+YKgb4W9Io^oYEYAsI}F^f)HYGZ4thJ^_)(7W7_5@bbdX1dvM6_`%CfWDY9V-F
z-1TRee%QPGwECV^7R+}RJNuvaqc^YO=&S$9DgL#c$Ds%NGC9c}aQnU~ybe&?8F2&T
zg{3m-EJ}+R?rWTs(h{(ko@$B0Rd{u{Wpzy3#ZJU$Vzdo(u`ZI-@zSQJSaI@Ia^JaU
zVwX6HL6pvN>8;Uu!q0;D0QTWC`i#x7#j1nVwfK?G*!S!QK2dPBP$~(vZ#&(tfSj2E
z>d|B+6~rk>^D<C#0zH@31L^l;d_;T<g(+mh$gP&4w3*-u1+~0<;rQUYETSzx0mxPv
zkob^wPJ|GHGxlAIQ=&^OSIt*YAAr-7&{fFD4<c42Z2|SE8R+#~INzutZ8^na{E79-
z;bXla(1`R|6r?ot6gzqBvMST=h_{ePh`<fL`8?Y`P5-U#(d$G~oN8y^l*N&jT5%Dr
zBbaFMb8|vJ5bvt;zZhm@e%nbc(GR9A>(ofFfvnom6h;vP0@l$YT_KDc3ZuERp$w$+
zBhYrbZ=g@8h9<%ChElL*-D-b|MLY70RKPa9IC1Y3s>dQSO`eC<Nq?j!U=+k<P)IgA
zm7Eg+G!Lur^fqnWy+-RPp^PDzTsHjEWI2G$_};LmhWD1w0Qj)8?IA2g^oN^{P{Ic&
zBnoWWi~E7x%2>I*gD}Gd=s58w<PrQ)%X7RBu^6RPrEDgodiOGRh<M3Wz%{)%TEvDA
z0L_jKd>9F%(?y*L`pIs2O#Ybuw!|$HF@!h`njUDwv3`Te#}NU~7;u$t`3b>(pczuP
z`Y~_mRL*099-*FIWb&I}l)I<^r_lC0WT57P`}&Hg)?B~an*9Wzn%rOXi-rK?IPMb2
zz!AmqKW?gYCJ58^O<fBF)&%PCfKIjYUJyiQl^Os;EbX-6I03nqD>+EVesn5Z;XR&%
zJAtE$8@?0UdKE5tiw+!*j@-gyH}>gf&ZIOUkDd+K4)cDs2PJ6%Ot3JYPk@r{4{&5z
zQhq0KH3FLy9z{6~s5ktrEy)?=)3(W&E<c<#97)_43AFjOoAE}WoXv1#67mQDB&6g8
zgocbRyj_!PL)1@mJXM#y=KNtlBoF4R!ONcTakaI|{c=|%w*0ZjRFZXVmk1yk4?bQ&
z63>s5oKvMS%|tKFeqeP1yt1~!AP_<Jgn)X|+IHTzw!u~W#UGKjSH5V~d%Jw+EE2}M
z-f{$BVB>Y#*KY&Ne}hjb2~Zz+b;_Hw;i<|%z{;-tm2Ii|_8nBIEB8se{N`QwKr7qc
z*>km@CVh#^S>t1TXN^0hIu?Z$h74~4)OvFi$}umA>=Fj7>X~r8k(GsoZ{z3|Lq^#?
z+%y>9{Uj7I73H%q#Qw6qAGjFm*|`&ZMbnIkR_-&w=-)07-`~l$@j%V6t(hqA75rQi
z|ASYiU*E^BzR^9<Yr;X;v5*~6dwPeuP^%(;&P#7J^uc+dC~9It&56cnoQU1_%B+2`
z->7Jny}O_&nT!F@m8lRG?boPITdeJUSDP)7y;3MTG7<F)oD*kdz`+x@aB==7f(tB^
zsAkGM6XD*kr2_8Qmfwv9e`>I3CA<ZEmH8EOdpZy{@>C)dPY`E-_MxD-3f{kv4Yl{d
zFPkQ0*jN4?nEEe`L3sl;5$)~`v;05k7ez62j;Q0z2SGXxnOP?;<@cC=|NUl!;JM`s
zwi{6b4=$j~o{P_cm!m&DI`B6O4EWz+I@Ft9#gDUc8vpp&|NRN*7d!ypQmy)_#{d0G
zrn5oFdYtrQ_kSsZ{2U0-ln=J=KmLw*{`aWjx{#NMEhP5j58KmU6T;s=ghS=fn)AvA
zzn}TvXq;Nex}!_Xy07|w|I(DqU|mqwlGE~^#l+u_@#km900;r%(ON6`moKdbF1U&B
zW8dHThChiXa5L8+Z<C{W`LO!GeCZE>XzRNfck`F2{G9Hi?R?9LtL?o~CV0G<r$@}B
z;?>{3sUjSWc(hHS;tBLy&CJlCUOT8nF&O*w(^YJTdae>6{OZ>Q!9~z^&{-W!690ec
zpO-YYJ*-vLfZPzPd>Mq#s?!t8s@oY#HVvTPMbL?uGWPOVujYBzZOCTwf97%a8Vbe^
zV%c<GtS*C6cd|NN=HIUzB^=xu?u=|M=!6v?p!&-|RrVc)y{RHXi=~K_uLn~#_S{91
z<IF|XE%JcIUb{m;kQCo_0^qe#JrGNp37PDs0g9rQF%6}ZmON?W)VzS50l|v-mk8uy
zXC1)BXUTvce+zN|%=Sl>^SB+^@|*M$PEJ%>X^&>h!oieTN1L-(cLK0!d$2>PfCRi#
z0If@!?f91X&*wq>4fNpN$E?Cq1Iw~pbcgj3RvONCpzZS7CR5Ap1yfnsfc4p?Ca85;
z7X*o0rh%clunt5tO;_73$~^}J(`kO+;>YwGx-!JjR%hD4k4!D2%D)eog8W=}nr}>l
ziF8#ul+=6%N|VY{WSgbX^E`L7*afBndl?G%a%Lu`?nAc7N`9vejYGhVC0t1?59j;u
zyY&ZwEPfq&Sg*)FfJPWfeKZ*7?6WtRyWeyaHgOsi9&#IwK(5>o4Wp2on;*dB@Jr7F
zVoFQHH^3#}rTJPZL30w$jWT$XR!9b{k8U`F6;mxx)zm_s3p}kWDZCc3WHMkmQJ{L(
zOO(p=`@jP5F!Xdz#qvkci}`}v%X}!e2BHL|xON>)EFixW20n`|7`6lZkg~2mZ8Sv@
zQIUY$kO=Sk`q1+aYNHs0#md%|1C|uZrq%9OT8p%Zc8i@k<tZ}cbNr5B^_AZI2C%6b
z9y4VL(&GL_aMVHs$4@}dagtjy*f0sIiybe)7pp+IRQt#=AOp$FZ{Z|7MrEILBU&s4
z>yM->H|;~Sz&FH8{BnbxINK_1We$*$j==<}UH(TirQ>nMtwJ*&5T!54oxTq=TLD=(
zx>wBkCq#d;pqj5cchN@GK<d|d(HFnXoaq-3Gw0c%ECU90#V_cP(kL`JSWayT)2<L0
zvM-FJw*f(t8sGzygObfB0j6%WPF3!>Iq{7b<PSRu2{L~HwmcaHMH`q0RL9wHcJDIO
z-`Di6L7#z<VH%L5^Ng%>{)c&Zjwsf9*+qxtg=MrrPw&{0`fBa2^Iji+oy*?Jxap-|
zBNQT8WHZ}{C)ExmQ1aT$ObwzYyC00xcO2iVr@?}9g}nVS)CX^Im=A#8eTPkN*?7Ao
zNT7?E<@HJs!6_`w6o}Dt_mdQlAfHTZ*)%~+yk*c-*;o5#P%bb4KfF?|CdYM>IVW$v
z5d$wksk-B`i2R1ef6IJj(cMlLSq{Jb)Yq{aikV`vnyd~PHCsvKz3Y~rT!Rr;`^6)1
z_2O*Nxe75uT4jlJQ?Io&U%(Sqa%_gLmwLqQiiC?I6V%5mcMAhl-EKU$H{a3c>`{BS
zQC6(>^r>Y~x%~0q+=DJIjM$(z>Mlf+*6y%JM+e@G(bw4v9=K1c+iAvatTP$E8S#E~
zratH)U)6Pe{e<khE_h9_!8@~pl9>oiFkTYyBVgB<25ha|0FBrZ0MnLnD7&r&<i&@8
zN;~jIClOFNqWj69O@o1$**^vFs7`<hFq~s5_y8WrQ7sGBL%_9H1MlR(G?0GNOB`&B
zb1UJ6snTz1mYYRD6|WoZagz0q)7%d<{XQc=ScQ-<Q^T^dU#;8#4`SyToC3*V?9CWT
z0F}gl@I=6&V%!z76ELI+z0_S`z4z*;S=CA+*cbne9=I_p`&fQssXBor(8vSwopgGB
zLb-9YaLwJcX$)|>dd-u3Y4sr1!q|HY_?}**rulF5N^P5ZelgaEU`ipN;F<z<IKx%I
z@@|??^DKtRuueiz&`F!A_2o<H=W}ncr`YN?7GUQxw0-S_+<KUOEuFoko`eGf0^bp3
zG;?3zGv_azSo0htX>dfJfVOdgEHSeqqphN^iyN<ll6V|hC#=>+CY`XI`=~ajt{3ph
zVzi|3kIz_$9vL*&eMhEE)wee{ZF+q_UUg2Z7HjkvWj%7N)vld;Qssd37RS-msJzU1
zyfN#^9(qsxKv8(2hMK@XJV67;cQKGy^QmGcBCEsL(CqlCv&J2~8|&WtumlrWPV=g?
zd}3Wr&`J%c@{O~}V##E{_kTCc%ROH?>^SQ##BaD$(|m|56+Sm6Az9IPFE)=~BzGP@
zd%F}9M0aVN^w=UF(1&223Xi<vO_3V#KA1H8>h)U>?=-6*LGLh^O%tSWlSn`B0B?~{
zJvlD2loO!+y-XLtlQ9-V)QD~{`wQ-;Wf25$KNble3odFXN01+rH>K4pkg&!Ag>(d@
z{lvpovOy%=?g$%@CX!NQ0Xs1HmM^%RxPID|mI*L8LpF&QT(Tb1wQ)gLz~Q@M67md4
zPlNEp(Zgh>_Q^{i#a$+rUXB4uIben8S8kW4vTN5PN{kFC>2q4*S-P~T5MTKgA3wa;
zxzhT@etH!AYl{Vj#V;;%m{ccFL!n&bKKNsjuRjM!UX!lWw&QC*O}x`R+P6f8ObkDL
zYpDaTe@-2K*9**TtV5Yh41m7Qj-ID0m_?hpJUjI{>WZKiV2j@mNfNCHsoPp!*_w5I
z?r{cg@O2TP(Y7a(l9Ys%8y{=5k2+bZ(4Gv?1?kvPO2tyUU0*Oue_BiJ)3v~b9N=ju
z#Jywlc&}S-)XIW+XNm<gfSrffKXc`nh=m=b1xnv<^7eh!<TYGsDZ4Y%7Q?2u^oWEX
zQ>}0=_S*6XPgwTy{+nysChihwn6LKFj=HSJJUgoOP8}6S>#kSm9wuJror=M70t&y!
zKeAQcOPKkh4dC2jg{Cp_4D>T_0YcT8B<nFq$b%7mRClNe--dWz!2<mqaPrpmAyzn_
zJR_IEzW)ov1l7lgy(ogzNS|)teO(Tfjab<5fh#=pPxBChTHO$*E;OSw<Q_2p`2mwA
zG89Y52?*3OFA=rWEi1~MRh_4xS{oZKrg=cMu#_Z)SbVIaah84St(l++AoeA2-47Q`
zHA9viKl2k;#;9p^A-mgCJx<K`f-y(wSd#6paNcfJ;Y*IIz^QLJ#b#eEWpDY>l@nOK
zE&1I&X~~k#809_cM3v39`Z^fm{#dG}<KI5ISk-6Ve=#2EYyG38##$O>e><n=(R6PU
zYR1Q@UW{8`y73Zk5HH^Jf)T+|5WZC6_Zrz*4{@~hdPe(qQV!}FOfzp5##C*RRK9Fp
z+btDlVcSz=E~c<#hx6ka#U1wee37a%#xSnSmK*nI#8AeBvdMQP!D8J*Aph4^>9hZp
z`Fa5SYx?GWmY$L7^`nqk-vrK0O^oz%FXsioVq37+FX4!y7pDUC(y|%U1v1^mkY;%-
z2o`^b->r>&mA2^!`-=m1d~xYR2Z&7pl~JzfSbqgq-2Oe>Iu1Q3Q+iBDP_*@}Iw^e1
z_JdEek>2xLaYlkg@+||LVw&OZc?%TvE{npix)=h*KXck*`3}MFc&nGTem2$@eX;?x
z4Ust6t()+L{sNp1mr!SWN$!3l3PaeNr|=M79=>}sOK3-5`_F?<9~aFVcWIBto;-Yg
z8hq&tlF&{BT4KV?-?5H5GxL});j^;2GCmII=u>+sudW9`YbT{08|zmARB^7DmytFN
z2NH)DVa6GoQ(3VrN6b4HXRQ)Fajcm2EXXVEr-Qu(Q3%AsYQh`U+hXQ4w9hit2JfXN
zvm_kVo{3P|UY|s6uwZy%cb_=*^k^He-6$@VZ>-8GTJP89{k+YY8sItF&=ED@s@j;A
z-6RxH{A)QYJ_DPZmfdQ#?8QjevkV$($0+{`(iesx{W&1UF>(*RFj{0-MBko_n-41F
zH___+jQz;E^Qp9+3K*2;owCXrPp!@#x`G{t6*VpWg27umn;&SIl~$7{nE}r1AAH03
zOE^1y$wgiXNbI&s*P_%jL}_Lq+Bezb++9$3wvq!TWkg4j;#4P!RB4-|Bb6)tDeF6=
z9S^VF-_lQjWz&e$eggJ+GVEZ)cWQ_mM$G*f3bsv~_n-kIjsJHd(VaWaKl-w!!pc0I
zL)0zTCmbe5Hk&%tiZ02mbb4}s3|%h0Dz5qYT6HRX)zc{|_k0=uY~NM$;!A>lO^z2#
z>kcg4i{wT4SMDP$Eve9fmtTh(qZT5FKCG^rOEl5$G{}9r6{fKur2FwME^*iZk<TE_
z?&<YFZ&+;RGhNN&-Oc;{uACX$KGSj6=Qoy6ja4|5qL(Rk@s}>kFTNrMVk^`uDSF7X
z^unwP)%CvW4jdAW8f|wgz~(RbqK&65h`S#W90<_8#*Z!5g0%QvOz<75sjm84t*NgQ
zHa39P<Rgyb%ZKzBf*mSh0Iq)*l@`Ep1`yY3i9KU|ROSzlj=iQbf3(ehm*TG(&VbvA
zO&xnb8oLCk#AfGG0`zxMhfzsM#HQlEGbR|j)^|<FR*{#KfjF-pf$i2`7>dWpeAG<J
z*tY}XH%PH~3m*KrVfE5dNbfmFs<u1SRo-l{oX#w%TT4I4IFFToNu-tW>&N~CM}F;|
zE8e_LBN9ng=bkEizTRrw(j&*i4a7?rHneRb3G?mu$ran4yws7tHt>8ub6v>c`?G@h
z2f}OPgskggtID0D@&^}SbC_@xFc9F(TsGr}`9-le$Z_7P!o3EkK9bCy?exWHQ3@~r
z|HIx}ctzDM{=<MMLo*_s12UjA2-1xx($XLZC`d`e(BV*0Do9I-3W_vHBP}gT_Yl(E
z>2DuzJ=c4`cfEhWyOwJ?B5>xM*=L`<KM8>7Vaue9XFjEI(B{gb^YPn%-<k!$=9HOh
z{k`(JKc@X{l$!w5d<!_>OKG~_L<yAm2>-l@#ATObh^0wONAr-=Lc%6?&Up9Q1Itu_
zN6L0j?jqU6Afa0B`!pNT+tz#uVoPFj(f8~?YV@b8e#MNM)H{`3tMUv-_h2UNB$5ff
zM`9|j6gTNa7cS*eg4I%a^OM>(`}<e2$?-BT{^NT7Eq^1>P!>#npQtan$w9O5OL7h%
z$OAi+%5EG#uv(5=puNu7n=K1OJ{sE{{6>jFbvhq)DKAY`dpM+=;*r)A^3!=w)%kYQ
zU2eEBMAK~@t6^Qko#g8a>?f}jjHAO--vFs($w;@$C-QawaRJ60YICC-S5CliafICE
z_NJvAUlBTrCm<wbf{?dn+5|kRu9wFd>={#A4LZ+jRjv~9S|V3ic-dwdD20*&Y{65i
z{0(rBau&&_A0v!N7ZffS{tI~d`$;(iSZOHt_b?c^{h%w6Kbk&uDH3-FTKJoz9+<cd
z==@}B7nA!Aj|GqDh}dE$6Sww#+4Z})@4*7P*V1T#!J!%yI1C{!3v}L;qX=1809w6R
zp;ueSK>Jl8*dx*e5AkcnXV>S^a86JIDRBF3b9Vsu;TIfk=$sq-K|ahjtOad0Yd>1)
zVwIvp-un&WbKREv+xht~Gv<bs=_Gx{06|v?pSSZ1cd%L4`Mh1uJyoOS#+Bgv8l)2-
zKe(LF8gz+9R<w%-!m5Dzp$g~*xr&UWUkrfkdV8EBkl78j#KsQi|M8Jx!@ok|GAeg+
zQ=#!T7cPTD4E70hVMQOzflhRLVK+-{r5nw3q!_Y-9y=l@Yk-HZ1Iq35K6to^?|jK7
zZq{)27clDsq5pHmuOD8^XF|(J12yjx5buHX0y#_qnnaivWQVUk@)N&GC5VOJvY-?}
zB4w*U`ajqBL}-oR64_)vR7xv9EKj5uXYn$p>fzDPSIfgLBUEW!0J^CUt^(b-V@R7~
zQrvT3%@M76(X>E&%C4Ta3W=H;OAn@12YBuO%(Mv<!BK3;t9a`~SE}}V&5+@K`HSrn
zpob%o&Apzvgn7d%W8<v-k1T+zKH&QNJZjHd>`oRUbm&<ro}w~|2l#Ki?O3_}=)OyC
zY@TNGJdL1bA|AIKA7klW_jWrcLyGafVluWw6nrd6yFIkJZOs<QN#oB0cR<Jt(n+zD
zCj`r69Fo6sXV*N~2Bzd})-QXvT(O7f7*_@0x0N0Mz4ST|rWKm|u;(i8DpdJed!IpU
z=owAtve@Mb+Yb@0XOl6w-;O2g#QtgJ3A_Y}{4db&O3)W8giFZSJy5+J_xQnuwD(N|
zwMcblIt{_9p&J<uJkyyrm$Pvib$<K_#{C}dfBT})UGTfQw}<}W!2Jsn`1hrj3@!n5
zJZ#Jd>wljR{&olb{XR;DLV+xgZ2yBG`^&K|Ld{G8jsvBCPJ91?eq_2K#d^tU*&kfe
zza0w)`W~4I8^eFB2LJO-XIO&vrcVavqyHgV{mZei!S`@88w+v&yEgYP@0dlS1Z=1W
zx%}7ux$^uk#|i-7LrIpO{67k?fZqMGjBR}VKF|L+7Ua1ggz2e5sOtamryuea33Zx!
z|KWZ6>%7MCJJJOG@!g+T<zL>q4b<_GJYP`zKaO<;vRwRQ_>i&jIZW0pUial;XBHNt
z_~+R8=L-Va3?~GwPKsO17prvNpZQ?oRcPa_n8x{he!Rji)y56=-LHd)rc%|ox6xc_
zUm#(%)Zz5g!>EQ)z}yTlxODN~`Fuj`f3}x!0>oxZ((inwp{;x?okME5aUCJo{i=Q_
zbp}cWIC@eg=M4clDboIRbU^nB-KW9!W%Estwl@vH5eJQgydW~<q=#F-@C$&Vx<LP2
z57gMW{#>5KiQBL=hkN4Fpx>9-B&#$ha-@nQ1RWSvIA$7IzkM|h01>N$ba_(A{!bu9
z^3^eJCshA5;Xc;=8Z{ICs=VzJ_=}yMH<1a@YEFtDO4mVwkF%-&S(*4u7^^XWp#*I>
z*H63WhXqCrlI5Uy&lCyIV%yQuT&On4o?c7GtP6@b($m(S<Ec9|{|xYnch~dn|GZT-
z-arG8<9xJa2lOUYf%qYbzRUy@uWO&Mo;hMxZh(#~O+CU__z)Gj4)Q~CYchR^gVd2Y
z9F194X{ANqLXtOg<W-bpM*a>7(=Dr>Q-P+RWsz34yYX}NTN0&*U}I7Z;nQb8Ek8)x
zXayN0px7sXqdwwA*8u0(q<vw-CC_1tiO1$6#gC9_h>Tb9@rLgsA|fUXZ9mO}M={k+
zkVWO~6te`dWAW*L3)ktP3MiW)*{aYR7-hbN(Gk#N+4ym%{1i~Wh9|q+_aemoT|~=O
z4iY|$L%}zIAGR3!0Tw%}JfLwZyJev(X*Pt2W@gbAv~FJYL$bXk{^R151%m-PDsZ_h
zn<<J`$T`do#g*pVaqW(G@Zt@pnda=VvP_mh2vgT({yA+%AyhM2T=w@KNx?^Da^4SG
z*L+roi;$~`0LoFX{naH(ZbNNONe-|He&jLoD1jyj3UYZ3+Dr2R+KYb`g*)Z?Llv0}
zY@)xMQPUWTQ8J>y)E(R%#w4IRb+>o?)gZVFPQ?{;hox2nSW9L4R>n>_c-D#4$b;a@
zd-WgG>W1F!ctD{?P;UB`rCo)e2asp!=cj%KrL_xbK&w^ssPsQbHXCwq+m~72u#q)|
zv9{hN#>VO6;TlF)haQG+HoNn#2Y<Hp6b%H0XkrrY9(aRtH0~*l7B)GxEu`Jo!Osp;
zhMtF!qKwYNP@-ji&9-uAsKs#kxbxkeQw5nBuj-Ta>It?4<I6l+ya}L|rVM@P34*6|
z38)qjDu6KI&P(<{W1xVvb%?96YXS*2blwAQGG%M0EkEKzO=FR&#XAQeD7IkN5xc+=
zZQUBQBeWi<cAtrA&-C9X>5OxTfBV&1+M$GA0M}`JcJskqWB`zWM8ux}-_Db37d4!Z
zeL-3LK8rsa;}aC|iyZ-*Ib~?n*wL|YA!gBSPk;^_vfJFg!akm~i2<nhNQ|RgC>)3G
zrkbCa)x8A>H!phTSM1-M3pJ>f*$i~p14Da}uevLbWdmVz|0m*GB9N}!+9D{F!vaJj
zalT)r#WbN0&jIbynh6TUDEDdAe5WN3$h`R_nyFbwpZs_i6~%2q)K=m1dXA{21O*u@
zxsA$8g}=Wh-wk(M-aPyTR;J|+J^T{3cQ!UbVUeMlmfSnf;1hdsf-{6lvT-y_9Mp~_
zH3>#ru-vG2TN!-vN=Fy}RkSI+kWJjI&w@VuC~M|@f#NT)7O{AQ%K0*^P#aiM8FDJr
zP<aAYat73*a<-WgftAA42P&10;~+-yBf~kcfE*~#ykD4^YJZP^cl#+zD)2I;mAA+6
z=QSR@G>%DMNwrP&rU<!tos>HHro)qT!aer&Q`IuweED>>K>qz8KDP>zMd=}7i$*sF
z?${6Z^-wB>3xt2)=MC!x_~2&BYU+ODIXJs?fqFjQx4aTW({;UWR32-Ny}kQ&&ljIl
z{_`pD_ZQ6aIPDZ~i0Cwmj+L-pU}IlV&j2MiW#6&4FF+rpFlk=lAsP2!|2=h4=XS*%
z4FT&*m^_qpaBXOs&-R_0fZy@}g*_>u)IqqdF+ZXFJY<l-)@4kBNzUfUT<EzX)%`1U
zms9K+<bJ`W5~suKm_MMSYT|*HE1y{k{aGpD4M=C;3Zl(=;6|JE6%v_OQW@-rh=fiD
zX22<4z<CoNgl$FBX3>(HVlZJ&Dy6f;m}gWhl$m97RC6^fR?~nToR46e1QX%%CXWUE
z!GwpYQTlr7gFKP?VNyA^TZcR>bmTJp7_v^O-m7Eqy8v$9Ru;VpSBCq4m8KDHj1#RF
zNq}==;yQ+@NMU^2RV<6|ilH1kv!)Qy#;P}`CAA^2l7*Te-bv3ebNzb1GXjAv(#(@o
zR#cJY%cgkSW5`(|^>+3)0>!6m9E0&mSGn#6(V2ios&AaL&5fRNr`^GO5v_eOtkTRP
zT}7l-ir|scf<v)y1Ic9SM&HhJ?|!foS!>X!kYX%_ZH~+(%!`>jBlErU4Un$ep7%g1
zfP1Iu-Q~cCN4`e~W?;7zUDI}c`bK@+D$(bex7A^wF3&;mCjVb4A)92ZsbA?QK6T#5
zYYyY3F_Z2k?<F>rpZ*GBLdXs&-`CtUwYDBP*v{G(%Hthl%qwzkFe;>uEB&Q!UOP~x
zTuR|X;CHurhWYx39Far67V%ddCM_*d_pLY2>OrxfLW*BXEgtukn}!V(2HHBFMzZ3H
z`{wNW2BdZ`{4S2s3Ry2F7Yu)2s~EVIKh<|qH85IQ+<nZ7;}!*;169K21K*~i8N|g*
zPID|@lAx0cG2y{sO6zqGL%u2A=AHCM=g$O(B#`1z;rsxx+1&o~no0~hx%kN7Bq`sj
zJ?xE}66{9}r$FLd2}rD=4i7>Jbq~<ZH$8>c+99Ho_Nx&GK0m0eZvm33DWm|%LO?L$
z2jkk#QwI1xcvJ{!Qo~=g8)lK0Ty>{4*u>%JBI8>m$GJLbE52IzRCW9E0-VPBT?Z3f
zj#@nWA|eaO&Cm~<<&kg9YLW`u=;T^~2<ri(Rah#p554hpZU}Tz_-ejy)ohKnvlL)&
zTLS#6UF!*`<ZOxN^*TS?FY#1&SIf3}>gS|_y7A+eL5ZcoBF~Jd+yZNi#7nXw8Y{GL
zk18@-Y%$|$m-i0XNb@1s>C|N;ct@yi87&d<)>RnCSDMJDR)^^s3!<=R`l`T|_WjDO
zGP?yc+}6r;9{*1p&#9(P7hE^mqnTp%a}Y?DxY>zL-)E@_+50k^R>JPvtKWxuUW+Sz
z)LpScQZ|vto^MZc_)IldC*cw8dV?_KLMqe9DNyDasuB71nPXBEp4eoiaa(D_Q+&ot
zeW4H&oA1V`rO@IP;K<;lKFx@^=ePL<yWkk`#Qj<{%yZjhopm*O<Si``RO08{){k37
zF;TNtJ^;Lr4)B-hsJ-Ds^7iBVHR1%BLkY7a7&WxV(wJHo@=!OKPCL9o#&F+uL2P}N
zvd4a=&UZb2S}@D}sJ_1DRYXRER^heb*7|P#Us5Tk<rM?WB2#lGbFJ>1)=ou@eTvfX
zFYQtZ+$M!J_WO#SlNW$-r<p>C1g#;3yOQ@wMq>WN;82`)iedFczOYVn_TBoM+^MSf
zroL^GemkQ1I-QNTl>5pGmrTgt^%W-cF}zgNf5*y-eS;zMQWZmA7*j1O@YY4Hnz6&J
zF|aTl4u5`_8T=(2BfHe3dlY$-SvqIy7!hN97%DNw->i<lk<J^te~9rP(9s*&&$&rA
z85H&jb_<pP=B4y->B{~h23-HcnHUewwad{lh-XK2vq=fLSkFRfLL8qBmjQWfN`OkD
zP*xW(q)vcwkoYQrat*`X*^=Hx7EyaUa1(tP?!lbeIQbqhc<G*TwP1;gJY{w@hcW=d
z)5B1J*O8awQt(06FfgK7SYJ-Jguyxz`TbNjpE}OB@fk1)-e>Yq%l;T~c;x@<z@MTu
z7pM=B%QIy#59xM7Q6$`rhE78AEYXF*@w|lR9_#LqQQNev9FOiRTNsAc3Er2ejZ;do
zU)u9lx_#MEl9}hx^=a~sXtukbUf4;G<n{hMgj#*brQu)Cnke>Lni#%QeR-a=Xsd;#
z_~q7Vvph|&i9}BAR&bncEH4^W4?1VV{bCs|n3SMo?p=c4pqm|Sc$0EMMGD*(I<P{N
zLK_V!1DW&|3A;~PKaBbSZ(T5~h%fZokhvSp0>^@)yzu&k-~14EoqW3jvk?$VJ-n+j
zztU7|d1rd~XxHj-RN!Yz<}2HimL-xCyLuaLcd0gIu#))6#MVCfBxRCzT%?GPh|A;r
zEE7UxlaYulamHM6=#?4HuHk%VpKN5A^;nbHIiBa5Kef$?;iT?T@!S02QQnXDbooV|
zXMK57I<9g^6yI4zy=pb`GGpt}lb^i)6>X+oV+{TFMy92sQ*H&h?sH`}p{KgNI#|Sv
z7rrWF2!NOgu4N&`I$^(QmX{bB3M(|jjzua{wb|B>1n?Qfui{-ZS3XpdnjvPPo57c^
znN<+-%b9%@uJP0Aeec`o4c3j)rC`4L$pZ%7dKIF+#?aGsASXexFcUHJ`~v$i{$6Rh
zHn_(YS$FH^H@Vxl_X$GNi+LlRKrvHZ)7=C<J)(v-I<pcGRb3n&hB1C>_dW_N=Et8Q
z>6E6_q$X#3!d=%`8mHSSg$Iuur#GotNG!5_!lFe>qGns-bNSP#O#`&vhTR@8(J9D;
zYJI~cJ_KKuUXu0{z$^vlYqvEEOK=xGlz@+IlFYTp2X3waw^(^q(rNqdVeJ9-<Wks?
z*{mU^3Q=0~j%dTFrnRz_>~Yh9;xYQBDC4VCE6gr=T|KI72LBXDBP*fK-GZJ)ke#OY
zSp!87`y`;IbNql*bnbcf9=>-QtS=^wJUb}hT70*@;l;Cm1tkO;Dr4cpUle8`G%f|1
ze=1jg9lgdqYMsQ*JC$Q>w`a7=gX0z)AR)8Taa+el8qTr#eFjg2@wSxD<rocgnKotv
zW;1sP9wyW6MWI8iY3_+sOpWvUmQ!X!-h*(taQGx6#jJkJdsabtRvZ2zSo~dvkNW5A
zJ)QWPSe@E=S6yj0<Z5uX1PGa^%<0XB=G4%FW}ne}RyK3@RpvbxYNf4vOQ!GOXWi-d
z5l>?nGh<L=|AJ~ra8gj&YJ5`c6uJK9diw7Vg6*SxG|(lVev_cGetPcBdtXaPA!;Y8
zDB|bqsQ8`qE&67ZO-Hj@q)WN(fGx-Rcy~EzFh@NoJ~k|o$O#jqCcn;Lm10+Xr`1zM
zQ1TpVQtV^TC#~tbGZvu7C&!e_LWZ*KsGw|@?Wm2_rU^^Z+!+!g_1~<DE&)&IQOVT(
z#C}e?rn|A;Qlzn{X~mwYhpm!*rmT|@5{82XR;QKgGwbtxVn>IOur8s2Z$uj~#&-`S
zhOf;vJr8Lumtx`_LL$1x>6)fkmv{JAhCkd2;J{^h&j$dr)?&m!zQaLs(`V!9^AkUV
z?2GjWmUTOEMpb0$UJ6al7yBkJ+nA3yCM>MPMElxos4L^GWmR3OUt2w3>zrqNtEd?0
zV8vW#$fz;pcrjRblZx1HO8;zr*>bZo%YlB4=KzOz{|CqOXdKeMm+l{{<@ZNg${r8)
ze>UOX0BTCYpJGL^#<kL-B0xiD9%lP47X6E*YwN-=VFtQ>ht&qm*XkqJ&NW%HlQo<q
zLSae!@#{(ssSoCYMg11p1x<S+I3Xj|nqtGIn6!7ruQuiy?k7nR<cMg!R75NMx|)da
zjH56P$7D|XDDeg%c8mAt`yf6PqxvcypLl_Z3ikq0fsykgs7@2<6B0b-I;)?%bv<v>
zB1M(lX;$fMkyH%#el$%(CzFS(v|rHM#p8-zvjX+TO_wF{wVx-n&sJ>R3U)}zzOuIq
zUGZa<F|FuDTQOKMm<t_=loY8X?OgZ;x`GAGNM(NksNgfzSz}WYp6H~wSjySL8!l3s
z3V+AmBtUjbiaeWSs+oqdeh<km2L5Bjz|ZV4vT(S}U1eo0uNzG<F7csK?ajApWGWpT
zncDX-GT)=$4Sx}0s$IW#rQl{tMQaUKM=iHbN1T_-H?il}$(bMF;&9067W<QbtRWZD
z*npw1<NQlp3vxX`UF?Mq-~*3!x5fw?*ap&_s7-I~G#KSCjUO0;#~H!!#m1{RIZJ7s
zl5(92%1DvxQ@8k~FhYhr4|P#(voD1^IyKY7pTm543F>%9Lo!t0h*eqCi2$J<??;V?
zlgYAU4F^Wc6q+&Zv3yb|dkx~Te6RKf>hbJx*QjpVEGJ)_O7R%8wD7R&l^j>*)O9#l
zuX?s*a<5g^--R7m13cXK>X4Wfv@L7EjD46$E2(|s>^)V2rWhXNk1av7oi_>PUqL>a
zq5>Y0TG`Uc6xfM|3bCj@ZS16Yn(+XVQuyQcL&UgR)gxYlqs?j8`H#h~opz1CmpxtJ
zEs@<JC3YiwPc7XF*F@JON=?XAl?#mYP-Q-(^iaFXxiVxW=9i}QfnMR9$!N;`gNK5b
zf<}$s);AzOpU|TCY(Hmo)q&l_4;@D}L#sR%Zo?6fb36H>mOS&MqNEU<=gs8oX<~;i
zR`*XIL>srsXxOQWSMadXRK?I1i%%-w?y~T@MS`(&_=E@C+7l(qh^gpjMb*C-@6D=A
z-SL$;7n8${BL6(XG=?%em$GNlI=$m&dRtR9<zKtS3`6`?t?ylHTE&?zp_h0)?8?@S
z!~vH7H4?1oyF`S(YK6hkuLU~!%o9mpt5KYjaBN>=@ndl{0TVZ+6#b<zD-OyYO`1u@
z&*OEz-bM^PW7OOW?D|`k3(mRZ_9Y+EF|D`?62vi`_mxjzorxO-zF%jVaLFZ#H*whF
zgzq8HHQ6x#$OABzO|SWU2c2xQix)|G0~AR6Li?45sE!4L*nFD^uc*&Ij@I7ah!Kjd
z+q^<Sy41$#-Bq)H+v$17W{3i+$tTeTk)NsUo0D&k;&YaodDRs;(`8~p7rDhQF&6sR
zg@@(Y%O5ZC#}bWUq?qL!`QkO3l_pRTXf<9e`749LT2E5PbzT)dz1WZh*F&*qJ6~E?
zexxiI7;O;NrF;E>^yV<(UPpT6FtugB`O1o2BgwCDS3)*P#|!u*)GF~Tk6OsYzrLm}
z%f0u0^ut3r{Zfhat?!MSBD-m6<yohs7(-Z8;*=GSTSrRg7+vfwehhr_rPNjFagm^u
z^wg_kr#=|>@7x}%!+p<CTcU>NK(0_~oQ+yHSDBN)Ej3u}v7tP}-k3vL9~J%PpeB;w
zO{%!Q2t&{(*!j0Ks=~9XWv@1u?sUOyyD-|b?!h0g+HJ_(zjhvi?66xYDn6~%Q8ZOW
zAn)o1zwAA5$(q${9kF}6fRtp^D}a4549=>xAL0^rg*I3KDJllME8QgmHKdH^&;4WZ
z@KAy;J2I-e37^*NU<DmO`NEBriv=s!?!6O1IzOYu^s3xyn=?@{(YyV|AHZ?tgt%iO
zJ1d@rKX)_-2T$aTf0g}>3Hm1rp0N(`2p(<uhF<GdPFC-nEa{LN*}YGb@-9dnkV<#F
zL9C#vBmqLLOA%A8LL4K8e8i%13LrF9w5C?NFQ4rVlSijm2jmBHTI@g09~zRF5v#K+
z!~yl|XB^Vz@P?;&DThjj*D>?K#j$E)?a6+NnW3pvL#K6QjPS?p!AmZAENUoh<n>f$
z8P>0d-})Ctyv$gyKR$L`z$ZuQKQXBH$L$A#th~m=WBY-<=&uoou%kOwhmkqW3OY_%
z;$yKiu?wge2?^^B5FSR=-4K?5i0xnwU?mE15uu!+%_=|@UshWW&T;$&_Z*?5IMEs~
z<I0uK@qZVBrA%;f6~J{KP*FU73uXs~?RR>(C2V_`qxmM!E>$VhKBtihnU}SJ@-FTz
z#3V%w-B=KuaaE<YhS#uffwrzy6D5C`?*e1f1?EAe-g{b!(+$0e#}Ylk5+wfV2<<6#
zuAc{Z>g&?_DB+V)AJa|Wp+ATinF;m`vcoFrkc*Ny(sxKJHw#sexCpygrYf}itY<&q
zLdjL(lV3|hpNH#F;V?^&*Wzh>xX|!%6AYztV;^)j3oXBqAfq|T?>nP(Ha>i>auH%v
zNh|FvE@L-KBHa<(SZ%*P*PnGaD>kMk72@;e%PmkVa4e0Kybs=G`7Ka&_MD08#N|Xi
zVA)1(NGLGX_`^SwrP7F7-VPO<r!m5%kUIt0C_30t=>b0ho+rF@5m5y+uSga0;)sRE
z0JE_>B|Cl`vlFm&^_RISHMwS=QjHs=c7-fV<oQXsUEbx0xFZ+NSSjaq7h|*+Q2k8o
z2N92xQNY8<HmuWKuC8Jm1O8WPM`H#EElYiiV+x(tz7;?o;hsD7+&Zrz=)LEnbEnsa
zv?=e1_z+tz|EAA;13@khYU{zpOwyP!$DA81Y!<UWE?l*}nXqRZv6|u}HTPV8mHv^>
z*OMitD~_rtP3c%)$J%ia*?1?yIc;Z})kqw$1(2M3RrccQSJG!r4v1G+2eH8J-QXPP
zoW}ClrRU|#Ks?IqtJg}m0wO~#6$l;AU80@3XV0P+dyaa}@YW|$4|JgUDds=srxS><
zIOFF!gfpWN0WXcn5!|ogUTPdd(%<33K3~Pto&$JQC^L|0EiCZ#F{B{>A!I!U58I?}
z`Vn4LcOPjm3J4^-tap#kTaA|=_9t->Y+YcI#SO%F;F9y_N50IOD%@oBH@K1F*5}eA
z+GY=s*g+~>0u$=VP}D<AEk)WmK?9AZ4TS8YSwdI_f5(PgLLaa`oXKr_cn8K!J=>nQ
zpQpsP*VgJTp^8ULJ2&YE`s-XcOfRF>X4u-Rk}u0F;o*?u&@W+a>NY)f18If)vScyB
z%~$=o;?%9czz4I>{QZ^)I09avp1yJ=e$G>2T;7q`C^J8EQ_99UX@_V-k1>atjCBx$
z-LV(wq_+J4OI=g0e-?S(uJp@OreRJF#O8k<nOef*jl7e)Ni+#0VI`v!laqa=WSC1@
z_Jj!$JWaSxX|~ZL%WzuH3M}@XUJ&+gICXXwd@Nzdf-7}@^yOmRhi#e|YlV5p8@0|*
z8q9om*c*iGtWrFwFQ1}n^c^VvcP3G0f=%PfsXphY>IuDGTS08U#o6{2yJ1uvzo2^^
zQ4wwd3p!4PowVbvTHW!=7$TdJ^foR-O&VFdP2(aSVq^Dd;ej^elWa<6AG(I<Z%@#P
zDv@NKT9jl}^r|UeB3ID!Uh17HqaWJ^VCv{rDDsfTQ?_mImBmjl=Dajp?DZuHWkOHl
zmc62PgHHC&>0e3`@z{*%QXK}YrAWCK#IwvZ6^7&($r<BOuVy{Fj&#CGu}5~(`R7vs
zv$n@z?jrWl_8{1`?cR?uAlpy5y%xak=7z!^+U2+XrNjQEe!{qVTw37lpl;b)htMqv
z>)=V9p84rm(w&ET_oYnsB&9)MUpr}ANdL2SA!6%NLXSK^xVGtX0}`|!jsp^tGQ%BE
zLyEPja7R!<DAFcf1-3|zY!2OG&SA@-V@c)$D`2n#iBV}JdRYo-GQt^_SCs+N-ZPuI
zY!a)KNSu8QsG`2gP#yBwww~pE!F&4?;$WHSa8Mkvy}M9(0v|q^$%na*b4<osGz#(q
zB_p97fBFu7Q@{O0*z;JQXx2x8okxb7#ZivWlL)as_o`(q_M3F^Q$7S~E3U@ewO9W&
zFR)({Fv%;EiMefnG4n*q)Lpj>I3VBVqZ2#YPJ9>=URve8W@r|aBe7x9{a=5<U#VZW
zP%Ob3i00h?AYQAh{$EMu?{E#0Z?5Rl|HYE`Ku=q%P&I}>7BqjQu5T*=rGc2-=>KBL
ztsrK<ME-sB|Ddk3!RK^?0b&0?4wdtp*&oTg*Y;n}=x=M(?|`gIj(+~TJoEoz<RNB%
zW%XO`>>uppU;nuT{8Fk)x5xg+p_U<L-_3oo-|2s-EL_3d`i?bw?XSQ4kJoMxvo9v@
zx?BO>`d8T@Z`K?g-M9Eho$-~55tKq1tb>{VIph@ZZ$#&R`<;7~V~-1kq83x^0>|>2
zJ0L<0ojCsF+GUTm4GouX#W1cRRg>rHPPgygv}h&z44AZPkZd{y>6aM5n0W#aOYX0A
zg4V7}-`@_t2acU;tt~*&Pr3nx(D!3wJs+?oegdqf>p=A92=V7e#Sg>F>KiW6RowTO
z7;porvRXet33$-J2$;vmfu>O3+8fEa-FFLlaz__%tocrWn=R{xn6;Ck`tw(LkIm0O
z@Z6$hPpWnLHT7D#5r|TTn&1GfHtRm=xgfW8YwgbqAj1YeMNUK?0Z=iH-oio<V)N6T
zfMAp2YbM`{!?bpOb_*i%$;6YzJ@Tso3NntyC(k>YnK2)#a<gswcP|&=bw@VwauDy(
zbsy|fo&Z;jGuSWKWl?8A3tW{7D0`TE<zq3Q(~N=#u(r}Uj9O3DRQ{G7fG|M|UXdCQ
zbFiNN>-4=XsHim8gOc-3Kn}7hpITm*GlI$fVPz*l3s6NFiC@xHOA+pV?X{en4?bFo
z4z#p}C0NLX>!iL3E(OweqhIIX2j$jp+5dU7{QC({gXzu@EfBf}OuO!+5}N@~XVW;K
zkD5~KJONvsQ=0!ycXxH<M`Pi=a7~>hx*3-LFpA~fQ7HN|jZe^0XT-GJ4RFXSP{$Ct
zOq1KM;}zZmV8EzqjoR5!ho3-ky!KRW^e&O&!1Hz=)M~8?GB)$p?!E+S#IbSER*IqI
zHCf&C+l7eqDoBz6xtIKOUi(_D;-4&}9*#|&$M1?vBmP4J79zteT`xw5bB*B*m_(!C
z`XFxd`q~Ldz%FzHem6I?@Q@|wpLJ-c3N1VUzWA!FEEW%tqE?l&2_C2`$ud5`C784@
z`>~+Tuz|4T0mH|@kXpP|D0B}Z?R6Hh7uvxpdP5hqauX6Okx4;n6;QABU>+jK+dfKB
zxAA*!KZHCq*%UH*29GFq(7<%2>nq~Q0D_YLeL`sl(<5z!B4jXJCJY42q6r{-bBO;{
z3}_ul*x@wSISvGxmuINjo;|<KqiljgmIFnL1Y47*;MRu|0EQB&g6nJn@9P2J=3~Z0
z)O&N3*-V3ST4&*B*0Vzpk!P14*YuRU;bq~PO4R1@MMBC7-A{&Wz`S<b2P_&?uzZ1&
z>m@-(HZck~;qxSqpWVFLXp`EVCOPID^-}t>?i{V63LzbM3eBbpT19_CDx?<LI>$C%
zP@|^Tk8wKP1vNwa4d-`pc#=S~wtnh^^%cM|?n7;=&Y-d^R$gI2u(jxibmM%J#rjI0
zsTMyOQk`)~Re0no=KgYLw%3Bb&*II6;Dc{(Nko;mOoFjj;Ja@Yls}Ze8T)E+49-xj
zHTeb%2vy8W!x-=oOTZ`c{MAQo=2Di}A`*(vij7Nfvmy`xRQ$ctjE=>BTK?QHU)H(h
zWp%uZ;(y7e0FUwyxH>?ekfBSC0&&q3o0z5vP`tGUYq{mvOtu2_4!@FM#}G-4d>sB$
zM!`oMYz7$^Q&D8wNqo+@l8xK5$Vg8>RCkld;wL7?T@nY<+q^YU(ak1t1MUSJMgrup
zPZ~!|&Fb6tW<Di`BlZ!=eea<X&^5r(7Y*!#;DDhrTfI;iz0k27iKaqDA)L~$;Ubmx
zW}#v>i^Bq#flLtG*ja4k-26vz2OqN40WWWybtgJ~2+GO|+5rwM%P}%#U-0X^d$x6*
zxzPJRbl+u;LVb1NPa+pUow%0lLlsvR{R-d`jJn{v-&!#N0!mw7TaeH#v<mi0STt6c
zs@ID|gFT06?(Nl1=N|qBvIQz~EC%4`^o#Y}m;{;^eE7J7%+*>@s^?75zCdqY%q}6*
z9!^VjX(P$mggLI&K-gk!#U&Dwh}aA8pzk|q#u(S9d9<dxQbrPXPEs@V5Tb&a=nvQa
zvReqOQpUmy&~Dv`c74>icP-|T5qrvZX}Wd2kADs;dElrVSPIzF;P1(U4=4jSk%B(v
zro_#y?+RS<%~jxR#{uDLpJs{2h|2K%h!Hvl8m6(xfv&u|f7r?7g5j}bTZvoZNmw0$
z*_Zj?;jsF1D3su297K0=S<Q|))Z4F-(BAke0T=gJPmRZ)^zF)t`liUE4JG3Nb;SyS
zIXjM4YS=C?T(Td_N!H*kF$FO<GMqs%-d!s%cyEKC)gsjoXsS8jfwfF~Hib)0Bze5}
zX1)H`S;2tpx7oX-qU!B;#CTvg+r}<yqCmc$BkUvW{>M+{$>F)M`v8)nQvbn-QVE2l
z0ibzj*wRl)M80Mcbck`zOLQZtIA@LknQI=ajF`bBi2|_~9lL@!QcGFLR)SoHn40fP
zB1^Ulqi56DKB%WFive{4f_;};!|GMkTtU!YHF%#qn77L-kG9pkVHtBBQ3M&Q7qH`8
zu~1g62IUs2bp5io{b!?YcCk{Q7yi^oJs?YSy<XD`bp%Bxg<cL@WD9KLHOVlx3d=&Z
ze$W3)mF2QnpvOC(Rkib>N9!pYVb?hc>IA9AI1iD*Rf!@Jb$XTGztO`Q7QVmjWa3Gn
zAR5$+aS!^HCuM)=o<}@N%y^toz&U2BaY+el@Lr144IA@<nd#b!ErJFy-Y=K-BIqh_
z<#CeCvD((>`8dTsVAu;<Q<U{?iAuTjsjv6#lRX_hC%yE(A8U03Uge}`@IVlJ1_lfD
z#XS%#n=y?`CJmn@TjNvEFoC5~GiI4>+^IPMnf%L81j(lIZX-qJ983^)JIvio^Tp;N
zs4w{1*ayEHOOrVomL3*og7t=L5``{i9h`AW498NI1Rgk;YX5BTFk0rW)NS1F4`WXX
z{f~pfJigyy?@Og=<Vy2EZ_KW>3rkog>7;Ew#O1b3ZX;tn7;})`ah)F?mQLMF4!*ex
zmw`g%as8L!!ZA0m;p8w;4O*0j-)7kV&NcS(l(#Kxh?|8w^8m`Wr$F*vr@!s}rGo5d
zS~!B4#aZCRWfsva-dj^-BuSuHjH-`^#~vHA4Z2-3W1m7HYT*+s0;WK$zF%ot?>87S
z1xX)A&&d|xGD#KJjm)TvlB5tHl$xeRmG0b^sgIR>-RqvWB)-7Cz>ISMOq9uF(*WCB
z4<V-e3F_Z|9m=cJunw~xHYU#Q#p-GXxDdn`Y!5oqhmBMYgLG%|B|PU`Q@s^5dvsGu
z!j23QX9+e;x}HInYw>b|>3dn`8gYcop@}>J;@mejTt*Csc+aGS`EbptzkQ)%r2gC_
znuCn+!iyj4pTk}>IOpOgEk?qg!?Iwhfe5Wjau!R)^L+vh9mV!x(O0hV+Hk7oa7{&%
zbu4@tHqCfk#ez)aCK(*@3z~l$(YfEg@F*~K6xI+spMfgueS4R(>XvGdm+Zs?#X`Su
z&igaLddDWq@!SSi;_$P&X(=j`O8i}{h&=nF=uz3WwOT)yYJ*xWR1cE%xrm+F>Zuy*
z8%hGf-hs~WIwvqZTj4XC1YhK67TuO&S=|bj<nlXys<0@Br6{p^%kP#m`AgEv>N04g
z1MDp|qepwq#PE8P&(@8%q!@|N-GzV)B!v>M;<}qfC?)37Kq>&?j*{KOc$eneCa}Bc
z4%Xop7C*+IJnNf*2S_<&$-B^~LN5hz^yL29*ccM6L8RFckzUSn`1B^buTnrWAvt#G
z^^n_1xf(Q+q%1V%Y&eXKv~o;s;Yysm=S30#EmhrAJ(y*5A`u5u5c$meDJpZJ)U7Y`
zr{M${sZ8ivmvv?VRt|6wml_9rx0q-3U^VdSoP}|Ag3iM)WGsTmMU8nM65<Kdoq~vj
zSdPmi{_1&=Typ}0yX5jBG4BimMD0Tg*gCv4<6PK5*;CXC&ejz;t&@-*b{vzzdGecy
zkjLiV$dqbr+UU<S4a(xRs}WG(BqGY)FS;yq!nW@8xcd3w5>1HC-QC-0<-u3sv9O3o
zlr^(uOBXddE{;8C?wsr0Hm}ia8{Jy*p*wdu^3I_ja<FE8Mk#dN+aoGSK$h-N^|gW+
zT0=AzE%iLy?x`Mp?ZIx8*dY7DPhS01sU+d1wHkxa;x*6XkKU?ahyKM9M2viq9YRFN
z$X}INbk^RKdw@{6X$QN3BJ~3~&XkvW%X3h2DDlZ)ShN{-T(ACzO^<TWnpKWWMM!~=
zTsm|E>n7PHddh3p(!(*=!WWo%B__$Fwgwfl**>$k$=4}2Mgi=`!}jtE5(_yd+>#u4
zeh(pQI;&Z`z5t@5K&kYp0V<2>qrR#?nH}806VuSqtXKFYf=n3Jh_ZV5R7gjpX5s1k
zC;cJ`o3zn_`*#@ide{yXThy2N_Nim}TAbu^T5MmnCqL$;f>Iq+u-J$u2{R%I(<>P_
z3!_Sm+melo^^Nr<Q@5~?WeUR*cs-_CI11g~6F)ncN7JPKweFWug+H#}r_EQ~v3xRl
zda=FwLDs_~OJ1(y@0i{OfnP1OA08noueW&^H4b06D0tX?g;G@hWOL<{eaEYa(mICn
z9|hZf!ZzBg%7eyiv1vJxFl|DxV;u-G5C`;ljx`S2zKte`3{kF<`c&hq%Jk$6`dT>#
zb0>@!Q-}6Ktdr@gz-kW2_UxN$XN6FJff~V;??OjAc0};PAs112F{*Cf2F1$q!OLkM
zs2$7jgUMt;bg1BSU(di8vMtwJ(q@WdW-$Yfk~K1BAps#nj|6^V#DYz3sr+oozMU;m
z0d`FjU~T_cj-Ihv6MOt1)M8WcM=Ylx-3hS5a=S9|KEQpfAdO@rNTJ$Id?A(O=?Ckm
zqz<~VK<m3{wiy<QE{E*x*0Gh5LbK8AT$#sVhfb|WaA90?i41;eGkugdibj1B?6V%d
zCI}F&kh+IIfjR4c|EM)J{41Z9ts~D@1!e~9EIdmP>t?bs@^13>P<i=~I*t$N_JK3Y
zO~Z0K18GuirB`~;blCK{&TQyx?6tsj@fUIDOn&a`0(zY?B3B3LWs8V}(q4h-kMxw7
zl5+uzdn(|uwHs?AyWG`+=Z@hO21w&e=5MKZK0Wys!`z%!JhxXGpHTE`ngaIbk*93u
ziat%pO_FDLk2=Q?cpPU)iMn<I)rl|nKybQUU1piifHdH#Uu7%<Cu>g61+d+{B+I-W
zO0Uk2&%mxiEuVqtPaKZ};mk$&msAwUg_?1<U0$-j<g=~e@jdhw+fbjPL@dq4G>gBu
zV*v|qZQ2GZf>NrW>xllP--bU8oBb8}fj4fOn+k<3<nS%R4W;PbPzJ>QTfuazi|{K?
z_FJxLh1=H@KAdsNR-Tk%S%<nk{7?w@RvMFhRKs^A1&s=Ja+%zXqeF{lwt`O>+WbJh
zrT(F$+|3zAvvJIGzg|HKZ(JiTz4+9f)c3&BIOvjQdHH<(q}9nvZ1DUzi)4Jp`SY>+
zY$AVy_J=hmx4%1pe)B%dw1JHNV@<yNO9lts2hxc@4X#{WSFlw0vO85Bk1&icV4Xr5
zb*pqe3@?^l6y0d>4@=hlEMaA}$$v<Hh0xc^%G+s^;gv_pbQIgK*OTbJj5NaS=$Brl
zU(YoHzMtZo^b_xEmQOt(Ze%xqTv&EPpQlrnc9%o{N&V>c{6`TTaLZC8d0;Y>!1C^^
zEfm}=q=vP1d5gR^2W&D&!PxX50PaCTGmJef8kPZ5o_jjSfXz<VtBU5X+4WfzN%IKH
znp3`dz(G#Vw)1YsH*>W=_ud*IY{o>OiS<Gs#a3*g4emGhwNV47nw1$MY?5B>4hSGe
zHe-TA6Y@w^Nrr&;+o;O##QDf;Au&+{A3>7I`>Ks8bC}xgC5;6L<hB)S#gzh`e>QKk
zvT2xOIAeK(YZXM9VA4uNb2R~?*T@r7+f%4M-rEbL&(yy$^}&HP*5f!3tD{o6UgZ6j
zmf9;z?3f2V-FB_CPBfPa8@;)@nht(2#r%q7j=agACZA34VEgLtT-v{Ffcrgg&<ze%
znB%=OJ~tm0jTKnN*54unWiQKoX@sdqckYB0hP8NNu?#am!!(3@Fkos*W5YMwnt4?S
zp6WdqhIt}Au4a*`TlvD=D^y&Qe6pzJ*c$H%H0{Jb_ds)RtC>T-YL6P#CI5aXYGpD4
z#jjNW#Fx)dyp4}N>!p(K^?P5q75uPXYU`@l=mX01q%FoR^1f*N=UOzCGW^^5?*#m7
zP|sRNKq|FaTAI|sTiy$tkq2N`Fg3YFwjYaIM9fcX@%mu5lw#f=#(n^1?bj)o;@o%|
zTlB5Q=`+t~WXT=Q47LhxcA$uPpYdFN`0`}8=FUQf|1jCaIoO$x;@(Hu{66b{vXc;Q
ztOr0T7q<6aV4S$61sq&fSd0i^VfBI|qYxhyiC@6f<(;V~!1k<M^v9-$++xz{MF;it
zH@%;Kfn98p5<R0Eit=3v6hF9qg1J8gjF_D_X(;-?-WdVCUpo&K`Svo0Rf3)#T&u)^
z?=<4D#a_f{hGP`NAjiXgv?CqG+2R4fTON*igRG5>^O*Dp*OsvthgkV~LChJ|XAcQp
zQxxW6O5yv)kgH5&X=irJP!;ggOT@(z*_as+j#H?O1<I<*2fSB{lu#~jA1T<_fjH$?
z&5x0ngE%dgF3u|x<8p46OZ-f<u=@d6#mjPn<eCt4cm}FI2HzKc?Wz1K8q5&|OZjkC
z;Ttl9Y430f>HXn~!?EW=vjS|{iR=Okp}kJZMzHZz(>@o;8m1nGUot^^)C-*e^|On8
z`L^Sgp{}9-pSUZtD0c7{cB?ns;L7f$Vl^^05}H-NPxJaPfA`@GEt*<F9G;U%jgKLU
zsd;SfkE>oIx>aKb?5Tz^axt-Y;uyVG`FFlvHD$x0GQXOZXo4#SR$LZVmIwBS9^3IO
z+G)VTiTtn~ZN-M@fz`l1fnCT64U|`CvjO3#;cw<Hx5!_Sy@Cnf#R|*lT3icm0UJR5
zXzj*a%tdKSa;4SA6LL&GNT3zIqm5$s6G#UY@T2?)@)^@~Dc>%OSPd_LPv4X)xT6?g
z#T~m>7^!CkD!LoH+3q7_uFk9%nl+6>TYKIbYdD;SRY9j&WV8ls-tBa^v!gE3B|d-F
z9&}q!#tR!QFIR2W<S9GJ>g@a7<oe%e!7rTKAmqsf%JLX6i%e8ZlPJg*mr-e9rGV4+
zs*joiXpqab?n5O%4LkBE2pjV&NyY0`r~WK<H#NjDuE*U|D{>`^<YoMvI+m^VAgi#D
z|3|G=SLF|YbaXaE90-|wOnoW=oA!GCC-)$e5Af}u%4&Py?3N|48=vEUQjC*x+(K-i
z$Y=P&MMNKb!6KK3*Jd%Mh55sj(fg1>kw$&yd!B(HA$Qxo7vJ0(Gkj4uJlVJAnyYWz
zlYgMFE04<Jh1pB_FlI*0KK$vS(3+DNBJ#CsQKkQ>MzBmc2J?)a-4}nT64auo*3^Ox
z^Z0O)#V4*YpaQ(`T(5sx7$A6r4Y!7O(muzr@**t)V9HXexMLMQDSg7RzO7>~Ew)Wk
zN+UvfL#Us9i%eW<DvK?f9ZCnxq`Mu8k7e4+f(cwO-CL7DnGE{@(i}^+EVH;39yIQo
z#aM9!RZIUOD-%s8BZm)uv<Zt${ju5IeDY(1H=*cOPP0Wa9wDraI_pM|AkGQE0cKLv
zYnl@2OH%8@R%-YvoJgD}g&ig(G)?}MP4ss*;5m0-))%^v_WFm=7rs5fC-dvVl|p?*
zAaxZ3Iw@IzvBvoFx{8znQ(Bt6IA$AVu9hw}At9aA$Jii5E?*%Rpm0NRd=IhQzT8Oo
zBx(IApB?TUpgb%krdxq)!LRLDgB14A>sjVHO*yqx*Vew7ND=Svl4tg+KacC(|2V2!
zZV@$fqwQ%}xc;gYwdxRQD<O9J3RcfjPlM~4(A*lz@1*oXw#|d$o4-x5zR4@EPh7<|
z_Fe6*y_=Y2qOH~^s1Md3m#Q>nWW4Eu8n#9*-}9WXbtRd53iu5l4Q;x=pD8#!jR-)m
zd{TU>8Cu1nnMFdZ%f0kLA>2@a%q@?uYN)>m`{Tulf!cSJL~@dgK}gvTcu#ZW|H?Z3
z@j49BZKXvF{>c&i1+N7i3xM=#U|uHsmzL?TGxzV;Gk~ayTF5H<*Bkw-v-J1ty#R1B
zUVLF<xc`InL4jk<RzCjpCu00B09i(o8LTK6iLyxkDAxJwdty!l$3pQKf_?3OypO+s
zMS~Q&kpKVX`9H4E|HsSoO~kM4?}JC1JrMMwx2Xc#zuVIH7O;MBy5`{vFx(1Ij9Cp*
zma18!MOYq<xYC^f&zB-->qZWN#7x(gmF9imG3kB{etK}|*TJvzbR#H@^u!nV@~S`^
zU|;!tz=%Hqom3yR^Gi$n_dtDPce?b{Z)+mNg^BNU39jt`6aRf+j3svi8HURsiw83Z
z;%I-vr#%o(nGeEn*TLpUJ+DBO(CJKVpx!pWvQ_o9*ub+d7kHLdfrZvp(5k1@Jk6p?
z>_17_BNk@TZ$2-y#i8M&aS;b{&_YBN!^RR(Uv#mFhO(a2fcq#W7^Z8Q2l?GV*qpmD
zX!r`3<Wl&qXX4q?;dlz`aM}mYYAB|0$`{aK`5@mz7k4X0>~1{Q9ql6^_b~<>-~@mq
z<Mi8bPBQI~vxb@GLcX^LH?*sv{BQDcC?)PDa3^cGOq5#x7(d*aC3pM&N_`|z_|IEd
zh6H-RV5#c-?Y6bK83^2HlVXQ$9HCYJ$N~U7L#z8P_<H#sgP3dK0pR2W%jsv>%YlEH
zUE-lP@QIR0?tW{wDvtqaw2*?FYw+4f<$5|H8<3M@IEPe8e7>FdCd;v-ag^iFk#rHd
zQL!`z=6+diaAD1H6ib<;U^TEE20tZkcv{VXi|;2AI_Ky?2#T2KeQq<;U*^FjMo;sE
zc4FvxLQfDL`rmLbzg+b@^Dc47SzGISJ&|{6Dj+C1F`zS0zr6nXsMovTZ>RQ-3!6&D
z6ui5xXbCVB%KZS(`dQCEooAIXoAXntmfsI-JbE$`U`*<Ie&mz+q_A+@xpPEw=8jPp
znGASf4e|jpixRjM{+uknS>ncBDDD3C!5o^h5(RL<`I1S2bVBb)2617~O~&2@#(`P1
z=ueJn`MEkDfG+2exPK@pASL7~lt~0boM5o;+qVbv_*CrKejt*u+<ohctHl}cyV*Hx
z)uq~ei1!c43FWixYU%+DMB65ylS8Yj#8x8gxwcEeQ+iBSZg}Ko&&HiIQ~-#g=F9_j
za5KP}5P@vKz~D;W(r<FzHg2TfbaV~a>eK2p@y^~$(~}|&h9PnlF;&tfaCZ^pP5UvK
z9)e&D9J2*<EAR-<w8n`M(6kY$e`nx}hBBA=`9Dfz{-mDmf1|WBbyJl6_G>JM<F#3h
zC-o(xMix;*IE>|CZL_UWgo0gs@)73fU;Y4aHv<m6QfF=g1x{*qP<)mk@gY^(G44N>
zIkxzvr$0wVY0K(=&*@*Tu*#^lg6%}TqRB0V@=xWX*Qaa4Hubv5GSwuunrk$c9>Jxz
ze8wNfs`}`hQ*5ne$Ew-DFGUBmKa}U;V(EgM^T^h-y#jw?z~GtGk(fe#c8ySiM<W91
z(+HNUMSKsxKIiMc@uT!>$<j$n0g9d_qZZhHk>5W7V)m#z%o#G+$n<B|fkxaBTn)6u
z2754k9RTL!mT8_p3%{~N%pFVtT~_1<kZ(RbX^Yh@ipGn*#{lxwk9^7E0LwiJg@y}t
zetrf8m@EW?WD<7M_>0Y7?($S#h1P}0^5a{FYlTfD@eXlyZ9mT{z~NKy>*8c#mW`aF
z)gUE6wQz6Sj<?^ToI;=nuVMzf1J{Jku2422i}Ab;{5RDV8Mp=T6*1(4P)$oa5#8gj
zbD9Qno)N93zY}TxHt?%}meJ0CdTXIkkhj*_bAgUsl_-ihR^H+<EHjhe)1C-%<TLVI
zV)^Bx6VcZ?<#Aj$&gx7OxHgTdKP?Dd>&-VIGA@_&oMED~)AKI=REcxNoKR;+-g`UE
z#;bp?@IaTXGw(fdhj4Adil`99{Qk;|+Zx2NdBz(CRs<LHOvSiHKgZY)#llL3061o^
zs=o&Ta5CcX2goCkAdoyX4tQq9cp&}_Xu|+YECp=74kWgyJRSp-58-|oz6}`@{5<Ev
zi{}zCC>1FXrGf$7f;XL_47w*KeRfpIPdGxA*hr;LnK^(_u@<|*5%@OEz=bc{1DK6d
z#cFY9BXEeOUkfNEy;V6#ZMF%^KsU7E<b+mN>#-LAz2|*q(nXL6xvg4tz`mC)dA;@<
zIDkV*v-I{?*1;Z|kDn-E2{9M%bu^Ml)iLn4L5;oWX|M~t5lJg_T}6DYRF|8H7BI|*
zn;@%slxqhdd9OX~S<D6AoI;UXla);&#DZy)3Cbp&YaY2`&=JXHfCT_UiNnzrh$S{;
z>@rfRpTmRejQd2vwt3?C*lPzzF;4M{qOmlfPYKpcoF#EMW7aLV_yx_V7SdI4mEoK1
z6G*jG{SP@nhy&kr3As7$0h(9CHA%3g9e|?iT}xT%PUd?wd)kT6D{_{Zxh;#PxjXgf
z*7kPEDt@@!1Q|VfLw|R(W-PmJR76U<ck|tjo!e3cDH=;|Lij1=`PM^KNH{rvtgp+9
zyLWt7Q^Jz=m1)L8_>#TnI)y$Jq*Uo?35OGWv?Y>!|N7dEQvoUKO#W22ku^+zlLyh5
zeTCt@h!I}Do<g02WV>4vkJbQXJmJcy>(>l4C06ewp@Esv4Pz!?(Wexwx$xswm>>?j
zJ1DLUaT_#XH^p^m2HO~MD02b>a&-t>a5WU9n|nczi!i-gzvhwH=5!PJ<qcU0qWf5i
zb7BYME^CCn;GTv{yhn-`lOo0-C80oA26ZD+j$7_34tHY$9S)Z{^<-3$$0A5)c!Xnv
zKzu)lF@^JNRjH_>=WsaE(y8V6*EFN|;duO1GPS9nR3tH~5i;SQIpkKf?s7lcxYrU%
z3M2%ARSkp<G2dKA4FR=e4}~+>&9F}<qrM1yl)Q9d7{PHb&1o0_5cWoI)#vxCwQFHj
zs34Z-@0+GUjDQf~!~(SoQ;^YM$YK{NmIn)4T8-{YQR6BwV3i$5M9Ft1``>RW3v6@n
z5bm*M6o7Li1iWwK_*~Nwjs-RAdTLz5p0F(rm)tzu>26l%e1-DgPqhT2rDf%Zf(Df+
zoYj-r1{pG*WHcjc;IM&GCCa>#8Cxu&Pg0{kxHGAs$vLtcSK4ADp-$(r_fvYPh%^hi
zXOuZN5Ks5gs#!~SVQ`>tT{}+ICRs#kH}Rd`<HSuDtw$CjkD67@KFuGqrHq)2x)xmP
z&f!dWU=nEarb%S%?>&wTXe@JLoQFk9@g6N;wQ(Vcc19miLuQCHGlQAj1k6BfD;e@f
z7JLt4<S3WdPil6*-~AbN+pkX~d-U_H;<Jh{>V1f?UIx6qcwS*6!60SY(;~Q9_ZFhC
zM`SkgNOf-Ws>Exb*YmWENPqhF$|c52^wO28OBX5y_>bGrQuHSv0_%Cf<i&3&ob1z9
z0;*~9i1gLy&FU_JPJLC-cB)u+3)CMqOH)bvvifh*;1LxJT(gJe&fx?*cnKehoYrbc
zi>C8MU!#;ZP&)4oynt8yxUrgh@h`Q(RU&~Ke$+!VuorL(r{Z9>5%U;lJ^k*BgbTP`
zK`ETW&r2=L9~a-<9=-C|;pEqi^MtEi^(Tre+F0s}whA(g8A3R@Hor$cC9d$eW0NCD
za;5yV3^J?>QzM)Z%`DD7ts%p0xRNTHvnFqO99!pS9wAB11?U<-ZCz<;iXn5V*&tgr
zzOn1e5!wxOML)5v9kkSebeMQ<;^B72`Ps>s?sk|IP!IilQMET{v~ZB4b0|)6S*vdD
z<=y1ZSJz5ZQs%U`H$5B@XPX4$w6VhqU@s+>S5~;uW1TuD$v=PYRV*${dZG`=7-J~5
zQM-Gn*8*u<SMQlPfLzktyCTx&!VA2&>QTMuyV+745B<)kM<No_$^o=Cp?j3AzqB@A
zqlC*V?5QW}=GrQddY9*v{JQoU@BHhn;emx^d`JJeP!YfJoaWiOW6ClsD|Qg;WX05Q
z?T7z_?K77t$1m5-hUY({i=@v|d57O9^)|aW%c&`-NbugjTZ&U-T^>{hOIbg&e(S7m
zJs)R|ayDPGFaL0DT+(8!U_RD;X06?M@^1fiwX~P#M*c6$y;LN9A2UbRTc-zp<fX7o
zDLZUV*s2Zbz2Q^-!c85AyXO-DAK2Pq#Xp0V#4<+YU+DPP_7a}N3K)beDpc^1BWeMC
zBFtKuz^Qjt{<LN~h_Dglhudb3pyzg`Nr`b-SU!5sYkrP!L&q?IV#i72pDSQB5N{fz
zbln~k)$U~}qU6}Ta@pZd<B+Xczx?anm~(RHsLHTK`$O;Jrvu+lVvjdPx&?k+Sg+3~
zC+QVo8Ky|aX5iImjU^4@W7YOkrd_4Ji!9jYLp7^dR!xDd7y@_Vz{%Q3&@|T-G(Wrv
zlpD!bj+rj?!#%6_y8T*~SBb``38`d6)2UgLU~w#8M4w_k!}If~d7`}4w;zvth~2v}
zf4WJn<WS1>Nr>e+E4cn!g4pk$it|Xbrn;`rW2PK@$&kp&7g6mCb$8mIL`l9=D^-&l
zB<EF@`+w|xWmuHm-Y+8RASeujg!CXF(jp~8mo%c3AP5LZm&lL`4lqbJA|;Yil7iAH
z-Q6&>bf;&H&)M($Y@dDYkLT05E<d1nF~dFgz1I5wvONMnOzHF<=!7X;_J}AOj)ep}
zp~^n_jNX6M>i@|}!~sS&z_dXXVZs%Ye@Dj{Mmu5&f5v<=G%znY#<5$;TAeY}#@_~0
zaQoR8Q79rc@mRjI2~(6^!hK`PIQ$jt!pn3@v+EzOY4cE5jcwHL_U$PQl9wLpgJiP_
z%d17b(RE-<Av66APr|#+A@s>S4+#^I6mZIqo0w@mtAa+Q(_tW-Z@W)h*Z%&6pWP-F
zA4D)h??a(3#GPQX^%KYrAa;=a7@&jTA)b^$UdKvG7A`43irrq%%4p|3)6EenWq1S5
zA!&oQh$k%?@)M}(W=(^no!y2<q=sSciLnHq>_;r3+xAi&xJ0DYn))yGmMNER)wIrI
zB(?QabEBd1n{g(mwBU^X1;mn%VC4E$X@?!GfZ{8{&qyfwvFyKsUy^l+wT;56W{wF|
z*2$Vbza{nB#}Xiw53*^nBg6|ST%}uY5B+0?00L}9@~KP+lod$^xALI)!19H&r6f(t
zB*qlM$}<A3R)Nh1^4JqcaA+6-TK0`-czbd%j>5TWPDxR9vxnVQ3?e!8A~w#*Iwd8?
zAV^5F!fXmu*1^Q_a}Doi?%zYm7^>3t+g2SlnR=;y4H?RyGM@JR(Rv7<P7m6$RFrC$
z#W5j-iU>{p7<qDgH>oOg=mwV-!tY1XO#+Pdh}OU#Nnzb*km(QXf9VdFFnC}%1k+#_
z=^8ly(g;f)P_w@*Q&K1bW=y}HPjUimlCgVOosjp9M1OD~`e=<U0jdlaj_{A@ZNexi
zwHISG8|@2QueQD5p%2N+h49A=jQpA2y))f9pJ!)e4j<YDEKw^#&hlxJen>ziswLt>
zIOPQXB3tu#eUYA5xmv<5x$Wb27HFYBPeeZ01sXY2lWZLx#&qu#hh5=svly^NSVQmO
zzilRn+}wAl5IZ}y_Y$3G2FoiiArL}UeCQk7)&JM$_Al_L6&vlQrw&4r!7NA<i$fbr
zA^CjtVkq}Mma{4~>!7!#ic17zf+_6R=Kd*|pJ$c%Kir>mMddR-HbPa+F$ul;Im``$
zp4({1H(r8~&lg*oE!~;ZVz-Ig*a^Bs-I(0(w{1TAU7ci<kh9h}=px#qYg69%uDtvG
zq@M%C8(C)L)0%cLC!+KPWMGZ!>+YZg*nJ^E0v1-tLTtxQsRcNR1_}58FW=N8HpaMP
zFd}5fm&)cvY3Ee`=PA@Wc1DSIbO0HKnjS(xe#6#`fma<|iw0qgRW9pyTcZCaZ~AMN
z_{;=6N)|2VV%Yn1NgNL%AuDZENKwJj(((WzkU~<90G<KEoi7cHu!7#oOMQMIB#R0K
zQpY(ez;NR=G6KB><!m`w2Gvl)Ymwxq!@tms!4Pb`H9g^L5MorMmkcoi2}P;?S(pMW
z3v!pmkYQ>Oq}x1?VC|WbOreC+emui`&;vOneG{C=lb6&aTk*>~NsO{^h-4)fJm9Gc
zcxp-LrM&O!cWa$H_=X6<mC)+(yfOL%?;*yPDL%g7xs^O$jnWeGHQbJOhSc~573?vO
zN-F6v$T3llZOV>zLS7aF1;6(8x_Ktm%Pg#Thz{|O{z<bc^vS#rNwRJTrh^<y?a3#I
z=q+Y!KA5uAVtm9uCfX9PU}-9<oM68peg$6BAfx2{5)7f28yJaPSmZDOY<OnGEXiba
zTQbc}Kb^=c)YcS()tXrEJ4-9k4ueG=1|)2%A27_5A>5`FH@0R|e}c(BX)1720dfk$
z)#41ux!Qp&rSzKp^&Vnq;h?YCtq4izWz{@+qLWSwzfB~bT3J|qq25~-!&B62#{1vo
zY|2kE+7jHiTbQK{fQ)?xlp_~(pj%Ouwm?9<|6t=S{X4<??uVP=KZJ=)Tcn>!>q#do
z>2jO|#o`O7Gmw%L_O0!BZcf)4zBh@+2X*z_JwuMQ@K<y5?_zE@|5I*kf#1(mK7KQM
zLg)8ocmsLKj*B^b9!_1cxw-xUv<-Y-r=N62F+_Ah!-}ZwNiM7dnFAl!qq=CQpyCKv
zpUqO(8Fqb8z3NUcBiGtXMM1@funppgrC}hG@o2t5b)jRpCs3QtfW`hQWDawK!^FC0
zL*3fcnxzppLc>1K{v3(+)YQpkQ+Wn>#-n>=sin`n@BQ8!E8n<XKd|vch5C!L;-2Qt
zOa|F+aFQCTf2OtG&<BU@W&OFQ0iUoSIjM#po?61izhIw7fjF>zVhfPrx?u-^9J0O6
zc%y{J_3&<dfNig#-VR>)#{MwnBB3P9s<*Vi3jZ^j)bHPhUWR%tNlyUb5A2%emG#e(
z)54q6nU>g}*nI;(^lc=5H1P|_P?P0L091w@j@-DS+kER!po|8Kj9=%5taxJ~l9w#F
zr~$A#t{+M1TV$oxf%x2**ZKcQGX+m+CKH&uOH#8ytqyh_Qy_VEm&sI@<?9mAWfNod
zwjT<rFu`~J69TS*t-82$Z|(oq7l~Y8eozL^lArU%-TuT8v3L0;JlDr+nlgV<i~m|R
ze;<7x0sqKomak&Me<>TvM1kK_;YC)?f5G=Nup8#1Ycco#8+eZoe)Fhqb)El0C8%Hv
zk}_^`I)Uz&zrWGHf1ON%wL*MMEBq7r``a~3z!oH_@ae7mw>Qm%T@xe~P5#Rjx{obL
zO0X#F|8H+v0ZU4tqW`l`>Ms)H<sI;zg$L}iivG)+j>3Mv_#9EWlwQDKi)w0WiW^wG
z!FTT-3TzjRQ!6C_#MAD}nf&Ixslc1j5IkDRI+zdIe1xzb6`8T2y(gq0|3H+1a|~Vc
zI5&3@Gy}UK0e;P!{oeC<xgN|p6vz8(RM^pn8w@FC(Uu`mNg&+paLx9$uI3|PMbH#n
z%_-V80U=4l-QejyJ<0KT4e<LJP(WCs9BGm?I1KhRlkY^JW}&8ljlz2o&<D^wpwQR^
z4g`T+Ec6K`$C+*^qHEPh9k!oE8vraLs0Vm~8K8~62H;Q$@lc~2@+U7oegT0#eqhMc
zeCjMWwFP>3ij0BfaqWA*y+a<reK}^p97+I0|N3A>=Bi^aYBw610pLG1sLk=?2Hp&%
zo*9>|M)G=a2&bJy&Rv5|2n~Veoom9)pk&5lb&hmxn$(KA))ibRqO+3&B9?eN$)d#K
z#XVQh2XqBB3}TQOv5R-daoC=36?REB1i#2tJ>5sG_v@yTS)G1bX#IIdj(Uq{U+M-v
z8ukj{gpfJe2SkM*2o1eU-%AZpN13)k*Ex`9R1bm+jf`3W=w+a}2k=<SS4!9jYWWMh
z0Nw4};01eI%z#RPWM|~L^NsJDbp*<%AllMQ+hsng9-Mn%Sn^<wAM_Ez%I=5^g7lhR
zynrOn<Ca$m0IIH2;_V<61*L>;2q67u?xq75<RjUfy7{lsZ>O)%cj)F0047rG^k_{}
zY-ox#Bc23A58+!D0hH1Zct8kp<X&QG%fL%yF@Gu1ZBFq8MyWPCfNJ{z4u{81bZ4#;
zmw=rZpqrLdA!*T1?%bI;Bjfu1vv3pCBKZKf@Yj*ruLparBED`H#ok34qtnLSgoe^n
zolw;hkG2m7O8;Cl`s2VwQ`_hU(f=eMsYCd+Wxo6JDKM$jgC9tP%^cY2t)~Krjexag
z9>)}b><34aEIzjq-7vg(VH<dk%!a{OclA{S*RXp*%njx@%`AyY!=MYi-kX5)g<(D5
zqs$ILMq)kC4BrH+u)vu9dLL+Wp#yk71)J!L?&M8Bu~UZK^@*{`IWJ#34*<-^%F!~k
zf{T^ZmiuHro!*N<$~X2<$H0K@I*8l(0xrmh9}i-P-M5N80VMXbnH06dIs=k>v61R6
zN3Vc^?ER9;FTt<hzcbL!hIQ<rS<=QL%k+Vknqa4MEk}R%m<nSSqW|YAlk9|N5q*)A
zh(j`Xt{IGzWSTlAQFh=+^5n(sr%ZU?iYOQ&&}`t7&hpVBUlEvPabdylh%$lEBWM^!
zaAsM2<rYA-q!myXD0GP635;14eF2QzJkVt_RM0*=&wqWEtr74neLUGWMM{i8FG3|H
zFR!y#EvL9w!DTBVc+SKHsKb2#h&NbqOzW}R!|H%Q3K#)?Ku`ZP9!|6TCTRA|KP;WT
zl-0hH)~@Elg$>(^I^j>twrXb7e0zV^Ffo#3j&l^g`$Uh5ibq0xP`BU?^lFW=OSj9<
zAslOc86o&4Ag9(+FLmTDLp7Lh(tYAPLAd4x+X8^YXXsF_H0%&FX3!`4Qoyfz+-U|O
z1|n@c2jGJwCl$nRmieLsPM85#)2Ic?AGl1vZ?%Vi(Tr;&2JMPjg0u(}aYBpB`}ui=
z+k(vs!rb)DG(rSG$FlGRNJ^ruJ`fu;`G`hjz`|nm0;6s=zkAq)^?`7M+Q?BiqCEp3
zKl|yyq?+h9cxfQVLrc$I#xj5(%S_d{QB04tZ9Px$qvmz*6}pHYat6r5i-!9;i#I3z
zA>}`Xr`&f&Tr}B#R%u!l1%DrR8rc09?h(%?7E1a(v9fxc4x0c@Nbr<Az6%+~dk&BS
zrD2!F1doH1A;BTKgNu0)f^>FFLy*Zy45QP9{tY(0UwbVgnUPVBYjrVhT7xc4=Xz4w
zgC2o%bAshn=?Sk!-JN5t&}U`QLI>r+<t%o;9(<V68t!!30S+!yT8#d;97N?8cfj*>
zSPFQYJ$9??6@%lKEk0ryB7e=YS+_@*cy%81Kw*XnCg_AT&x7rVI3CNjXH*I{Sr2a8
z67M<KL^<X=Ea>?`exT;--|f(8A_~`Sqx5&S3jUt5_}A7v`cC{iB>5$67|9#~f`~lv
zuuwI)6L?l8sxqH^#NdZrB!Ww(X1q7OX<(hNRPhqDWPSayDyuKgOy@9z+8hQ`NXPz+
z;xFJrGHL_svcU+90GnmZ{D$OmQm2`m$u3qU#;3|{J9JZC<ZO=a7X&*9&3_6IV2HN@
zlEDkAOlBmus$6q4pd^*k`|wr6ySZRtPWFNZlB!mH{0X@S3|5Cf7(?0#r1c?NkS22m
zJbj?{x(+^JAADy0M0>9}FqXuzmbsSivKd4}^`MCN&++o70Z-B6@CF{t<D!%%>j8#2
z_cv?cew>)4s1Oalz=N(bF7CkGM|EZN@l5l)n|{0KfPj*nOn?uT-`Ac~7e&FWJwyF~
z3YTq>aq?ZkK%BLnb9o!MGvYjcciUudb~7nUuu8Qr=wV0EI^d8ZD?2X?q<eh%)?|Ib
z4>_S~zzclXKbDF&cHn99A-4XikO!6C#WqkmUpwWld+SEZO8!o#e28<lA1s`<GjrpM
zMxQ*Guhsv^YP%(66G+tT_aGpnvPc3`Zy8OeY{W>@VsO4q4*NB`2Z%O5mq&fqVH6C<
zhQONij3B=gDHdIxUXWv-z3-Y<1=dd%LORiLb-3LIqlv85=^M$(;oYgLO=Ps#fZ%gm
z;*&6<W3UPH^K`9B)Bt;f`_t3WMC!R$mU{x3y>3ywbRg{d?DmD%c&`PdP1fhp%>^2+
z9nB~IcK;*`dHwP{HxV@aacL%N@fVG!c_x8KdgAqsyzXIc)_Y3KT^`#<>_-)on{8b!
zZK>mnB~6%}?;?jkHDgXgXdl}qEw6+R^hcby$5bd<v~ujcCra=>TXCJ{4r4FM&GU<z
z75P9FVJ#Bw6lwb+Hh!$Rw7@o`yFQ*Hd}DgjQ{7jzm&{39ME?CZ43BlbzG0fliC;wU
zun0T1fn#%NYO)fZg=F#so*FY}?GJo`dP3geyX044{$H+j(K?{F)2@E>=IX@+Ht0fY
zyEto}=L1PrX9IzBAb31ms`_X7d?8*CAHmx8b_xS(z)`Efw4v4Rxaqqw&)M{*vp_eT
z*A)=#YZfL$X0AyRc3Fa4B0aQ+&f(&$P&r`mKL`u}wZ7Rboe@>h)47MBu1A>cbKPKs
z+(09Fh4Oo+o4&hV{^`WZ*uLxcF~)k(!~m;L>_#A{l3T7{2GgMX6Sx9BDgy8s`xSLi
zOMtzCJ%y4Glg@ZetL5=v8}5w6^=ch0!Dz#i&nXRME@G8CRsLEs_{4JLsHS}bVgPAD
zPorxK(sq93g0md+8;7a}5N(C;QBOhM#JB(1q>lNgbM;9=CERl(8pAPvBP^nH>!lBe
z$yU?cscgf$5Tc%}XscY_8=@M0eir*#UvyvDY6F^~LzlYedQa#oMHT1LLzX#`5aqPB
z)V3J6AUm4Sm|ftA{vvvJ1bU|U>rH50b+}~=9Z%UhaSpN6*6zyypI4nl+BhD-4zZ?p
z%j}MRI0dcS_`znR9zYY;YD>_IKS6G~WqFWXAcdir3X-a#?xJCm8`EUUsa<VZ<Xs}i
zjND=kx5}nHyXs@0$fyS=52{RCP^Q6OORJv&4iv9MS?nrQT`5(-=<2#XlU;kh01laV
z7s`WfcM8Rrtab3LOL3fE)^N*Dp&PF*o0{`)bwVHTioxO1bYSlU#$TJ;F6Yh7yiCzT
zzsjqf1?FsHsfO;Ae=&ipy?gU(NhUS?mz|yko%O~f-mKfThls6{b!AyOgFql0Yg{`0
zW?jbxt6vbNyD55%d3_Wm;?_Z&u0R{)Uo(PE7BrVEyv9>t7Q5S5_S~ay>KCSn(}WvF
z9kGF9DtL+MNjiRNRbI;sLxx3RL5sUv1Z2h5Q5Px^aX-Quox01UW!Aa;qxII9+iv`Q
z)}dS8y<MrQ*CHaiZng01CXwc>@@P;T^c~twF{=8-Od+|(osO~$eV09hOh4rvPS!2P
z9afg4W!OmJ$YrQcRTI>{aJ&#nG>#0P4)PC+=?UV}boZBQZ}@<(DjvvHik0Pi2(eg^
z*m9D82A;yrNycBaP2`THjQ!AA`6~IUVn)1a7qlNM-E42fy^&56#8x8_kpf<QF^<A(
zVx!B16rn;;l9^1bD=rqzWP>ExPNT79u2Lg|i63o)?D33|Jag5YAbtQJX`(do)U=!7
zTmq1aM^MA3(xYHb*s-3Ipd?5}t3X~VZAvDH77z#Vaw6r!+@L&{d-C<_VI@^nq7{IR
zVfKXYPT!z9C>j5DyZQXL=rs`a6swCZ6`pL0rvjskw@Y-o$b~I_5%3VVyG4psonN2}
zyUKO{xc-V@gEL5asS^_-ho(pCImqi`gNoiyqlV!6X`WEH3M`Mnxb6Y;gJ4EMsgY$=
z%+X_Iy3Zqi>vo=wL=XI=$XD;cNr|LA@;qo!qwsy6JiQHb)kUMdsG{6)OBCImvJg9^
z)ZJ%46(_Vlm%lo)e>=BuGvJaMB2h>T`m)TqY37%?S0Wukkc+jAhzLkw-Dod~4cpl^
zWQPEjwSXbkJ_qd$CK_YY(eN@v8~S<tfu5}=S4bN;5HvZ_d&0$=+YKAH>_y<+_G0YM
z11}|RyFZ(~^8U~-8;oF2L{0^7=(;%uj_uD6o?c4{wpiDg<exjv5If5nnSz{8ZqE_H
zMHX+-?$`|I?Nr(7R^FY|o0}LZGur~-O*!7-=^dDt0{)WzBUYZMq000_q0PE%TRxZ=
ztBBCnt~8F*77vSO3JbH*+~Y@nEif2EZw23Kcnk(gEa75Zh!Mt}HI67Uvl*F`tILSl
z8+x8}AN6R=|C^G8vVNA|F|12Xy1PZFmQM$&^LbL5Q0DMnNKc}C!Y)fzNZ0r<xq*lA
zWS(VS>`A9J+ap!?(g>V5@7|!v?X%)1bL@HkTbBa3hPN(qiB&zmPIJFs6gjs>{@Re`
zCykVv;%mbu#|W+x3jHJ-$YYAchXo8@=r^Iq{)ak>Qr&f%cY}hs=Fr6GH|Aj-Z49aQ
z)F`8-*$i7}L&DmRu?#6;uSqtV4%@0X-d?$t7_T%O$;T)%_Vjeuqr!Uow6Q5?km4-3
zQFU;pu{GAno}wW4_BQG(CGB@~5~uHshOdC8CE<RcFs-5J+&!U{7erm^2%C|0fybkg
z^z+mV?Q!R2P_hu7ERv?-Zw%7*>kOr3-SU7Oblb3=eUrblZRhy$9_?7^vsLTplPAA#
zwR8H(T2>Da@>RCULL@ok<1NO^V%(cvH!s|jNPJA1`s~5f#diZbouvwSVGb)hcL(&o
zubZ@2*+3lic+tvwQmA@Hn{P)>8z=B3#@U^@E6vgTdDj{qS0onY)q^U%5N6LpgP*nW
zdCgNNC77X)`5^|`<@$4HM+Jnc8^fEM`_*in^pmt)^+fp(8>X;Td#@d1%SElhIyxb`
zp}Ql1KFcoGDxkaUSk1C9h#7QQay#CT;~T0N%Km$+vzGyWJHKL^KLQF@zN>^+KkhuP
z{VXshi{(tzZ6KdRvk*o6?#NPBx1-eUA?{j4VbzBy^YeRS?HnQBUGOVmUMrCP_KT5R
zi{@RF-x|oRm{Bu|t7N)XVMMwz1qX;|p9+YFNDA#N3=T`7(N9F)pJnh7pdd1h4`JcY
zx-Pe`-yiVI*NoL58S21;kldnIdL0&QDZ#Du+E9+T-Z-q=`e4+-P+(F(pP2)v>6WN@
zLC_vr8ri&Ma!-*$!Oy2Cgy6=qfocT@g7IiOiJzRiXjDFCUS#$V&`9llBpVvrPIq~C
zfD_x6phPZN+Px;5{NtsM{gh?{G5XC%GF6hayZ)Qc8x*A#D?EuiC<G8y2<k*`WD~5X
zuM%IUsUFs)T=6RF3KvUzbBm1eAP*U0oI>i5w-o5VFK8i3OH5a-wYGoX4H@C0X>AVc
zGcC<;d8ALDD^W5!=s2v=i^8J2v<K_%l*`^R?;6o?SW}O7V`&ch(Q<c@v57y=d0i~t
z%yXjeI_*lL?a0`?AGD#D9mG-OKc&(Wuoo0DDMCIUm|5LCZ#b=*UG2*s(975Nqx)Wc
zprsCdmdeU%db(CKUL^c9x>^WxIVhJyaC=UNUEj@ZJgLxA$+BaLe#MuH=&QrJhJJKu
zM}3Xfy-J5fd!h|OeV@~(Th+Q*^-I0BM;D@(E(JOU+$Q{u5}Qe;Xgx>^=j+kQIxDF>
z!Fo4Of?PO)wGZO9Z*L!&<V)!KD&}`~Sk~!3+Eb{QBFR^5=#rIefz1`&G@QQ{flNXN
zoPPcCO1kBNc|#xX{{HSf^F~{$`e#cU`S<Y$W;+8%&I^=r#b&`44|U})R7kEl-dcM$
znBO8l74H5~e)xMy^SQs=?fFf@v-*u8e9PQY_dNIA95-dWPan8UmksCPS?_Ta9#At(
z6fiK~*gln{o>iov`$5ogC-`<#t^M1EY$8=sJk~mjEzw(CR77~3qwo#tu>fUR(u4eP
z6G>udJ28!3JmYBsmu4CsMQ}u@y<`W}R_}K5nZuQ?M%jiy$;)(vwSv|<2Ir@5t$)(^
z3$|S`2;AH@l0Ezs!0Xe>@D8>ROam&27qVLJ(t0A#!&QS3@6ky}+uC}^67@Xw`V;gO
zI2*$aplvlN1DKm~&8?0%0X?W%$H{3!KLLhC@;atOO%K>Z^bVQ5|NUuch#+lo0a|=&
zvz6%z?JVg@vyp?Vm1Re%I)7w-IuUo~ywF4Q>T)^mmTN?Tmb4Ny&`8I~f_9P@cdRqU
zEmM14v&&zd=}95wDlipN-OUfyYPKA>BuQm)oB(w$#=0oR&7cnD*T+I{MPwh&JbR6#
zyl$Ld)MmYKZ?bZHb>M)3U93%ScYfpQYr0dv;Fmh&;skqP)k5Z$1i?z2GKn+A_OI`Q
zLYv;~H4n@4%MT@(4nCJ%)ET-o2AvTOu;)^l3K3SE((UWgjyJ71WB@4}X^_LIN1`W7
z@vJh6V+^6^KeKma7iEM1#lY&Pp4T1<yH22MQ)q+yX~5;T1#|qkgxQytz6T}kxtl}~
z=|NJ!GZI`D`{12CDWc9u#j?GUd0uDJVUTd+A<O&IOx6dn8j%Hsnzs)kNfk{+jT8%m
z3L^~-G>5;B!^grsBC<l+Az~v;%ch(NIy3TB8YdQ_(VqjBabFf%75(@$+lBZs7AJc;
z2XO{EFS1;5`A#ms#Q|QZsY16|gGAyh<#uPx6k0X>w6U*UK2qzWP<LEu6{K;h++pgf
zlj7UL;UDs&We@MWC_rgqdv0zK8D6N(iPgH)>^r<zu6X3F{1o+UwxkuMrlj_`HD1eo
zCgWg>`cCJa$xh{T^*O=H35A2bZH!<2#Y?jGi4P?kTFhIuNz&WeYblI0qaibUM{Kj?
zu)NS;7YcaWh%>fGFw~NGCMl%k;?M$fztUy8vj)9s>E}l66voG$M?MEEPTyuU-em^(
z!Q(R{cucJDCu9#fNW*mrN+&}s46>J{_)Z@8im7YU_lL4ksMfOUV(DOKCWdijFVA&5
zE|UU^P>s?!%UjuryZ5Vo?^M2m@<REyT<0q=)bq_W(o!32b;7U@_BO@^I4fu3^z+~D
zm@bbuMdt1j1pZ2%^#~ZVT-&=j6jjpBRY&0woZs_OXG-%0yL(RW$yQY~HNjpVrD=)5
z8d!!bIx6p2B8rYPoc<2ZTT49~xDkCFfi<Y)GOF=jDDJhHE_kcvcZGpAOh=y?IbY83
zrO6j5Gbu;g@AbeAMDh55LZSQP00<zvP7`Chc6U+@!i7xQGU=775Adn(V~)UQNM(3A
zTherkrhsAANED`D!J+?(=XQcE2j^Qm)Q#<LL2YDDm`_PU^bqHoPt4QEB1CVD=6Abp
zuNk8Hz8y}RKyzaU67>@FJ1;44JzBbPUCG8LR6O=2LkQRY!RGKiVTAMhH&>x8m()<I
zgth+m#vcf9%<b_nkZ(me*3=(ZFQ&8Db~Tp{TFsTVKFtHqZytDlbMjBo5)`~*FX{Ek
z@1kryG1E^_MZm#*qoN3pQ&nCqv$OPx!CHx8l@1}7i39icr#38Uy$Vrm3V>cnI$;sR
zwGWYWcm71!R1&t|d2qMHeb24T&AZ8IXchnphqjV>)%h$M@`FU;`h%4RXZp*lXMC0(
zx^~Y~k`;1Ct0jsbEmwZ-t;(wS6DxlC5;yAmrNJfiVfYgR1KScHYL|Lna_owdYv8TI
zkkEEmEdwon$HrUnw|)^_DeH#8q;^|(uJ3eiXwTqICQ_s@B_^Y}F?1ZJn9X`(-riZ3
zRS~lUw{L}<ZR4MAgh(Fj2Da6{<+f(ePAId=GC9CF+4Dc8^>XP>mRJ~xn~e4h`Fhj$
z3dQWpQ81W1s|(8SB2jhMJyocZY@*pZ0Xo@WffDWg?ipmd(baM->M>|Blmhu|9seE9
z%Eu$l2L%cEk%G*GL>0wl0o*Gei?)Ub!*luc8z1p`WSS=3`DU6}^wSyi2$>~v=>rfu
zHFa8a*0_;B2^zd+zO$u}Z@YRhp#L2yACVMYOt~&WOUKq0uM3hM-_IK=5$uYZqMn#N
z)j}%Y6`}&^E6JsU?sdiN`z2-i%XC8)(Pd`<`O-dkI$bP4smdAYnBV>_q$xt>Xnn1t
zXRO0DcfnSz<luA8{-2q%;&}JjyjYpv;0Ee66#!AlScB|-`m#Tz4-h9jar28|@ok2c
zul%s5>>{+KieMqf!54ewXn6JG8~2D@o}|WvmYB^7_^of+Kj7BoW?(u#4iGlXMwuEz
zTp9^&Why>wMpBEqBSraePSiGl3y{5<F~xyBoHV<ro1c#S@cWxpldVvA@dx*q#_cdy
zmpWNixx+0=v-lpRm(u}4N9#si6~9iZe-U6Kz&7+ZZ-h*P46e>nUh@`^e>zzwOz3q*
zQ|SdNjFF$e6IM57PL#KPCQpfo+&Y~_NWXbv`rgD!5@#xDr&Xk*cNSWAmG&d815JoE
z$B3)J9go~}qJf`M)ej!wOo7Oi%p9BWQEz{8!7k!x{0piC8W>X%P#f0X7r_(A613&-
z_G9BF6|O+}lMGbJ!sy;<oo97G*Td@07LLHQOKN!{o`f^aVIUE+bozk;5n@yfgonB7
z&kb(Q_}vMlOCDbxS<P<TQ+SFD4<8cXuXqC}f94YjR6l-W;K^LLBB%L~a5LR6ZrMD|
zUmsvYJ(JXPrG0G!r`JC3>qpsqA3yNPb7=~8^ZNCLa;P~($!bn7Fn3tgUI-f%`M<s^
zEDg+x`-mMcF_?jV+vmEiLPcmS=0cA@SBks5<goE}@HPJ7n&7uzv{510d?g^EZr!hE
zBl5!QqHlPJVv|gn*d)1i59UPx^;<)wXaT{+35klb{sXJ&8m~#ETCHGzt^riOLm-+I
zoel{=lEW?FUl;}%9+zFw{U(Xf2|;W>Xp52`=7$s020^Yts{6eT9fA0$<i0J7!6B9*
zY0J9$y5b&wZliA0<BLdv+4oIWYiWU8Q=nd@7~gx>5u0rK@G>o09nu#l<jvrXX9bmb
zkH^fhlsX2dW11bp7A+B{xe$L2`r-AJweAk?mBkvMREQz+IdG-8f<OxW`l78jD}o#0
zW}kjiXFn6v81qV6vnk?z8s!xVdw1j~hR%jMH;$2BIe}wtqXx+9vDqlcOBZft|GpP=
zF>^kbj2i)jq^=t}`o{5cPo!3-2g3h|$ID4Hn`h1uP$+!a;iOlwPe%=7XiF|J4kc{R
zK3`nT+San8Bmn1KO;>}c-juzk9=Uk=6?$&?&$J*qip@Bd8_hN?c=kw<h((fG{OIvV
z<@*nMdY%}2!`ES(vOnI{Z)JTi8Dev?n_#eUOnPSZlzvX`@;PVZHqN{pL0#>ZyUe}$
zV_+9uCr8rkID7fm^$sRd_gOb@WFJvIUaP!g^EG^4atI-Y8@`U>ncgt6I-M|Lmv@_?
z{c6Qks#;MU{SP{d)Qk1bn_i)+;z>Z8ku|x1?lbMzN8Z5#P+nkg?~rN6+5^bpiD(Lg
z?L&vIn4_0fYNnLfXIxEE;%L?&2uDE}t?z!s+{K)t#>#jaRSJHB9*kH7lOPITxa&n^
zkO!55vc7ZShRVJ>fKBQ}9(*H~`Ed<hGi{iecC2A>0{kEytI&tnyt-v18>A5q(1-x#
zbJC%ZFAril>n<i1t#A2j``5hv@vf|q+{{!?R$Ss4M>C{RPFt#7{H0;wQE%WwF>i>$
zXOUgLr!0$sbh<x<l!Wk@8%)_b-V|P!xCxEgA!CTVVz4fx--K3s#_VEb#L28EX;o*@
zGI&QRk=S)k&COsVf7$vsSYL6bpf4}=CbOfwxul72J<$;DulFdBYVnzC?@dqk89x}F
zq2#2CT_SwRF)__2M&WoMt`Mx(Xo=0~IC+BI@wa+*(&J7cn6jjwA$}r4EKm_}IygQx
z^>mkjutzm9R!dP51QHB!>Gzmlh@8-Ms|>e|p;J1%5Li*E_vnj>`-E`eV}Y5by%r+e
zcmqVH--*y`u}w)8fXpa;rK`^01xUQ6TmHMTKO=O37{wIV$=|Mb_~CUNF6wRf?M6zX
z8Y9`Sl?eR^q#Q#qDQ^Q09t>5m&yROII7pQ7#bdS_*Xk`~-6JylW)__->k>~B$y?I5
zD>(4Y<gD@rH)(;#8E3sOLp8vHIDeC3S1+F9YNWP=!HoB9Jr%U4|4Qp<U^B&JGeiw}
z2Tl@WS@K-h=EJhX@2l9m7JZWF=v_>X^o(}2-E^gdAy}YFMGX>MLzDSyE^AHA3)EZp
z7DqSb17`2$K3zEXH4_|qb^n@#=ES^yekzW5zv<oE72f4Z@ZxYXrn9$Nn!-MVVHOvN
z){Zf9k%Cc@AGdRCiGmwPZ97T_$FraG<?Scj#Va!lx46t$@_++Tnb97WIk10A{Fruq
z{xm2Pv!mp;|Ipfr9oKzuc0;IXs#mhdiY9w#r6ZC#*{vgz)~j=<YMn2^w*9_Qh;Ss%
zDD*;>>R-pgd036`4p#dhmle9}V5v-yJingd0Pjw~N{u={8Jra;MrJ00P|!yy(#|wT
zgEU{)`EuumXbuJ|(}jaw*Fwt6+MchCJqr5k$32a$0q(YE<<WC?C!FqO_wW5f7UQLY
zO$}qAxaIT^t39|+a6$cf=t*=uHag43Y`OH9^-oa}yxRx+X6oK%OJPNyUaS&t@cm<@
z1My=m43*ra6MUF?vt_70gT%VO3u?gE$*Ux>vqd$|PmAy0{jDXe!s`@TjE|#16t(UA
z!v>GtR`~ER7oJ?>TVwpkd-xw2PB$wM#;lRnr>gu1dD6eOsmCDa<pp1Vamc@KTmSkt
zb}tjpz?u>t5k}nq=P&#}FQ#@4D0K$Z`b7VJhX3R11aGV=h(~E9^}oCrA66A4rF`;l
zyXfEF4tT(>V%0p2+*uKSyujc8I)GL4{J+0IC5+ryeUM=tYE&aL?i~s7Y6`OW_x<z7
zXFeKV^VoXQ`zEp1t6O#eS@!`Y=JZ%~v?18r*M=SDm9ap=4B%X^gDOzjSdaY?w)|vo
z5q&VG*?}?#Ur<je1k%Bn@A32df_sIR{o7K?Ovvj~<E5$U*OyA10#g7x6XWz8=vfT`
zCCxRGg_Zxr|4UmF&tg+Tdu4d?KdF02K)x?InLcs`xbcKhLJP6=ABHT2VP!xaQx9y2
zbgp|VH$e3&iD{g9dWs0P>9+2MmL7CvWu*?ZK+=(EKl4hNA2Xf<c~oRrnE^KlA@RZb
z7e(UYjjZk5^FhVvv@6#9lA$fu|K;EB2JuB^aJxB%B)KNRH-X}W1uO5F0TEF0*2_rS
zDKOq`0$t=K#NDyhrXVUaco(+K&K7qDE2^1k<EkbE+5rvm=@Ivxj(V`ADPvm5e@+tw
z@`9iSOrh($;dW7O|J>*~@rjIhhj%<_*95!x$E*gK;rBv-myJ6fXmJwMUj~;u4bEeY
zmRQ%-t2Ktav!kxpoXFeY7%4?@|7GTcYAGoDz0Tm-tJBHbIq!h_<$YyHo*&EozsdpT
zH)EFNf^UC3F5OV<;(-2?BI-ndt7}yta|q;8qe>02g7!u=-?y*@fkX{J?8RtR*urcG
zZVA3B^ldEbOCjX&97{yDfB2dF_WBCh&$+ib^Mcb4(aCy$Q_1)xinu*-MDKv|Uvu$5
z-yKlZPzUr9bj~tuYvRv<u!SGXvzV5FY|J|FOOa#sF744<oD0pQP3GXfYh(Wu_4fzp
zudiv6-Z;E@b;_eo^!>Ki#p{CfSj`uclwG#SDM(}<ZcyF-1=?<JmVz|Qio=#~l%}x^
zKfkT!%>W_uv~?LE=IVfC%BXE(IZJa+Jg*BYU|W!J!w&0Fnon4{JnJI5mqGglD?1Q1
zE$p2veF}NS(oVCY{rcC*)YxG+D*_l?RmQMRNFdh|rK-nv5J)ccnfz0RE}iU19D3!&
zA_dD{Ar0i-y?e1Z4<=>nKDP{^7h~KuUC4hhwd_xwE(mZ)hhc#N;?F|B_zW}=@{oP-
zl-VRkgTJbRQzZZEld&n^taJ-c)>)^0FlSy~sW-kBBDv2xv~+PYVHmh0>)x^DY1?Tb
z=z)}G0gM!MzChz7_7kY&I9hv&j!%GcVx#~lzucLZltuRXJ6b)#pt-;!r}Bt`*i^Mj
z67h6Kp0?dc(?+;)Wz(9Ph!88nOIb4wx_G)o;6i6;5DZli^E^cguS*Jv;`?K3g*nJq
zL1YUI8&FNA5kWEo1Z`iprpG`1(piKEd+zqgk(qY6D6Bu6bZp;!f10;duE^O7ZiSZ4
z5vFMEy3TB}Ac5TJ!}9CI4MJU%!rQTAyvPC~<H-zFI%83wGb|l$?1EB+y@shiKUVrB
zD27b!$}S7HI40ZJXamE*%VD5Y%6Yo&eN1%0ica`d4*n<h9mZOqn`$y5$bOz9NXLV+
z>wG&*+yy?jU!xvx3T!Od%yKB?ntXX~1xZEon4+6KE1OL|Q?jhU#n?)L-%WSSKH<rq
zmFQj;`1iD)e$JIS@fT-dc_ZlMOk4NJ3hU~6ebi>4gxcOy60O4%z{WxQ>r^FbAvoTD
zc^mqq0>{{x1nzMCt^|ys8(^6s{#PJT=y)6KU8EeUw0QX`R?Y_uOS*%2_kW!(igmPw
zQhSN&gbpR`DX1tgKb9wOM$;?x?qb_RhOZe0o=^8I$v)TuiN74DR9k_10R><&u!v@K
zj5Bv8EO7o;7C^ec)`u6G^1>ajR4OGJ8c%N6H(itXLhP4gkNjEC%<g>W%pH`vRK%s^
z6L6L#;|F7cp8vvnI*U$$2Zav68r|(wLj<!D3d2sLFmkUQ@RSokVhVbk5iW+oZ~Vm9
zd~rm;R3!XW&~NL-%egX0kKAeSHN)4tOFjI{s`0uvYOfa%^Kf$vGQNE3(84qW$aMjU
zsa0b-W_L-okmj*2s6!{5qR+WOR{&3(Ia~5DC^xRE@gyUreBlAK)}`5{HPu{^qa^ai
zTPiq1MUk|{RgL?aoAv8JNlm6d1sRG@(t89HoRv|Q0T~KSE8s>iEok``(<9(nF#==~
z*6MssNjulFP2-70%bgtzf;QX~gr7aw`<q%DD8!Tc#Lo6(Hv3TWAMG8sU&NB$3?I7h
z8i^QLJj4h~11m0OX9n|J<LzEa0mH#jRRPbz`Ow7&gF7SDS>E?2nSYJXZd}T#XYi?;
zV+kZOeobx3WTsR#ry@|lJ6LM{^^T*{cH+trr{q(o7RzL=dyy1l<0~t!xAyuxJ56If
zzjYsc-6^%x+}w6uI|WCJSLZza(Bg2MNi#?|k`>!RzdzmN?>C7RFi0|aLuj3bL1w(U
zK|l1;)`H`lB4RR_0)E>7V_fj#Qs&gzx^tCczz|k)f=7%B#Y#>Z*boFz#upozpMtM~
zE!idDtd|YcPig3C7#K|D5G^~Li?*#27CZ-b4&D$=X%#$e#7XQAKjT{CAf19OS`*o|
zVd9=Du*KCh@PG)9X#ZPSasUZ7l6tU;SkD&3*eVO_`&DBd?tgi`O+ol1K^uEfpuaa-
z0WxlF4hd^lKLaUYPjn&A(aB$SDzb~agrG^}lPndylm=FylJ**K;(R;nkyY{l+R}QU
zTB5`1S4uV)8WlshJ@Kp>?pKIH$C^RIdWtzn)2IhKA@M5^TDg<?sR~$F7~%mj&z}yA
zO8L&U4W6ZQdw$54qqh_R54x0=PSt8|soJ<_PE$|E*ANgd#4AEvK}M}{8HkFj1BwL0
z6r1(>fDOuR{btNG3JP<8I;gmTWXo`d{bw=_xv93*rH_&3MIHPEA-9Bd$_ItGT6&Iw
z;7dnEuhU$5@ky}8Ok(j_v_i8*{WSqXWYcU|OG*4b?mYD@H8z}`muN(x+27M%#oo5&
zBu4)8;f$|i;(mb3MJ>soSKo^4PbU&2oN0e41W$Cq5_qH3Rwqr);_a_4A5`h*!--*=
zJJJVnu?o$)=dZ?7H#4Zj2shlfiadf}#O^8r#mB)+%tkdKtl1+1k^+m^yT<IPa_nMN
zxc>e5iL>E0N;>mi!M(2>w^Ek#jT3c;V*8_VtnR;%R*Z%!dL-;*Dq^~%)1-eWRec)T
z`ysCl$<tmJmMs+FIrh#ppARpTpBZ=P%F4|jq+Nv-0sWLOZ3{n4-jS^yentkKz1K!d
zyj|3qDylRg;>T_#Od^rL<>_k<cx64aJ^qI+w<OTgHprDNA8I1LR(fwE%bWsPL7<jU
zqNpKgt=EBm+4QU%yVk{zqN@|+k^T(s&Jk1^kVE$lnuzd73HmxOgTtR$8`=USkg206
zc(b4kEscbRxY_ynUxzFbLBu^LOl->vbNDn+GM-WTH4vu~qIel;8628?pLzQ+c*jzC
z(*|ad*IIrG;xAOrHE&*BxN+fT?bVs12AC>gU=r1&NAVdr*BpIL$)_gr2FV9Kbv?4u
znzU#E<nfkx&GJ-rbr6g3`~BJH&c~kM#G|8(&Fc;-(IQ}B{w#S2+;-ck!=J+Yudc5x
zGc01t#!Ww1E1?@Awt(TMHRJg(IOBDU5!;%f6v?-sR!|D4l6im@aCuD*?=#r4N+OH|
zeGipsAp2+y&hXk0k4?4t3ct5s=bYf_=gX?EukH>6-9@h>X&EbyFwz6N9jokn3Ur&D
zm{G~4MyVT8kDGS*t#c`}Up`$ss_HE;VlOupiYe!69J+fZo6{7l-H{t-KiiBys{K4n
z=BNIu7@1v%{>vVjSL?24166lT$Fsi`ZArn10ClSPrh4ZtG&2&&jiP?%RNdbf;+}1x
zBiV9vuZj>!k7M4Aw`t<mfa~)s#S=3#Nt!){+xSYG{TyN2`te-IW%GVHLrXyt^6B8-
zlKBZ^v=Or!^Sl(h0^D`~0jlM4gFK$vYh|(t;s-ELf?g!oCEcX3l9rG@&{CS(oBBW>
zFYWr&@DI_xTTefP8u(pTBPZz^Xd3q%%umgAp8O?9?Xj|HH)wz2+UC0U3~wRcu~Ug9
z%Hdl}77NVFMiHln<YRcZEz!duLwYg>{+SVfX9n3NcHCPZJCOn9&?<kRpI{&?XSm4_
z48L6#x->Sc)%pyR7I<yreo}5_<Y5a?8j4-d4OwL~dryo3>+G2H=a)&)X1yOB$Lc}!
zzARc+NGBIT9{q3|>>V_{hb@2)xhK7<efpeAphfI-hKN1d1e2a>fUb_B(0_|8VraN~
z5i$7|#-aj-_J?7ihIK#s<^s@4p4xa{i@T*y;y)lQfX5<Z!7y$e21Fpw$MhUKOy+eK
z+%c#2pHvh3!ucnN<#|xE;1y_&#Gt8P1f@TGaSq(HCKt1o*~W8RsoS(tp;|;6uvG;)
z(H@@zkwBN$qRbbC<LIgK2HYHoPxAvYFYdsawp2r%!O?=&&qX4G8oZjEHsP9Zere|h
zyLAHOyKfqf*N-aStkyNE=7*eI8cO-Qe=zwB-~H4Y{(_%U-??0lgg+4y>V#B@Q+v<$
zrrX~+z4D1pU7O>HcER)V_i2-lv`WqByk8ih-<y^XwHNAVkUQC`H>e;U@Nh-A#q@tb
zCM>iS74lS^^q5s?<Ub7r1+(9W->lf(VHD5g^ew{}&XvMlSAcZD-x95f+%`nj&iVT<
zEz^3iGfK`-x`5|eJC<kCD%P($LxjoNc(>sdqGxlr-tu~g1T92U$8e=5uY=5^?q*$o
zHXpOfI~G;Rc$r8elG_c>^2cPx%4qi6uF~vN2cY)8m!#9=ztt);Tz9gxO<|GdaClo<
zGQ;<7h?xilN6<s-g~W)5AaZir%#UMtAUiw6Wab6$*ML8g;@o^Iat-2VZk%$b)ht1P
z({A?Y=YmJ_uT$<&e)5Nv9tdj@tt>ZG`oY#7<ZC?3RzXjG4iP7MA1ftS7TiC+v-V*_
zYn?4I1JfZhE^Wv7)tX_a>ann<&x&>1-azn;16q`|=hDnK=MZg#&z8gDX^@Yt*`sx#
z1|IV7Hb*@uH{U6-Y?B6;XeY7HrB$7UbkM5^?k1EUHeRHLv1wyS=U_?w#M12m>k~hq
z+Id1<u+CSKJiPD3K4CVxSsQyRUW7}OStZW@3aIl$KH`^USX~Huk96TDCQEWZ|I&@z
zSEQ}z!jDx>OehsSpkxw#XE5e}qlI~<-={5o$e@jXpu0c1QZcGgxL;~vZr{NqqDxL+
z^ea54aC~-RyU3V;|EZl1UN@&;=Ka)d{IWN1rzL;9;t61hWgw&t5e1!<F|l@6yq-<m
zLQvO|KEw$N9ZA6VV!lVZ!7LqN<oisaj9LE`hIw8>7?O=)F>!MR2IynG90v(_wvy25
zCh;wD#-npVQ55M{?e=Y_^CvHwTn`?|Am#<3f?q7gw)T-u43|Qx*y|Y!DuXKGl};Nj
z8YBD&DUIMu@68g;mq`!3ms5=5LwF~&K=1~vFS;b@wS9tp6J=lHPTi-N+JK=--(@Wx
zMg+}FB+s1U-R324H&?M($3(kHyP!iJwa%`dU67jc8Jtd8aeMd3T|=5Vha5dTJbEm1
zN{lr!&r0>W_B1=RdB~8++1itui$udH&0Su$)5<t|xJDC*Wb9}Y-G3IW(?9!do_fA#
zBCwxL`p&cQD?c8w!@Sh-&`;={t`0rtJLpy4vveC`B+j6m5sla@|FlX|5}a8A<Ok*)
zwfiH#YP^d4lBh(l9aNv^o@L%YiorK@2w!xZ4KbFi=aNSY8@+%d_?6;fG3Tc|N=n;r
zRgRDRZXrbRKd{i5!Nb7!m>HIY7+G8%UGHvbIZ%*RERx0)-$kb=!_E6#5mL&@RwS&e
z;S~w`S4`GU2ys~w<9{ZcPOnbc=j1)7%HQo-DNtu;K9Qx(F1L0vEZQ3)%%G_>h&68-
zKj=@NbzO#IPq1NL{f5QCIFnB3xW+(gCmW0(CO~Lr$Gzp20mB0BBDP@0WrtYvXjXY$
zzCyS!;*yKUd23KCbtuZxtxBfi%19Eiro(*L>*B{R`1M24q>yC$M0H!1vv+bsSO*y(
zzL`FxC9a_&VK)QG86_djfh>0z7SJ;7FE(bfmyfSjHd_+TDMl4#Y6$Bau=h5dJs_~$
zDt*FUPxLVKi(v&<krl9leMR`CYKouE3>&U^u=P8yNGdfKq84+wvN$|%lvI8YGl?=>
zu-;JJ>Ft<up+xbm9PPbr%AX%@x~O>Y{ibcP!^yrSkwrD8`RSUr$sU^Pg0=bNqD9lY
z9m)%l+=l>C>O(=-ozaY(KQm7B*9DJU6sxaJWK?V{^Zy$#kZg^=8~064(5C2Ut*4YD
zDo^CRzz5B7Hmn}^UZMCFz?qm}xy!P62n2;{1K4+}QW#lIcfjG3Z{q>bRfzJS)R?c=
zI`;Y`8WmM%P|50<Ty~;GjzL8|u+2Y*HH1@sw5~5)V@x-oj(*o^EK(?OKNA#nT$Qnn
zMME|O@bDLFQW5AROy6^LW^xJ-fv%n<hG;OhR-6^K4@EThUn&?ZhJE!C4`7fq)%m!m
zW0I}oC$9WLOp=~q-6v3N5HgHT;s%<{+Wz;ZYpZ7=Qo%*78b!fGP^A6rrFRh8fHR*J
z(6GRRIlblO5iOSyBH%>2y~*~tIzK<%7HC5&G=(+Cuap6$-Xm3(9<sBv&cQ&9sVWbS
zjw@pO)3r*UiSaf71)2O_O}|6XzPi?!D&GjvSy=J)tiqO!|MMT7yJR<FHOb6%p3eKN
zxOj&SA(f!b>pd@OuHc8#pprgJ)uKUkou?H~4x@CdVo1vUshu=Vqaw*iG=AtiN~0c$
zD>S69#_i<@2=)rmc*Tvb?NXH<Zi8$_k*X?n4V<S&nm;!0>`}V`WX$IVqt*7?TUt!M
zLt0Nxr78AyFRag)`Fu!x7S6Pb06|omb&aiqrLeF5q3AH1no6d*HAscEIkM3R*~+j$
zdKV$qLf0lVv*_=(>yY<+FM-=Im_Qdeya_ZWKJR&QOuNLJKsK;B#PW&5D!*1&S96!J
zImY?RECwNqmdDfzF&z6<y;V1|vrEQ4)3E)DirZ5CU|_ktSK>M~TCm*`x=H#PWN75-
zR&lNy^UYy~a;K<fkIQUlWVrqcdjLk`|NF;l5cD>UKJgx{egvc6rIEn*Y7t?cyZ8-2
zsU54!*l={>O(c~!`OF%oiHg@$$!<fnb=uy?nKi<Xd#mGa8~g8kHRJ(V8&WLHERs3|
z)IEL2S&-BXxT?+ukzyi;t-&-yZc_$WB0v2$>HIo){mablI;egztzNihbRz{nn#?on
zVu%{iSP)BZ0Zljyf)b0r0_PtRf$i>DFj<+Ew;c1Q5)U*sb#3yt52*j%GKK^_S51up
zoO5B#W#5J$72Ea6_L6ZoA#)gRGTv#fV3pM$254Xn;sT{y#R7MQ|JJ*-4UjX{<rh?R
zjll(A&R*03)Sm*<t+LaO0@v(}X48iSnbm1Sukd#YtegbWihT3o?keAC-+FEn<RTok
zo6EO$W9%!qxcI0a%T!ciiJ@j+b2}(Cc{M>~UE<TPAk``)?QHrAp}Jve$5`g`_}DjB
z^UaUvk2JDkR%2F&9ktWPdfonY)&F(*fp=PR8tYX~@Wp}mf_k;9`_54sn)m=w_q7%S
z!nO7tl8y9<aCzU&(pwwUlSoWArcKD(z!evtS->dzxoJp~(S^tZh{igTj>@6%8h!9@
zZz>3>*5uOvgnogJL4Sf|w{z+T^XT*fWSaak5v|p1f0!JtIsCE_`%D=!0UH3Rd4rb&
z7&^C(5W$cLr6m*uqKL<{{SKZ^60G{;4jht_J_J4RKZCk%PD(zLB~IgPP!H}B5!W44
z8E5w9nAt`CgctV;m@+g|!Sa=m^zdk7w7QZe>VSi~m8X|X4~zT&;IwqdS4)uVOARjn
z2yBt$$cU#i2#c&5aLiLAVMD%q@6bm&1jE%C-Ro&C1Z`Pm$D>av-dy4Sc&Y^9r-wE`
z%OR^~?OVrBL8eF@HjrXNz{Pq<$Baf;Yb#v#w$!2A4-h-O75xnoV-uv&_jRyX%WBG^
z8AQUB{zbs|h57fDaZ+dv+9!M`e^pT0H?|r2JpLW!fds-%msb;+e;1!>h<icoS?XA>
zX}sVo&zz4*dfzWzxnBJ8>iEtxz;M!iKS=w}ALDNeyj2Fnn*%k?7o~LK6+vH=D?+f@
zATW{lUK7{y5mk|lFHI?pvIvsRyOg9k9Z*K|kmSg@HZgt&%!8fM`nl^;w^KP#!wL?p
z@O$b*U`Tv1nFx7<*0E_H3|<PzT%k>>x+ZZwj69zUMMN7ZvDBy=t)`Dq#Q$hkcneQY
zVA~gtV!(vcMOgy}vw+vNTJ}<~cn{UzJ>mg<cl@(#M5D5?UNrh~*5;|=Wx)EV)c(=B
ziD<)+a3*lmya}pWUL1-~S7Ev=mIWiG<w0rmQqq&`0VC{S$pFkq2v^96bg?993{s1<
z@7>t78ikAk$VeBP*!NTKqb)U63Yex57KQ`6xnygm0}ng28E_dPJ8yBtL^t0srSQUe
zX6^aNrrHYkLqtA;tBp<^zgWg%=Rsa*(S-jnflh9dTINmjX0Ygb*&O>*X*c8#vir2T
ze-vOY-Nai7=aLJ0I;ZxM6yEKM3n4`%!$?#!Z3i1Tb8XXJUek+4a)y(WZm?DSBJO@-
z3}HfPe{W{SHOczoJ}Pyu@q}~>W-Ht9MN)Q26t@dZ>!yjJ#r`o=GiPb2F@;gqe$R@;
zFV?{&qWDw?YCA|bH};bw)FR7dWS@$#wN3~8@{EkX_(H^bWQp-L0v-kuT+W|mWW+Oy
z?~{t6s4cVZ7nN%Dea&&TD9DQ|n02jlgGnIH>ZLf2AFI%;@n}W`!-QglHp-jrCYr=P
zM>;Cq>5IoXpiyP0sT8GY?5-%D^cN4tS+*_{+dRq`o^snFxxi90#k+n$kwl4DEU_KR
zhCBzNX6Ar*gEz3}e4hKGnw)^woDWwQ0(5RTUN7#87MjO@k+;dpfL^tR<vguY_a}Mk
ziMI57L&9VcT&(*z(@ps+#X=ly>N+1^<A^nzWrCv4*T(dVe_ZVU-6TMG3?cSp=lf8d
z?oVOJ-<aah*p!DmR-+(<jXnU!Mdm%`ML{E3+i1?e2`B%T!|7hfyGP*uol)y8yxR$%
zyIu%D5&w}^eMySDjk9)}@uTN|{Pth}*dGtr+c-uIpZ|S<|6zas^#zR&_&>pSNml&+
z^9x~rl6VK8!E%MvHU1HP`@cW9`yrOP{{L_2zmWTXULkQxJS^sMS0M}i&ub5hZ6)8m
z^ZTc%3FCsT$X@UW6T!;{(sZ-xuu&AIAkp_r(@o1^V?gP)u#XM0iGlyH392(of0}lJ
z^qtL4;SKF0kgDegwvq}=71pm`4-iH3`P*Hd1Ebi%0ht2?HXy+2Rc2|NO34)92gug}
z+ygR7iSRnJx8|DtK(8J)0k~Pvs&a1-OG?|mQi`ku!1mhsC<rt7V4XeiVO@}{#3%#X
zU-0)oqcPyd+MmxI1poLA|93ve*D25Mm>z5|_U|YKay>hVSU|@qMCWUGf%I&@-?_3_
zb{oLmT;E5rtk2QhF*ch3up48N?g4C{+ChI}qQStEjYdESj*c3D=$LJ%SMxrg>ETYi
zLo4eH2pul2mp+vK;#&_0K3gJGC&=w*VFwYyE~LPG9%R2VlyBhqG?)C_n09|$(#jvF
zzZYUas4x~%@`MCD;pupsfT1ULL-O<U$Ne9^z5*)BZGRgO1VIF(lt!dWrKC$mq;qHl
zq(u~@h6Vu@6(yt@kZ$Q5kWi$gX6Tk2I);vKzvtd_?m72=*SB1Y1;U!Q_Ph7~J<s!d
z9@;68+|DF`-0j~4Wk)D5|M`L6zMPd<WXkm(*d&{PT?4K&%AvrvCuIiQ<x3h;Tla<^
z2a^)(Zvj6K>hUbn@0X;2q;UoSt>p^9mD^?*G%_~>PmZLiWnkq%3c0MNQTB`i8_3>z
z1#ZZo_mxxS2nW=jeYkD(_qhl9K5+e!oVZ-w%FmT*FaeFe$1(HtTs*u$K>}=~s>8^b
zpPhay+XmLZ_Ej<!m1)3{_E+A)qU^b0IS;<3!YRO<e|?!RFA>MpzIYk*b?Jkq^3oaL
za_R&cl`McIxoA%hu!fQ)=a6-o`{~2}t_KVgechyAw1a=x_xyd<dHDhlU32#G1($>r
zw41XN#0wXw%?S~X3a5<d)q|N<7!E{nsFmKbRcPO;y>|~dHGEv;n9Ovp-+GWLQvliv
z`M?f{b&rqz<sp=^=#(wxW);2CmtH^h{io>>aAA>z{H`s40g^@Nb(^mevz*#E5N-c2
zYv!%W-8;St_-B9zZ0-{3dAf|V)B~;~S-@yl8e0z{GC#XUGsM5{A^K{6493+3i&@~U
ztQ2RT6~B57`0sXaey24bmRW@rI4Sho6aIpHS5%XdtUnv}cMp4SyGObN=n@+GwvU2G
z{fyXixp`L5|4^mrMcM|W)rj!|DWm+lDQ<o~e6H`F`7yC3oPiS->SJV$v3gyYmc|O%
zhAo&%0EROR97qb-Xh;fZyVP8Lp!>t~*2Q--Amq_$aUEQsPAMbx|9QsRV2$~F<6r5}
zy+82Dy~+Fg4{x6XUueVEW#RFYq^<ctWl(hAQz(dmb%5wu39H{MX>~hC;xGnemWDu%
z0iP)rKpjPE%)x9Azmr^@;u$Xi1A2=9Lat_G{cH7MjY1X_3ih98mOX!5w|xaVdn>=&
z+^YdYQ*;{Fu!DT;r#<|5t%|T!QVicz$_Z@hdu;59fHr|?r-2bPu4l4+qoln(O!_Ha
zxnQ<F;c4*DQeS!t(3V(qnx(i;R|Dd$l05s8#^PNNp4fQHu9syhbDH<m6S&AYNzSWy
z^%Q|d>Of(VSID|szcQ2|`q!z;V(Gy(X@!NfFeW<nyECt(@-L7{-IC2kff30Jn9gUE
zk9}pe@BSgZ0Twq`+}zw6Bz7w39_(TP_kV@xtfa7hyTkoT>y2mirk(#;qkz|4;x{tc
z52Ug08pNfxifN;ICMXrQI~&~MG@{?X_j%@A-jm_h&U+pelXiQ=pv`f+Qs7Wb{LrT-
z*hAWTk4QR7;;Hd=jqpB2iidcaMxGS9?~3<gUPU35`ryus&(iSW#nFr_g2uKLqE61E
zLv#;6d#h5zZ!WiN^Y?ii-DA{Z+m)!Y-8LB#LSa=g&JPFK-2`=I&4=CU)SB<8@I9I6
z(>aD}&xU$*0<<CsvjxN=G(MLD>&}HLF!89&0Lrp-AKcQSH}8qSBif5o<0%I&pJ9!s
zYt9zmsbK)P%6{!-_2b<opuX`ku#VB=Q{rsCAru5$QI;2$59+2)1qvX`h&<M7H^w2W
zZ+)etY~iU-H%N9#aHCCPfU>gTM!+xN`%IIUtVjA~9t0HIn1#PidTr|I1i8Bfw)<~t
zOP?M5R=c)>&Iff;<%deHHa}0^XJGHUz<vyzfI|R<cpKpNA`8?i4PE7)VN0Q~IiO#$
zRagNdkD5G)A{nT~m!1Q1FmAC>?1LDHVBzF!cJTE@#(cM;)QYZ+E6))LwdFYdHtYID
zGuiVZEOKQB2tGvn;<q>$FDxRwnII=b_b$);8^bs7Wx=3yzruPlOKt3+fB5J+5EQzJ
zT1f}(gNo?MBm%gWd`N~RHJT7YWN%uFU=5R2j8DZx_pFtTxbM1$7?Z^0HqqvVi=uuf
z2Q;z|qhZ01+{Ey<kT$ceZ}|PX)}Bedte=NE=cLq4(@+|+5oGkI7U!FNoQ)3<BBJV|
zxBFxgZz6sl@@**O<i@N0x-5P;F#q7@ecd0e<g!t|SM_cm`b1IcxYFNAQA>K#nB|)Y
zf}a<^WVvy?=IBJ-_j)9TJn}17GA<}iw5iPyJ&KXiSZF)Wk|`xSZT@mvSd=nwGiW_y
zbG6es)X~}h);p<{xUYM8naNTvG0Wl{-!i_=Kj8myT*8Mdtafh$+Ki?q6S4FrLy1K=
zS23yv;-PU~fT<gCS&eUd3w46KNZ%l04VdkI^AQVDnGz4)d|Ybgp5JQ7cxFS~mQRNJ
z5}009-EszS+U5Kbw7zk-!U{!hzr`ldY$te0-?(~qg!{#d7u@AoM#&pfKthv9OGlx9
zDU9P~isrL=r7ah}C$HmPBZ%x&@n3m{mGu0ecV+-Ccg(+AwjiIr?mYRgCj*ExDnIbD
zFAnKkw4f>O29I~L^@-^0r|^R}gzP83KRzF2r8)(pNyJ&d)ualt)85P^e^JT6hCf~i
zEFl^U%3!Z-6*|Q^ZIWL6fYf^d-`>T6R?HWlG*HJRO3iayukWkcsQ!*7e$kd^F*gaK
zZWn5-DV|&>9BnR(e}3YLms8!X-RUE~D24eI8uwk118@H7bc(<~TCrCU7pf*tdw@==
zZ=r#f<uS(%y9v~rmvNMr4V}MmG+RV{d_5gM)*%#;bT^(i_H~RhqqxQPkgr{OJnO@S
z?+=SFG@*~1YqPN39dQ1qich{Dm$%bJGLv}kmW#Q)VN+1YSDUOoX#cEh>gE|fOG@YB
zD|5b`*jySubmrRNmKQF&H37r3qh)@=m;!d;i|k*4j#c!uq=)eWlf9?Ije8cJn-pKx
z4_PTrAlGEM9PiI}m2*T`f6WcuJYk4XKW9&Fe@^Z_>l6J-nr#!ZT&}r-TDsnMxsL`n
zd}JELLACgFzc}#rvvvsFbWzBI8CZnWA1vP|pMhHRJB2VwKE_8eUp+44_J7w9R6Mxb
z=D!ecdr^vX0FUd-GAI=nzOTi&?Y?Ko_Oz-s1=_&MhrT5T?T<Xk81DAs-+T4l0Ea!+
zo_!DOKMh{U5LN0e?K7VrC{fDV|3o@c2;Q#0WqMLh-A<_>QT8ipg}QiN0SH^)os6>+
z7LcTtt1y#WQrHjt#+>%&|80LP*Tznso#im7^76ZTRZ_s{dbf>m9LZwOaH^iuBVbbP
z^cEKLheQ8Y<#(lbS-wCyDhlr@Bqug6e<yNQVo;;^Vxo9UzDWN5R=os8_OJ)YZn^Af
zd-Y-jdDvH@T!566^?EC`QOI2i_Q0i*58r>KA=IwJ`ujlY72?in+wD&+(U;PEzHrd~
zEXByH?lI(+EM1GmVu^mBsD&~k44LxgtncG|cOB?%3o&O<)&g7@sjq7PSf?-GM}7#7
zs&>3$F?I~^5i6z_RN>=Y<azf2w0sB}ZGz>?Gj4YNjgpXAt*8IC0k8yKW#Ab$h^HqS
zv%QnEZsXZ7I}y(60!x_vQn%H=wLtVd;wEvRPwy5|+vi>?IWSfy{?vD53H4<@$aZuZ
zgO>%)H1j%NLHK_Pe?GgzKeIkJ(7JO~`6@rEUPAn(6sOnMxG~Gl_0#smq$)0^_%{(b
z-X5ZB1G)!8v{NbOl)S|a<2T<`a&@Xtw)haJO(IqkJ#$1uR>t)A_Wd$2H`=&ZH~o^1
zSIfPsPJHOJ3g9J18ya*#@gU;9_D7*OPf-olti(5DM0Y$nY8OpwX=r=jdOfY>0%0Jm
zerC^`2t+L|H$$>GXEGV8Y2_L))x?Kc))o#nD4Aq9ZQp}mlbW3dMFV{$NLh|V|6YES
z(hamNO|N=9uO_)P__!AMy#B#<dk_AeK2*d_%fm?XW#1ivo>C2txs@#i>@A0S<?!oi
zJ3g!h`p@4YWiIvYmk12n0JzGnIkfWDcBQ@*`pzi7vRJ@$&>~;iw;h)6IL)WPrHpZy
zNhr76*~D3@Ht~h@4?M1++&?l47(R^a<Rimk92CduXmHo3{EBCinjFoc7Fl54@K$qZ
z<d^i!r!l3_9f4XNsl(My3n|%OH!gC7+4WoA2)%J5fhzuW7(TM@;cNT^@0+YUr0+#9
zU7d2mz*&4aB%WwTti|T%<~C%(qt@nD%Y&5DOvEv^7tSv^Y7BCGAhl7mj#8`fb#e4<
zGci>jsl6G?dCd0m>W`lMA$zBrr{!yY?@)+W4U5|`p<ky@e4EWApV8=>zu<P3FJ5~H
z6yuW3`Z{1MG!Z2SI6C<fq43w?&L!941QyLjEkAuUNtm;{rwZ+=`8|Mvw&$H~7C}y-
z$f5p&QbjW@g>9$+?bI>?h^p`7?PQS2diNGLS=+lRE;D~9S|vi?hK;keF;1NKo~8`n
zyg$JQcT0LQzJ>3{4u0}aTv(iAB^;cHrilFJ@V=+_ySC6chDl}3VV~d?>~>)dkF77!
zGkuvWDBa`rFlB#tO~_^9Y=+GL=b|QiHt#aSL+5W^Za`cgEyAi>bE*Syoj<{1^1waT
zTS^-v&F75X&-3Yix6yb!U@29Stt=<Dt(q{zv#Vl+GggDCQ;Gc+DE<tGhzJFrc&b}2
zkH04J@wA2yi!nvM{z9Y-lHv2ar`Kv8kf2SM;my+^V*6g0$tSRU5y5y?Q{JSI%Z`Ly
zdfv}e;&&p@WsMEjbo+2ULWM)UV|%G}WuS1>YB89GDa&UQbfCsG-w5ZM)f+samVXY}
z_d<}MCnh($8HiBdR&D{NDt=mx6~EWvvOFzr%nkwBARqXoQHAIyKMwf>`v&hc>v#a)
z;|jvxZ(v)7w#n8Cn_wxw4h0lBT=^d&`@>V>;6J8EVA2b=QK9?k-tZq2R#)i|Uw`a~
z+&Sz5Fu>EO%~tD6ZZ<*et&&nVrMpI?nf*y`zuC`w8ZVXd+MGz=bA^KVeU4#Q+|nX=
zwqI{Rrhp_{lolvgZ<?Kl_fA?{Xk2@AyeXv*k3bbue_v9oBr+UmJS*-H)f<LB3^VZR
z;Ld*HRAk4fz|!`kRq=Ia$UE{<*B<F1Q8v}jzfTQw1QuWKKmTm%ITJGTic<ZnPqN{d
zA&i@?bvgKo#7_`JZdNyw_I8o?@ab)rR~>)H5uoo=jqubik>yR8{@aEZm4Gwp59;Kv
z-Bhx`?bqi(bd8)w_UZ5~%Ig#bK37>eezBU+f#A&h<HsRkEGglh1~(!t)}7u|AAmyJ
zt1r)Ni|Hne&62^l<<X_o@`!I?6zc%36=X_7h0Tx}bj8Im(o4f%lBB(}YTO%jbdvmH
z>DV=vCR5L9)xsA0(U;@-qs3>wkcJ-AC69koZs<I{zA*w&fyuYi+$!o_z~QoBM8MOL
z37Ch8T2KYeu!b@8KlRRZChsjtd{)aX812od>hQf)7q~_II;=dX(0<?<>0&ZhiUb0#
z*}SJCA$}Z(k^jSynF`OGUwrKk*KVhKI2)=q;~eIk7Ar&=T}$d*cY}H%+s3%>$MrkE
zIv2l;1Mq?GNkV)*7~PJPxj%E2CpBnSAWNY7%{Px9R3F6Mw!-^Doi?W!db$0wufXE1
zOTp)DvS$`E9gx5j4SOn!-wQmYhhE9wGdqH&^0AT`<ww^<PdPR0xC1{Mwn9t;NfQJc
zUGTMV_uz;EWJBMz7vv8LOCY>~!~I4w4we+~TH7%VK#90;`gVv1PbkTX&7ZwZ(M)x}
z^j#IL(SpI|+j>yMHUmq_wLm37lZRvhu$yh6AAP_L4h8TPZC+jhR`v7uRy~F($iL-=
zDBe7%eQbU!*vT-Z8XhVl10P-2T7<%{)In<1Enxgf4|!uwNFfludxn%s;$mW>c5KcH
z#ec5^q3<<s1EiRmZQu&@L+>6wd07#t8zEcsL=@2{QZ|iQsBhcS2<R(!7~{d6yHFMd
z6&|9;k~2Fpf#fxemxMIJ9|Fj!KE#nNC~|Boy){^vci*Z3)Kf&}b3+*^+LH(~(i?(h
zu2bM;;CMC5{wA)fnV5I@g;11K)NgYAo+rw$twS#3zE6sW8kNyX;!--7ydm}YnwH(v
zqd_tzsk@M`mIVj};u!q2Po+rJ+Cp_Zek97#f@nsG(JgwO;XCgRz<Ej-l0E$_zXoA}
z#VqyF?y^oe$AvaRq2`GVAnx}`CaRj7=36*j4&k*Hem*TXM*WGF%!ujQ4RMQ2&kL}`
zIy6sz)Zcbl5b9gSsTn)x@3^{0h5rY%Cvgi`;oFwYJ%gP~k#F5-HSQ^twtrVvf3WRJ
zBZRNns<8b^{ZcyOl|0)L#gS=O%h#l{+ujm2?D1+$7S)CC%JE*iZL+U2HWl%|9}yH6
zxM%aqVf4xPXvg&H;YX508)yUecL#T8t=hU$JjK0>Mjb}i*N@-PiwPjE1fvL`NG=bh
zFTVwYadMTaue>3*TSHdAThPrQA@A+#GJQ;*{f%|WOo$-WW1GXRczN!^uzal`Zh7$K
zSGBDFfn+NnI9~r|9ZdQvd+m_t(5xN%BUF&(e!k*<!n)W(p|s_Mqmhc{v7%fQ%d*se
zmYKiSn?Ju<DuuxRku-GRf4`-F>w9{LgBJf2X8-S@2MsuDsJWz6G5^AUVUT}6?nj(F
z*MH)w|Ck$7J0ZNlI!vFC<3CsVe~yK#yZE=$`a5*@2f7~%r2%waDZKJ`wcCFm6hw>v
zdGUWgzkh!pD4+ti7_p2-jB)=QME^MMe_Y7_&nZG@Z~7S^#sZS%61V=g!vFU{FQrNl
z{^J(>^F(X;z$<O)Zs+#+e;(pr|9Mr;?aF`Og}<HXOGAJG*RA$qzWsj=bTa_M@IP+E
zpQj53NEW74DGPeGIj_G35;9!_sd}1yGSJFu01Q-vHY8sz_1*#|w0gQqjwBHvic1Qh
zDi$FBkr+UR&Ngm>^_vRUZd<-`G6vy&(C*<*=LARyft>S*z@Cewjmi6|O?;)17l6{j
zAX95l2Vb`Ub+&W94e{Xd`RWAWV=C*Twk^&}NF#EJ`+<~N6m^`U_XzH9N_k7xS#HcO
zzX@Zkvy|+7@wxFW(xT}K4N6hEwdgeq&;1PIhdus&8w;j3Vs-879Eqx>Q~S?GlH>P%
z7n6Hdi`t{Ap!_!gn0QKVwNe&>J}&zmYnR7AKe9r_UUa9Cwzee{wfKQiaundE4#v(%
zw;CzPa9<j}zwQJWhd4<<A{37SIR6<)Ad7->H0L{XhdyxYWDP^;q0?BUh-t7!g|D-Z
zWF!K)1F%2`-=ke+nIxCG+9LL$o|CuKyiY-z-C%?7arBxMCms<?0M4IZK9b?$r5<W#
zs^eo)zrDIdX@Vn=nAMXg)?>YI9V=*_qPFGGUI&);F;HDN?kx6b(Bt?5l+%<cSyO)U
zC2lW(jeF+DKwEcyrqhV-HV{OQfDq3D(aY#x;7*U2f`#@xMXgL{;S_Yw{hn;U4kGvR
z4~ihiQONn?{{8z~zMw%Xk#`K~y?@XE+^Oq)iB%VcxN!>i`<p{ZDAjew_BqmSr;%>(
zAyjqTOjQIBhky$C0oogk#=z>Px}Wmuud9Ks3l)un+nW%%AVqY~WEwP9RmSkbILfzy
zX>%zMH2oX|rrmYGc$xBpy<$__s5%%|?;V>#cBX1mnW)ofv|~R&R?QuUzoxAuOMv|2
zp$q(aW3s2aI<UPS|1;MbiQ!Mk0QmFuiXpSDpw7?yu7Mab*^^$#N_&6Oq3=rw<BL{`
zEva0yr{#SQ+(GbDtyy~Rhehux^YbH^aqY`o5ywHksq7B#(+V$sHl+{cZzwryaqK1q
zqbdLPSips_f(OHT+3beqPrPJF4*`!uutvv^4OPFe8Uuh}Cu2nO2LK3niVi=%pyON(
z1%n8;N++SpUSLqUm0(v<D_0Z><7KG@NJwZ|e5h<syii+$u)XEXWZ0$n;h&G>nwE4o
z5~01AzOj=Cz39`p_jDq8{WweDM{77rUt+C%z+nZ2_M1F<$&db^81naFqv=dQ6fihu
z+!?|ONVkOF&ixt6S!tET0U+wMJ3-7jrSdk$?FjTTj(}`aXP&W46<XRHl+UMl4pKuH
zjxO%l98M0Lo+mfkE?Yf${3LQTmH`Kgl%=-QaqgRoR*`h&j`A^!=enDZDz#Rz&C4Ye
zUNvkO&1M}l_dl&z8OgJlg=f@WG_MLB1=5qYXRg2j*rVy8Ki{FD;w-vMzXc`fz8dFq
zchRYPF&8H%G%4T1I{m@>{=xa;A@jzboXPs{-@mV)Y8x40l7w3u^z!uwD;$TSk%7L3
z(ks3^Vg&ZQbY{i-pVu=N1f6<W_7L59par7POB>rp;QIi@W;k%S3*^9Kb@c=8GIOIV
z4tDqqzJA*k@AXzeP-E-$j(5)njx~?9GdpB<&ZG-gbhJ-PohNE53QK;_ZSho??<)fR
z$<payk!jt*3$uSd_ZnhyoNm1jE<`Mv>eB#-ErDn*VcAp-;w1Z-1l{l<E}}1KSUMVd
zb>i~NVAEs0(ezyA+z7W9%0a^7V2@AA;NCp;26ds^ypOZ)e-a+bBC}bpCJS)5o%z?T
zC739x#qfB8T#NB9w?W~R*!d-`72u%RqyUhj5j0yh<HcEu!gDj~;;h-v#6v(^E?`Eg
zbjkI73Lp;~>}nRlU$+6Z63#pT&<)o;izHFJFlF40w?T*w$TNBvOjH(N_Ej-R!hp=q
z(k!*aS~(2hB{qT_IyRyf;!z5w?M*NF-Q55&Q5#Uwlm9r8Cg4%bOQsR~;Rb{R>yRt2
zL1E8pnj~UQQnjIDop7yOM#zr;;CTg7-Hkub<VV%GOL(2S{~i7xnIWIgfLJ;)C>^*k
z?}g3w#QQ8zA4q|i-{01AIbk%;62IRGjr&cQ`-Vm>enT(nlSF1FjNUL~6(pC#!NvdM
z3nuwnT*Qwtuh_|_I$389?~SU;J~N6|g%*w)ylDAR`SRE0W2-?b<4U_Ceskd-eF0<3
zOy1AN%N^EDR`*BP;wB4UTKkLzO~vju-r&7b5Hywi_MZOS8SP2rwN>H|7aQv^Bka<#
zYl7RyZ-al|=V8v6{)%xu)DzH8lgyY33^$QMQ_uD56v!34`;{m4T|OHz&ERnV#n%Z5
z15MBNF9emF`-k%JsUP!)V%{ZlPJS?;A2-wsGh1R<ueTn%&t7E4+kdE1SS`=SYme%x
z$bb30KQ*|GScK<xjq7JqgzVMw;*9FR!U`M5I+y7;d$BT-hiq=^@i=zLb=Jby0+)!n
zh%!?gL2J}k$S3=mXIW?_RIL<c#|BBs<btovnA0jir__d>v84uN2Hhq`+0DCS=kodN
zJ?roiG57OAo>EciXitC?0#|7~s7ESR3Au*ikZx|t^TiSNI_ohRjaP}EbJX3`Y2qZ1
zrS65SQq@{cTa<Cx2#hLn-d(o&m1JMsaFn;AmCur>U%uZYd#a?-gaW<e->~lEU_qZq
z`V1zsBKxzUxA1qNJ%)_MXzy*nRC?$E=(6#V9>xy=I-J*KF~sheQ(#qm+JoE|_Ai_a
zgZe9H7Y_i0zXlrMUPwz|WhOv-r>eYE;T?oS%9v*pm0+oK)p{V_0I{D##~aYG)`Jvt
z(@60@J3KfR{hqsBtAYptm;WmsieLD3kGEpwG+pmEI#|fGV@{~JMyE$&g^yNFPjHRV
zeYj~i>>~D-9vpN@CN=7Z4*7f~zLQT_v%taxdP_|Kn76gBQtPf;Ztj0m6Q`<I`=DX@
z2<ykrsu>?}Y|Z;KBW3@{`Oc-l4*cDnw;D%mhz3_>B3+Je9%|MK3<z<@BE}aTvQPV%
zOE#kCh5RoEMSUN1IOQ*}0BN0DbZe$8FRaOb@|qgbozzQ=5lHY89dTb>!iy?-{g|BU
zG9Fxj=-P37C90#Lsv{_p_zg`Vkto&b?v(vmW8mnOj=Cc%hL?vyItRbc^tllobq_Nv
zV~8a%J|#ah&Jr<G`z1Fdn52S^-m0h+2oqTr5mnmuoS@$(R$ex`7EKaVC7(51)1Raf
zrV=YC#A|;oy7z&mttN`VJlx93<8qTSOduBhh}XXVj-6x6^tk--%PWTmwju&Qb>@oh
zqn8t6$F?6&?Z&&V&iRkqIhH%vn4Y6p-LF?{ZeZj;4xfKWTwA^rV8i>s51;%YZ|d{_
zP)bv76lUT?!AH5t|KMlH0x;oVub3tv=LJ>;&fP0(6LnQ2y}4oO+gwCOhW!c>iBC)4
za93xt_WD`eBuFU7)P)-J%H?I?WaG&Ymmh&OiH*l|xYEg343Q&iF0j^>`A!z$ug_@2
zZ`{-cwrI7}ww^g(d$Omq;fo;`e2BRq;9?KR9cnGs6P$IFab*Z3sb5AGE=cHMKrlg_
zE!Y}-9aBHc23iLJqP5e=2w3?kAcjXE{GsPdz`IbC<F(2nEop;O?;9!HowUT59FNqX
z)w2XDgL>fDy$0GOD`?GqZm*}CB>_>d3-Yoca#wr6yPH#d33=CJ-J~(UfAQ(B!j4cf
z=5;BmcQ6f&DKmnvZsz61SyAh4pdoHnl`4r;gSD^9*?d2Y$XBNS^y8;?TeFrWtZC|@
zCp8Y5p>A|fS>7)bND;Ws;#d39!<LZ+A?3Saq*xx9^NdNPpZE}-Nl4NY@+;a@r36DM
z@p3NCt6h&i<@XXC6mjo*$n6dFc&P2!glJo?RtwHHY4~@mES}Sl%o5->&e{~3X!Oor
zd><UQXvr&JPjym2xqI?5u-;(f5q3fMM1bZMs*#wYmeKn7#8H}e`kEAP?={|xpf~Nz
zJBL>zvc?~sSXpX%oSZ0+=w*|oUL4XIPfaT{p(7doF&3kHv%$RE9s3!*F-p6-+tbjN
z|BDG_dNZA6I&U(Y_*+q4b^)otb9zZbh3Ub*=+Qc6Uz_BiVx)cEuJ@B7ZW@Z86~gmi
z>JR)Ffalz<sE4l8u?KL#RuU#AS1xlqH0;OG<phLI%a_K3SrS^*?lCWXPhNvS%y+wG
z%9mX8_>|nkU-j5$Lj!z`cCoh=u(q@fKg~C1LeXoqjElT7l`qk4(wNter~&>KoF?OI
zJBm};YKb(mq9ELA-$*gvce<MyrTL<}IG)G85u;M^P(SZv#+TSLS1ik)i>1&@keUd+
zdjZ3)%_x5z85UWr9B7>RHE>}9#gLe=A;fT^D}ybPp8Ke{dnZ47xkfKHi&9epD4RMf
zVEN<u+0JU5r+sp+&OBV#ulVILCZy}g&%D*nLL9!uY)Vs8UnAorZGt^ECW)P=JGng6
z=J>Bw5V{|=ti3OgZXF`!^1l}Z?6T$3=ogqy)JytFy~*(4#G~x^Kd5f%O#L!=>y$~7
zbZ9cKZm55ki?+x(us2D&b|+SE;TJ{u2mwcD5P~|H(_O2Cp-9ad3+tB>>zaz;$(K9C
z_Djd2(@f>PuLi#_c`Ih%;RLG{I8!9+t@RwT6T+jQ?sZsgZztS99slAa4b<=7eA3uo
zKSptst&^?(ERccVzVK6IX@-~F`NsK~V44i`++{LHn1&VsR(K;P)WJTGLzda<u4dlO
zeE%<#j#tHMNHgDJW!d_>K2j$}xMi)4JU>1-5S8FH&6qi3?q_G*kBE#4<>JLz=t=aE
z8d`&lAXsy<8n^Rp^MWGj8j8J`MY;@e&9OJTMC{kQ<}xPhAI+o?P+#XNp5(-oB&l`m
zjuUU|z4XI=MG0RtA}`>G3Qgd(=LyhlKeu2--(Kt76+#r%@9`S&Cjv~f9ZLn<(2z#y
zR7wOjCtPI{mowAy#&q8HUCawBj%?W373YLjg0s3b*|&C{hSobmFn!uK5=E91({u`P
zYJMNXYd1`+Iq-ajo_x_A_S!w}3bAEy9GW@|2$%kh-_*0P^30N5%{bmn@YJ`GKVFi5
ziTyCF(4aKusb>}9Sf@u`Q`obbWwK+w$&r8xMzG(HIg-ov-(%IC0>7NW$=P0}n1tzj
z9qjY$TW|j91yJ59H=9>hmxiGWGSl~e(ZXT96Ce6OFow}}aqIrXVNfOhqW(AG+uWMV
zW4S|tAu9KOskd3~WOm$hTS`5<G)lolTs$$#oTRd-siw3#G<M~5C1w4g(<!BUcO@yl
zTp2Z+EFMt?^(IN{E1?6yoXyw<vTd%1*)_Plj}j2Kqsm8Y5^JdQ(dlYPVv!8friWOQ
z%lnVen$HMSl-|?L>n9C5VsZ2&qk^St%7tR@%&h2;Yb%C|yfjD3LIR*nM!y0kmt1Yn
z7o3bkJ|7Y;Xf}3bkqC422qxrBKHbj^%L{_%M#78oI5XgQjt5Bd%ba?X9ebHBo306=
zQywxn%fadG*agdv%QY2|wF?<B{<WzJc|95~^!eUnf(D;|-5ZL1c2GAOnb(~wvqN2B
zjC_bCDbKR+J*k1)eK0W0CG^TO%c4fd`lzmLIX#lmM^v>86%5s=T4^DCt1D4>XTD!-
z&n4W{_dw4!QKQB|(TwS*+yMWnP1C4c1B(TE&(S{jgEQZ?y7tE5JC0BN`A9<>2*bC8
zpP|0d#o+Jw35<yjQtB#B1y7w@wtS=h=}LioJe4<e73IFVm_97$d76^yByCw7tlH#A
z4O1pKG3nJJVXompvuR&gu2^<&?Wl92){`#M+>Gd1tkR2-ZOk_$4EpwkMvhn^y6N_5
z7t&%83%Z-{i_S0F7jc*9<IkH2B#RN*)QZ+>%eSZcY--QPv{kVF)YBEE%^MqZBOIEB
zf4=B7E{t0ens-TFyGCzKr@JJ7pmDTcN9)!q66xLQj_UTm6foF*{2-s=bw<Fs1;@Y-
z`zz|*J%(D_B$>L=N^Mk|+F#X>G9q0@fg+ipbEJyxKuKDf8b|MIIdJC862F2LDhG9&
zj7qcXAM3PwUP@d7qMwkkLau8HH~_rbJ9w{(^!=uWB&TLj*Ao!8!}NSAwk7xlX}kuc
z(x5!D81*O&Tw5>aTd8(6%<itaz{U8Ku_4CwON=sAW<x)2h{RN=$QvV_`w8TQkY{s_
zq|PgZWuT~?P?O&Tb`*6X5yDPE3s&e^&4fCAUew|pQpE$Q`Uihj6y!O^cq>H%gS)*w
zy+)}T%527{-}N%tYN8kugkbSt(R&=b!ai3^d}7!)GDq~sF3_h_KFWeh)WPkA5RUGt
z+_zC-6^;nRiVm)(?5mNDfftBB)|r#T-2wu8DhAZ%K4etW9hk#Xf>%+)L29i_CQB-N
z4vPBmmATqx*K!W#&Y{W;>Fev6@x`vWlc3(L#SEA_)}EsKIMnu2S-#-Cr8C21hY}4j
zrW6a;%OZ+Tk3}YGcI+d4OfiA^dP~@k1HdBX3E0r5d-pVe&K`Gwn{m=w(Eqz?`d9t*
z^C8Q%ISmDH;STz#V3uBgmEbLe&KomgSOm473z7v*$v2RJlM@-*>tIj^Kgmqk(Qc9>
zprXdXhI3?MxD%j?pRz$_YCs*yR>AciPaaS$FOZwfho!!Shg2~SBAC`vTF#dQ^CnrL
zlSETu0sHnmyLXpB<Wqf97_;9}*0qN+*$Ci-<nQ)0KRXj~MUHoXlat{7se)d6+jh;G
ztWZL@h~9=S0aHHj;pCT@kYZu!;4Im@KxjH4OnPR5sVMl?xW{|AOmgxfh2pQO5LyqP
z;u8hfq~ABgC(oh5B_)I1dbp!K2{4%XAT3ZMRgqY#CkS&-ItoTnI=VS`(;ootrN+!k
zB7*rW*~ARD=Wu6n4AjW>MmO2f3Mzk0QSDAUV?bE?fqphEpeVE%#tvyOx-abj{i~s8
zg%h?Rx{6PmU5#uLAb)k=-)b}VxFhgtjXW1iI|W<^sO%1Qz{c~FlZf~{tk<|nzYxFJ
z@yfnbOnZ8SOo9$Fo<;&thbH>nUUX*9ANk`~&%WjD@V#=Uu!`q!mk<x@H)tsiGj-Eb
zAlKbL1v|{bN7zf&r&RbeW|>D;uNDrWs8gMWHtGWXQ6NU6#8_VV=fi#wz`8kN_lXu4
zKTmBMP-w2wjlj8wvFhg=xOZ-B61+(*MQVg#1(F08=u_(5QX{#}os%1#2g`26Ve|%M
z)%f!{z|@W;k1><TF8;Ad_bi(`#_fW3+husb84%x=0Q8{GhfD{-Iba(jJs+*J3i6a(
zkdDW@%T{zjGVZGe#~>&Ovj1=k%$Yla%(8~eybDQgf&y{9J&bo8qR}P}eK-ghT9D`E
z=F+~@!bNW~CMz&%`K~e`7jF-|Benalvk0JOuyIR6KlG|kxzbJ@$gq1O3iwedEibtS
zZ_nuULnoa8{S9APLE9{+hwkbE+4d~pTx#bgw$?zT#V7%-$@IeLxO3y_Qy|lO8UC7E
zvj!+&Yvn37;u!>^TS)ldW&%b~utxUzk}GEf=mqH}2!Qel#l(sy^NsLTH`EA}MA+|v
z&PCU|Z1?~5KK*<61;*jvYEE7PNDwa`@m#LhEU;Z#4N5V>9nBUnt*vI!i(ryIDSC%!
z!R^)E@jrTtx}b+tM6=ySpCDxZUA>rZ?||XB27l}zbB|yH<$8PQWh;eJ^@^T`62Cxt
zhH@(*DcxMqb7CTxt@bw^lBHKs3v5mUS?@_o?Tgw}X={jwv|@%N8sUt*EIAhEaIWV?
zECqP$VGJ%0_410#S8um5{4O$8>=c0&2ECJ9{Td9Bn7&L@IOy@}kqq5mGu!R^EIz8m
z)N%E|n@d&R-OH}JCa46IKaJSckpWW0o=Ua$#a{PDztdHKaJvT%0yy+Xu}+g`CcyT4
zvQ*<|nMU;^v`#JV?cWcyy!t0>{Re>kF<(w=GhF9s$gzoIDg*eH=6&44jzZS2bXSYN
zTb|6`)hnnR4gX_+>3X+B@t_#;>9z}TzT%k@<Iq@YZ@yUQivJ}SI6bP4q0UR&3MM2I
zWu_`v6U~h+fZIEiWX!aL!Bkv7^BXT@vGy+3TeO63Nnn4!hW~aj(=1%gCFnS*Puo(h
z|7q}era5FQm*1$uAx6*n=$(eb4>)IPPqs|hkxy2lLe`|IFs&|opn`FH%_9S<4<wtW
zl-bkwL!m7k(fvwdf+budRx&~gxl|5(;W=1fbx!slSc#9Muzml{{2INTiIma2M$<a?
zEG4x}Xu|h9#lj>{cLzF2k72zUBp%%M!(M8cfK2b|KCSsqV_co`t$h)V!g&PBUFGex
zq$=uN)%ha1di|iG;D3VUy4g74EPB}-Zmj>hqYva1yV2>}A0*dr@o&yYnK`EDomv~`
zmpaM}COk?gl9wL=4nE&O{$2dRjfW25K?o?pIT4{7ZujY*Be$048~H&ke9=w}dSr6R
zS9__*i~@j@vY*S^Ah{0dF>rLN92<PQ3au<6PNU@yidTR>vJBcGC+mbW+E>?&=F})w
z_yUFH7<k`|(w%0X=+QnV@<Ef5lw1tWq?t7>9u*Pdt8nlOF-q7-P(bd~n2Sw`)QoAl
z#Na4S!wffx|CqJ{>o3`vsTAVK0<)>A%^`sFkJrC5B1IZkJsZr271NRVVJs`Q>uA!o
zoi(rzk^y$T@}?H-IOA!q=Kq@6@M_7l{HrvAJ}?yz7apJ=(8BuX|E3BM_g*L$n5-Ce
z7DARfoHL;+XnJ|`izZ6!%40VY?8d`tQN#_7nEr{k&y3?Q2-oh1e*V|({^Oqi?TZ{W
zK-KjrKF<CfYX8sg$xQ(z(1#Q&zx<E+_^-7pk4p}x{A~c0(q9{pzkUTRV@m*7e!q@P
zv;N<o4qhI-R6YERyZ?B{{yO?SSOAXQ%MDU~#sB^3JmAv{Z?Js-$I|)Nd9u{v;mf6a
zJKJ6V-=D6l1wNe`m;1v%sGPrm@jhx69J&c_#m;{WvcEn(^)2}He4>Xh@QIZlA8yZA
zK536ssC`mI{nuaodCr(>xnPM&zk~?6WZG*Y4$(+3%j^O7#<7kVUMG+NRt4dA00fBI
zs(4Q?_G`U<M|}+NcTV2|a4*dR1(%yQ0G6Dxlif>!5JO<bz?xnqR=~~4dp-uX3eu9w
z#9bE*w6s1If(4=y@Q<TF#N-4HQA|w>>zUGi05Zhjps`^LxU+ukNLc&;QZ@g<4PK9F
z7aah`Y-@n6>c|;>K=1Sf()kN@7*9aSx|}P$X^L(z<Dv#&mp1@NP%?0wm8U>F#ph99
z>3*-mRAl}IxQe=-++Rb5#$u8{@69WX1wy2!pnWO7Gy8;xHdB`pvB703=p_H=J%Pr?
z2@xP2B6urqXyyjMwIRh})^w(JzJB5O4n+JB8}X3m1UwkXQkyt?c@AZ&LfND>ED!Rg
zhd`pNnj2&}^aysp<1~s>qY~<$%CbcmmD^c#Cx{3UjP7b~>PLW{rB;vx$B*D&Tm;wM
zT~Xn^zkYov?OREa#+AcILN;B`7Xi%3p_cF@ve0O!z@Vi0>~zC_3|y)e&{G>#3?}E@
zA!woHH+q7hV92CGyQ8D~fbHise+1ZnwF5=VlYK8W55M_7W^<o@>Gwu?Z+bjXMe%!o
zjUs3w!jgbt>ZjvykA)2vJFL4|28s!RfZ+;97;GJE(n@%Xs+(m1+o!XU-2tbBddRn!
zOs*&24OW%${9@(@d5tB5Cs7F$n1*AOPNvJy?hdkngvbbRy63KeG}t>%ph^5<#n1#F
z#=;xuy>vZsV+aa(`I3G&5s=Uf15-~+QaV%@><Honbv7Oow4_IW87)c><>BUjvA+l!
z%gO+_DYUi>q<G!sZFm2A3IDhh6lY+<p;L{_n-}~<y1@Z(quj9U&=0)BBe^=+o`q0O
znki_patCb97+5UqAcxgaxhVKC_v?~Thy0Q88rR=$y7M7$?D7e*-!l-^>kbA`^-+sk
zUKIvvVt`BvNc_|hG82&3+ZbqcX&%8$tblc}Htig{-dtLr3U)F4omWMor1RGD;`7bj
zXiiIJq2(J{#!C`_sZ;c?jstH(9<gW0Ki+4Q5SU`HtD!F9Zi>_3trH_Po&?&ft3ikc
zEb(iHhc!W0feC3a$-N`1Geh|{@H|Gk+j70`0%4ATKi6+tQt^GoHj(`t_y_LPJl1)@
zSC+~LUyLE~3*|MpdM~mHj5(jLL6)uk7v<bi0P{SF)kUAA(S0cXrc$XKXhq;2EedXF
zxmy<?xP^Bc)e=RsQs)87yapg#QBv!n8f?z;WS-0Cz^p5FQw>PB<^JzRRnJUW1IL96
z4gnYULg!`y5Q(izQ$DNPR!x>bqjijo95BfIjQ!7etL<a!k4Iz^LglT|uEJ8jK^;rT
z)csay!C}bX)~|~}adgA-?=%eMW<BaX)P1lPcgOlI)Iwi0fte`q*3#<tR^Pyat!$ka
z7ud`?>1g~0h6pugKjZ7owJSOiIzyQ3XPOa)NHX-G?phwE;=omZ-q7#pgC^OkYLT>$
zK=*$BMjT`sw2qvr_uARC*f*M_Giv}ZTaC~BomcL2+lSqms#%B)%>mCh|9CJ@Zpz^Z
zvvJ&E)kNY!Z?7p?@Aq)p>;ePiE$UXCN7N^po60+Wd9hic-+1FbIcYzC6rRGhluGal
ziav&gEHLYhLR36%Rfjg%$&D-AH6rI0=U&RL;RYfAFWOr>ZPZ)?H<Um$b`VMV(dlvu
z`bVcgT^HzWNq^<VIY6q!EbfwFOyvOVfjO+y#8Oj}G>FG@7Yk`zIy|(=SAe9U6qu)d
zBPc&Pe3EL`;Ic1&&tEmX-!3q%#H#Jcbjy~2^?WPBfB3WUv*kT$5LdkiK=zs;iCB!T
zc7zG3*8Z4a<4GyW#kyeYbFSWO?_vp&kU$Mcj>cULz**Zmug6$Foqk=ST4&DWmyzC6
zQlLc0lFN+Tvp+ZoSrV(jY&L?>Y@%rz0Vx+IoC)+>l$(|zW|qLP<ya)i6!($Z1qwg^
z(4Cu&>S%#^Ee!{B`o7j(d--CZ+MYtW>IT9PE=U^l28k5x2<ljFV+-m;GStgG{t8x-
zTBcOcili)ld6LY3D9DO9I_pu5j~^hx*A^x9j7P%$^RVR9h<#+OKAk!@3nvP>eJL~$
zo_YimAoE)^xh|DYo!K<oM00YbU`jqX%$+8L_KHBoEQ{ht-n;sH-g(5)7s+C{NHN48
zs0ky!Oz;2bpWZL|GCS?28FpT%-_yi&N=fzn++$JG@3bg)XBjYyO|DwX>C;-4nN1t-
zGI?Ru-{1Q(-e)H@In_7O*0<H#tTUw?YcS><`>Kc@I_ed<&q{&wzS%UDlz=n!TR7Z3
zQ-yA{NSp?(s40z(jM8Tx>R2RBu}3dFy^xw<y9*VLxU2;JP;#vYF@c#cnM$Eein8`}
z4j@bBOL?YqzJC%{%F!p51<%toLY`nE1wb|fB3wk?K=oTAIOReey{0KmyypjGJw=OI
zc>2Vv_FJF9DqPo`lB6MfT29i+9hV=F21oJS%1H?zuFV~CaFG)*m2{bq_1t{y4UDEu
zSd5!mAYNyO*E1(I2X2fx^3ldw2K}aP(ynM5f=(>c$KRk>*?>Ro&>Q0LR=b}%*-R|F
zBKtDoj=Zw8XMW@f&qkgSQV+`n+Ax(u^zyoLvr{lZ-Nes$J&WBP)A$GiA?@dJr}^2{
zV#R&^Ks#itq{sRyoXr7FFyTnhoo^+B1_zje9Dw%?wKeRqPtY$ZN>N@F&vonE*Jcs8
zQ`Jw4zsH7m?0`R6AD;phxeKml4KGffxbOTYo_X+2SEv>tHgwn2D>QI@PG6(~Mlk^k
zA1SpCdGHOpz%or+HBjcK&k9gwxr4d~Q&m1gx*R`VT-r2kIc6~q-1JbR@+-Q&_v<D7
z+0-zVHt%QADtqRJCnVW^w!EM583lr-zchb38oB#<SAr;@_H<N-ri4$j7Q;<5T|Ucl
z?RvC%jivnvYlyhwx!M)KMPBXHenMUBlq1sp6AXF9h0u*k%wct#%UnxY5tfyqOu}nF
zj6$2%TB3UZRIi+Fv)*|6oizBquU;5l4k`VJm(pGd<=%vao{-~Vj0D;Ikq|L%uTg_n
z5>wz4TvFw|mJu{1|6xm^Tr_;d-k=1}dW0hewy7h%foj<R2GVa~X%_isvI4|#J$7!n
z?CDrIEsy5%g$Cl!P`h?fTPh7)yQdXsWwWJY;2b=|xd8=mmm|GF2v>k?Sfl|QWZ%hC
zopV2<Szatz<r&5_`*AUiXooQO<9u<gflO$B2w8*!>%K7NTEGM9@Q{O)*C(kd?*Rl*
zxw4!qmkp|Y<XjMfm<Z%R3NxkZ$*Hs)i99hx2BRG{K4ZM{Sx;qf^5k6IZW35hUyJq+
zr{Pys{MmxMsgtZIuSm)b`_?zbSpm}#kpjK2MVvMVW?V!?av2LqJu)@(2e^(C&TILP
zv=gbPb+wCi3Br9PKz7mZHEqrDbBb@wJ$ptzH5mBkc)vF&W-=FLxFAnJfm!QOL0DLm
zKk%+vKP?XM+1?MAg$W?0OgePuch<&?M#|uU$4uCm`NHBhEK_grE4eJZkhA!9)El2S
zph~?7|EgDmGbIv3|0%sb-~kzX#3~Ph*WcSOrH+rt?`#1t>Bz7)LrceTz=_QGj%`AW
zfVl&2uw2GU$#sTs0(q;Z*8UrA11~2OgMZ|T<@m3S(UC;ca7U0lzlao3v1&pY^p^9P
zZ?<J=)1K%G4zNt;$~kV%X^)hFI_$=QmCXgWx0f)a@N6EtTFWgwS{%IqU8J1V+8cCQ
zB`rZyPfSx;k+NHXNhK;&Mxw(VGdMX~E+AW|KVs;(tDNi~sibK$ijx~fF2`~MFJVKo
zNg1oDBd01o*sP4ph>g=J{thj8hW?wJs&UhyNZx9<CIZ@<__A(H$y3>zNgw?XDHWsD
za3^2_UO3N$nOKg3Z*p`0Tnt8>#Wj?FkUjIPIt25@;RH8A<P}UAY3du$WsCai(7hEe
zz<rQVe=BP5Fl{4Y>P}+DbU9w-G84v-UR~yHz9T82Y>X1po4Qqxm`oNn2GnrVcvbC4
z9lbD`EbV^jE)NM{24%XbvzW_wusrsKb-x|_BUj(k?=i>(fjTuN@lda?g_*Xwzi1D!
zPrI4_Ln1<1U&aR0Wxlc+8gpAWX~!!tByFUuWc*<9ab}Gi^*yx1<_s0n8Pg=QUvlT7
zSnM6WL!Yp*&i`<K0Kipv{OgOs^quifa?-payC%mcY`(@mJ-yGbFgKxO7-5=1^@2&T
zu&9+@cdo_{oby}K+E(&<^VmbH7VPQPoC|L-+@|N8Ef?3#HHs<KZKt6R1LsP;uOV!k
z8FUX+_qil5In1-*ay$t5Gak7}wSK(Y5E4p~<OF0$^FjG}yyGgWBpEje*P<68vG(&+
zons$u=MHF=w&16;DVp;ZJ(4;b*%<Odtq6+`FYegSuw?2gx`|LVfP$ikHm!%*PoG@u
zgqg!%bw}UfQ{2-q`{f2ZBEMP`*dV`BoK#jEXnRi|`&^a2!2_}WUIX-g7cS34J>5;|
z^*sj~k>4athbgr(<}xWFgtnF3de=!S^tW4*ca7GHmnJc%soXdx#}BPb)rjwKP6-l7
zS|0)-XX}*MGM^Cg^V2asokMig9&q5fMH)N>LITF^-3mtsZ55F}b=W63H`Zmd+Ye9t
z&JWRKsw1EKq>GlMHAfaDYt)u2-BRXKeQN8m_KoZ5=UvOib2Vp~^)m6#`C~}zh|032
z`*t~>oqg>H|Cs&_BHlN}<UYg#;{C&O0u#PRQz8?R!lfdKJ!OU@Ftw8M=c85f&h6~-
zVI8R+HipxEvb=H2W^ETDiKRFQZ?%ZQq?LjhE-5Z3wozXrWxCY3Sc}=!D2PZW`EgNN
z<~R4!Z46%7D2!^C&kZT-d+gQ_qu-xu@ll#6^kU1r?=i};{I`VTa8~l*CBM&?(X_?Y
z3+6XXJU(Z0;v}$N8=?3a_6<}GHngTMKar^O9At`bH8(v>*4eq?aING+zK2Pra+A>M
zopkEfinx!VOpa;M&$k14)&}yHbAvtrv2KBJvZTp6fGtG|3)Z6Gk9c~&uvRbYOTYv*
z2{?9{g$jQarNp-@AL|wSQK(d0i{8#pdD%{QEh=J>j!2XKcyBGy#I(bSVD#i`3fR}n
zSTi;1v}~OUCXI^u9=#NkT?XaaY>=DG>olu7`MJ@U6kYdq-334Nk)AAmT%CYv*?RuE
zbKV@S!6P@m0be+?Ljo@UE8!Qj_r%@@2pY#*XARA+gpZa5MISNsEl!XH!&v-&C5dO<
z&+5-Kk!gfS*c`{l`s>L|EsrAV&5;&DYfVWel>*KyDLn|;Dbt{!_V#nN+|Ho~V>_Iw
zvX>ns7~U5>+@DBlEYoo{rRu$<d%p9*Z+)O#{3bpR<#b)v9o7$=B=^?p7An~i$GQbI
zN_#ka%EfgWN)M;rRYqDjogm)IUc~(|T_qYK9enr8nOk4i<5q9J=hvSQxwi#H$Y0YT
z2vYi}B>$-U{5qd|#-X@MaLAjNs^{V7Oqo(mJ-y$X%6PLM+iglV{7O&F>11Z$9qip_
zG3Q&l-bF}<FA`2KTp9V}=);xJ%9RsR-@CW&Dw3b@U7=C#?>)ApB@%JG$`@2VVM9Nf
z$%zHqP&0lm*nU*wY^^`L2`-+St{k$2Y7<3<hsn&8v&+LtHeWX0(afTwEAE7)D!KCU
zEG>nZd(#lm3qLh$s)Q00vQm?1vM!Oa4hgm=OSrj^?A#<<+{5O;3|+LP0X^;9h<D?t
zR2>VWV;oTAHf8BJ-_G<oQhAX#$#rJZ?$qy|capn3?~TykF3L%s*>U_Z2tg&ngqa;V
zu_+5bQBsJ64V@j!Ba4<9mzw&b0>Xm*o~(7%Zlu!vo&BGwJjxr=zLh8vZRI8x(va9p
zRciA7FKWx$SwoRUK294$r(+;1z@*rzJ3W8em!oCs6xlw%lc;w6{Hz1JV>;H5#QYcJ
zlzNL`(7fyoskB4X^elGix|BM%>T3yiG``%*Fll-XR%IFl(?1V<Xo?1*VzdZaA+#RP
z!<?`MTr^wKiTNuOiHbCt4gnRI)0~LSc!pD1UgJ3>wCV7^4T}o8Y4l2&1V5M4Kmdt@
zcGYmwB$MSIkILgrhI<@E7kOka$m)WvffJgIeOwV8tLNO##X<0y`n7&Bk(zkwecwCE
z5WUu<F(F9M66IZV=p{@|{_7_9>+Jd&vy6tMb!R8`LLz3rh>V11P3De8PAG<SVn3c<
zFF=*lvSMdycL^_Ks`9s3uXH&2E{HgN2XP##8{DR43p9mx#P7jNK%5Y|VatK=LRPI{
z7VC<}{YU>%Z{7|Z#>ce-Xv}*OyO2%#XVp#E54x$Hg2d59ti<o$Gf7L#E4l{pa1W<R
zHB{ME3ouim%YJ3s7WyTRCu)XY<n~Mpi9=DAl(LL~vQGOMwG4Wq#7fBnKJ}Y#oCdjf
zEH?5a%0<56xf!V9`rX($u}mnF6fHHt{mfl(YG2dEIdS$sPF%^%v7#%=tKp7;tTS8$
zFciy+hPo_coKre|gw3aZ(#L6bbbCojCpD4#L#>r~HVMCVlv@U(qVN{mH@fGfVx0Uz
z7pOpZ7JPG&IQ9^1&&O?s@A%Zq)DK2qC;s7htsV}v;Wvq2=|{U9o-AlTaLVaKY`ncJ
z9{Qhh4hOIADu6Z%kL4YS@5xEub*`DG5tA|jcmhx~J}5Sk{zm1gqfM?aPk{jJ!hC3l
z<8Uq9jECq5A_l!Ae3=N|L4cW2v@&mqW05#Px$Yn*b1!<+v$6M$Y}O1G3L6}gT1pK*
zS&!~1)iT!A@GM~=IEF$H4j4SRNrtx7ol_mLO?RhmVolJ`-T8l}Pkwxap~#kt><37K
z?EdO`pZ9}6N7<*QNJrmo<UKcYAk$p-sIzVS*aIqC#8-1yc^S~zaOv0qt|)KDt~Qh^
zZtttTc?wcQxP}Hsn0s~&btJgnGs`xZkQsF69)s-2>faVBRcvAYC}TYYCG#*DEKRQH
z!?F;iI(1B6#~P0i2(x+qsla^2wFXk0?stQo!9oX&21k$Me6o9CCs8H7_T=!_k&Wix
z5KORfQ3JGO#e<$VB!<(V2z$F<P!GsJ)rgnMuGavL`)m2iV*PoX)rS{I9_j#kkaI%<
zKk5MJ7d>q%WcF!FuG_@LP72f8(8MxaCToF_)q$M*mO{R@m8Frg>_lfRMQ1>EJQ<SI
z<JI7Ooy+nm^a%pwT~q>I>*xkILa%C}7xnc7xDvJl{P_3it-o=UB3Om@mH>Mz9Jdq#
zLjC%>K{F)0=8zM;qba?OO4s4-WCl$GMTf*spa&zcnK@k=lx6h#QTRSrBVK<3rS1Uj
zs|(2qDq8#2g9U~H&of!O=(Z=a)aa{-Ym-pH`uir*ta^yOc2Ne;0{fQriMqryoLm1U
zF(;6ymqe5ts?P%-V{ho=Xv)`{M8fa!9`K1F2}PK*Xr4so)5Kx=GsDXa>OD&mTL$+(
zch8#;mjdh$T{I!#eAhnj5`iQk*qts6W7~Lm(q%v9Ou#of(|DBEv;>XS*jRT|X<|FT
zEcnM`De90;<NG_tk1CLR)~Gl5lwO=3534R0NQU5{_e(53yhJUm)4RMopiu>|u5JR}
zmOy<1d59OK2c10c8N~ZZk@|9=*5Kz#$LR-vv$HgOcQZC>v*259=<jnt5x)T1q98-M
zy`mKJ_&p!J08QjBRhRZHE#(&YADpIj#Y9w}2v@f(WWvZ_IowR}oci*hqVgH$G|$Y{
zNCce?5XdQl<tIf<0q`7H&UzSp#eu7CJw8+@-=oO<@gvt79)lv59JnUW^BkNUye?u<
z-qgHr6-*KceOwrhgt)=tn3hftaYvp~Y1ReJp2ygC6Tte(<f|r<%#E>u)3(wHGhS0B
zq!nN;Z!0@zDnAHuzgO-~;{bb!9g|zSJ0W{3LxN*Es5d9Xtt$!GhMyVx*^N11`mBl0
z?L;^`e;4}ujl5@^%60955#1U`zXU${>!5%jPGY~jp)U)Sm&rVOGJG+<Ss@YScpkt(
z>JGQIKm?FN6XphmVW<~_{pasIco{}>Pdmbxz*%H<g3MF{ZK9@<no<bl&P{Ztt{lp^
zngsC`I65xdI~6>)mcm9k7vJ<(;_YJ6Bmh)0v?DfDxg2pBzOgKwb00}{jl-JrBd8z7
z?zo-qBq!UQxjH&r?dC1W*!X|#eRWurTl+2`NJ*%Kf`ow*0$U^m!9g)lKw6MSBnJ@b
zE(J_NK?J0`m2L!76vd%CM0#LIhmP~SZrQ^A`JMCUxz2Sy|L~e|oO#!~)_T`d_kBNT
z-rbik#0U_|<;G#MF4Ny$>5r_#4mqwfG`LQ#EMLBg>Q^xoytMw}JGL4E%lkr(kq;rT
z4C4$+7J$@n2>IPZi2Y>eCj!vx8_KlwE9uT$zR!tXAU!)Z^(LUpjZMv9O|8>Y0wQ`O
zli%SNpGrb@Ex}&@jx(Y6zHy5)&vZPe!+!`Ly2=ZPX*0$JqZ4nBNOw8~{RFJUgcG+g
zE6<1ZfypJOBsP7+Uaz0VW}Ek@c*Ke0-!-|wbOIaC5$-FZNTJV*oBR-9eJvl9j*O_>
zk*6wd$vBy7jIqa?5vClL@>Ll>1f9qKvHd?^c@l6=|NIv(jlB)x`Wl2flon49p8k*-
zoaIE3e?YC!vNuC5TZf$TV(Yb|qFme88mX8vvJ+^G-Z{*OB>!8q2{UKhJM{gEuk@Bv
zX#A^6Mr%koLhy91H`d}mRQ5QK%HH?*zjj_?%L-k*3AMN2X=;oARM{hkiQdQN^B*dE
zpqDHe@9JQ}{O4a+xeC7?vc&mh>vG%6F$6@444m9lO2N$1-M>qFIs4TsT#I?NOIqm7
zfQL1H9(I%$u(-Ouzki&0cksgvB_*Ys_WnopT-#@kf4l_it<^$Nx~E^r$89rx0!!|B
zBlH-4=Vi7gMskSw01;cU=7oE|FSq?wObK9p{<mKz^SnIuqJe#@_`7q>Q}|$}TRE;u
zY+Y~XP1B0P$3dkWf@C{)%GQsh$-rFMT~(s_^Hg5E4j&I$l2ai-tL>xy`>huiTnP>B
zXD#=3UUjQ_6to{ajsJVM?zKAzz4!OfDWvs4YIp&>rY<mv<~(XydH8nSe(1nPIst0A
z8a@aF4$1hg0&y_YX?~=w+{K|AloT^<M!%YVzE5*fBk$vG>O&^DuKe7pd!&9e9A!E=
zw;w`FGUWyMO7|;Kn+zz<B2+jH0M}$DBpf@;&rHTnYvGmR-afWE*1x&o;Nu8_w}n(9
zJ+G_94_zeiG9hbeaj4I@nLGLLMAU1*qI5wU!VI8TwxuecI17n3i_HACYY(e1LR>c~
zjEogc11Uz#iWRn215!*Yhu6F*`Ch{|EoR6!Yps(&qCN?p21lsNYC-{Zzea}V5R`YZ
z3;8CE)i3z8v8;6t;8-K2;>bXhlN4$N(Yk(p%HTr>!o)*VF{~L1B>?LW^v~u=1VPaa
z-Ee&Cqa?b^r@%GQb{<b`0~b8luoJxO#=Q?#9O{wq8~&%Wj~)QMLU2HArym($hl3<b
zQqH{{B&xfh$KfsD&iSq0n?#yIQ9Q_kg6n#Z?SKzs7XmILsKyA9A0jEPqFwSX!)Xcr
z+LA2=ZvnbB?j6l?);<pqF`Y=RlNC$e8Y@M*3f3G4xJY@E5fG`GB6>j~j(8CZkW=aA
z$u7GE?A#Zt9FpZ?$@0-r3dhT0b?E@vA;!42XIdV=0-Q11Dh$F$gf@_uTsD?Rad6z{
zqSnXd7eJXKf$CLK+G95W`M3}%Zf1r3rSDw>C=H7qHlH7Uu|s~&(MXD~i%1BsJrA7{
z-`$G_xuRF)z&2fgVU>;oG_Gr1JOM?viu&d(<eoLV&u3MZAkS)ALT&3f%|8}%I2x$8
zufKo4RSKB740L<e><ja<*$;e0sVo%Y+}ZL#Mv6%#9YsX2u|79dRBlr-3+Dk8v4dJ{
zj9s(9JPhi#Hs@TbbA8`TFzzDJu<Lu&1X6B>P-V6&=WLlz1AZjmDl7*Y8tCOtSeyd=
z=2h^cdD46h;MP7Px3)I`xOz+hu+>F|;}PkQlb^-`9CXAhd@aifmPacv*I~NP9)Jj%
zz_!wE{vodSmH;H?LV0z{I)X7Le%{-<)~tM`k3~)E(XGedX|IC(b*g41dHH&^kXL!~
z^J-ijNb0JnW@Y%o`KyW0F6^^}-{$NpouGDwtJ6{-B3^0_`h_KZvDii@K=^SSlJ9G1
z5a6H^wP;Fs=z`m9R9Hf7E@Lb^%DrQbflrbp*;0u<wg~rg_U#OtcO1_(0Bo>_AX{!J
z^&S9kLsjHXt-p#Ns?_`u9*s_rytVgO->$8*Qj&_2(=&xN1{)j5&E+Y2r%&9$B)|2B
z4F=(SQa<1|2~QIVcJ?nY?`#T*5HP$}?JXT%s%+S@D^Iu5-Lf%JWfZ?!u{I~O*tEHC
zU~$zre-jqwulinxkB9x1dc}`=O1>kG+t7}m7F!QWhPv_W^vU{r8BtmlYHIJKO{oEa
zPMh*+nyFqpYx&$d=E|q~+4&H^fO#9a93_3r{RF&UR==wLnoUVw3X?I|TG>$rB#74A
zgnnx5D3g6wf@pSdHL9VMCDy}Mwx&sFTGjEABQLQPLIY-(;&kZ>4T=0*c-YB3GmqXH
zPHnSG?AU%G=&2bWQDT6=PXKb(4GL}(M9kqyQmg)ASk7*c!$80He}2SMa)d}|Svwx(
zpRgkJS|c5mw$#YgTVnTEmjKoNj*DW2MCu#m%7V@8ulb}E@`8M47ijzqhuc9j@nWZQ
z`ERTSE)-%&*?KQGq1-u~qTAUl;Yj*WB_A?7+#Jc%>$JA)co~pFlotm|KVUXfbtW|S
zTt838>zjb0ML?Z-CJz}-A3p3U*-QVi)<I6}DUkGQ-8Ytv7i$Nhs8X9dYa{n6E=Cvd
zP%(E<i~+w4k>b9FK)!m6D4by;m?Q?`*-!kw3HnpxN>KXfiM5rQXbz#A)-AR+bTx4=
zv00u!c{AsjR&YB!5LbxfX0(rTqpn0(ppg?x)3P{mf6H1emP!o7M+&Vp0Rh=lecbMT
zs1Zz17_!sE7di}*{G0<3{gGMGFU!Tgh4J7^VL9N=ewj@(MM;Sm#CE{V=gNunwq#PX
zSzezXsqY<WN|dQ7-LJFic^haf&HU5=5wPRQSXtS3eV_T#Iv@_NoqH!t1k`eN?tY<L
zVk0jNWR-1AN&;pAZ{Ud~zIpD_f+XbyHuoicqL{rNx{cz*&q&PXN=G1@nUB!VNl^U7
zL;Jb`*D+W>FdBa5QdK2j+Nvr^2(~bvT?`GkE+G9|Dcu2$QD6_$<_H?cl>*ZPrOz11
zu7(P3CM#KY+AcTa7Dwms9HaBA>CRC@&YiGsYZQ<tyexr2@*nrV8LP*@IbmsoP_uQk
z;Q35pMo!0E^G*Q~RbTJumH0_SKLE4>xcuH>godRltVszQwjcHvFMUI3^Pme{%d3+I
zaat!>%0L0no1HY%C9G_RGEKY^WwT0iY*gEw1WYNN037E`9Ji#c+!>My#|htK=ZaO)
zHEy1oyKr%Buke+uMm7Vm_O8v<LE#Zv%0@^#RWAOhXtYueUC(vjw~Rf!qu#&#Om=P}
z5O(m^_E^Vfi6Un?3z#d@2<VHb=^)FKon9*iI>7tqz5<lP2or_#^#HU6$AEeP^M)1z
zA~BsRA4*uihAqBy;pc|NmyUco5pA1Zv54z#7lYuO+8Q(wETg@d$48FwsXbtU-G)Zy
zCm!`K3gUWUNce&n1HK2N1o{lpF1>47&tD|PP$OI0QiSpG*4Ey;1hUK8cjW8s92ze%
zsvcNVi4a2dfsms)CBw<r2pVIU&j37~f?Mw?LX!c|G3;-gD`hjsRX+QHdH9j(fi!0a
z&SesrI>TF}u&*w&*Yyjn_&W^+U74&RWs@DpBWUt<c+O><UPXfU!t!$(kC?6$U#-Fb
zp1%tu7cA&WP#my?jZV=mMlwLmPdYdHYqNWK(={BqfPj<6{g(F2i#M(;PDcmTK=E>F
z4iU+){8+~S1G3~@U>~u<VBYj9&yE`7S5)p~knca<0a8Qzv5K!Rq?cH}uG(Dy#Ml#u
z3;+%u$3csq4u>|u^<Luv+!RLln-L5Y_HdeQHTjAAG_y>E#7B-RJnt968m<7EJi<30
z-NnuucjM;w(eS$a6oZb6Gxr=!^|iRw{z(j8l#)Du5wETQ7^u|Oia*^J$l5?0>j>Hk
zsSDs7ma@iDu%6Vd=)t$BC|DyPJczuC^k?W*SHM9)de@pmzlI#B=kOZd7(NJ^yA!Hu
z^Cs%$?gIcJpy`=$sn4g^F%&Z7m;_f(gB~2Erxnf6#9513#PXF*7t(+)06`74q-$U8
zciI5O4bn{r#fFH@u%(y1+Q6=%Hm-bo<nG%A=zj#`b(hCI>!C8;B;^PJZ-*wvXUjZT
zkUwEVopU!MA>h6?5E6jDw*h|5lyiVtlgD~>Y{EJzi)|p%Z6jniibqUsNgWd<(d$H~
zpvqoNdbH)V3L&L;v_7xSQCc~#6p!|?Qki-7iBj{WnSmsoXlG7S>GY8$86ldxg88QO
zL_RzU%KbtR^gkj+C2BWf6jpZKJp;$Edi-Cy&Aamcnm(%r!SYX=V5cRC>Og-?**(vx
z&;S6TGVLJeP_tPM-XjgLk<2J_Ml!2ZAWmyNj{Kw;2I^%NQ$59P#Wu$1#<)xQ+PmoC
z=>I{2%*MVot^MnZQs*?J2?W{FTK-wZoncoHt=`@Q>7Lz8UcZL0W$S;vU8ZFx2&YlS
zUfepFcSJQ_(tt{Q`pzO|@4vqG>+Nstk{w^$ij5xJ2O*IEUVP_@{G7V~PWyj1+*Zi?
zD|hhUNA3T|Lq_nyMIIC4o!KVIW5kyO4x(vLr>gT4KR(*mtwLuWiqo+BGCfmUS1`zN
z&s}6Aa3Pgif@89Ux*t^=8bHted8)4HQBQh~0l8gph}vvZ$83i_PJ!a<0Tp}6e(lM8
z2h<yq#~W6wJ%=6CmtA#2yT7Own|YM+&K8t-zu~rRpkEY7FIeADzr3zlhPD3ubzH`Y
zEwnvS0oV3o^~6F(=!W`Q;MY-VlK@;H+QG$b(xJfQlRzQv+UnpNs>1$O=1B*?>d_h8
zBNno`w2hAE9)~d<6W{uWT+day*q^X<QXA&s*`HR^I-K@{>5cWJIeY$l9&0uV5lp=v
z5evfD$6*V#P4t$UrP@jAjo<9_GTRLo%~fX=BC{TRJ$T6OpljGSPZ$2;$Sn7M0X0*F
zyXM`Y%}&9E=9Ml7d6+AwW?W9#YG<*A3OTCY@C&Cu8_*dANrBAn&?4<MNX2zRp1Qv8
z$-W&nOtPGQNxzBC#ra+O)JpaFunzy67)9C?E+swbPbBCcS`32Ldyg4fUAgCR%kX=I
z;rVFw-Z$k_{f>H^bT28W-=*>MWKyvOHl2z*X6qi=Xz{|g>!!VX=-D%8W)kk<dIQcC
zvL$b%^%w@IS6R7D*(u0ntk)%%7+}u59GN(IHf2xUu$dj^y&=Bz)<#LNCZU3^+Kg6O
z+_^iGq6Oo+q?nDKjqvs2x<~Jc+{FCsUq-q(@-xqO9*bD%b}hs$Fid#gvU{JU>X*k-
z;$Vm`Sr@}j%e?N-Z&V%nY*K@JF#MRUxi4qJZn1k*CgbRhXn&4K-BF4K=TUV6!5e<)
z>nU=^KYBm<x-RCgi+Q9nr7NAMh)zCMP@3^1{pKD9;{Ep~Ej5RGL=7-^X>c@schaXq
zx(-g-khp~j*r@6F)lIi|-m)80zv&*D|K+2$N5(6xe{s>}l9fCOz3$6tzQAu1r3iM5
zvpy+zYUdfZ<rhe@S3}DBsH=?$)6Sj0<)f&`L7ugBjdROy+Ocv3M0*(^k2(+`)%-`V
zLo&t;A)}03G5zz^AG`*ygslHi_>L;@?@OpcmR3Uub?u+L&F_!kmE?V32wh$pX%O7`
z>t0xT$M1=S7Jbu?(+Zo&+7^DK%ogZ1J|ip0aE=iCJ{WD$_>8_))07zPVs~M^@&j2;
zbAmH~zVldnW%)?{nVFx)Am)}=_moH5$X8v3yGhPgKeCc*)#6TmCLlPByL$1wN=KQ`
z<B6&e6<6|2rwLNc*N#r_!?0D2ezpARq-s?hNPdo35{-#E{U7ci3hrR!;7jbE%Ly?H
zkbrNCk}&=!Sl=<rUgF5pbo#H~X|0{pgF|W+5-wl=L$$#PmM+{_FYP}tq`Hu~cUgFK
zaOZ?<Ij3_c0VR)ltFo2W*l}qBqWlvup9_<sJF0fSy{{pp%o635%KGOYH<O2tah*0s
zI~R8AKYz@HZ6m%*S>n%Iwj~Na?#dng@~<)ddK+jAg$JkINzQ-%@e~R8INul(_UAQW
z$H0ot?~%p-`N#h+FSuYzx$?hejU7b<%6z%SHlrGlVahOVPP!R61XQ7qC9+GzpF{Zz
zo0^nB;ZNVuA-|0Ew9ykJwFSfwn~A=1mkUVCjg(BS;wIn)u%RLFu0AFF_a5ZD4a`4=
zx!%#P-}6Lq;TyvHii<n4cm@(rrz=K~o*p_=K&!4V(keLE4>7j+W0mmDVCCy?3M{(2
zfK9+%pS}ll$_n6W-AL{oNK(09BhPj#USTmrAO@F*M={ZNkRF~$4+;`_de2G7?rI}?
zx<HhCg8g``8Tv6AUtx7h09?hn!=>JFR}S61&jJE%Z83!FTKlXzekPMR9bs8XHW>)E
zHZa_VR|^sL{Y#9X!G}GJw|juJk%kw9-j;<xQsW;KL(&xxKI4p__>zPa7(yZhO`jp4
zXnRPab_|E=>ga1jZ}L1O@ep`e%DnnMc@ku7FReh@zZ-g?7(i`1cO5b!^ld?(&g0GS
zUKl`c2*-W|uEk^lA`6D@D|7liqqZ<YpAG77E5B!yHig3rA%~D0Oh4VQ0?^t{AQ?ZO
zr3bFJwmnd+i`Jk$PI)<WSXn^y1%ZR{&Djd;+2sVin36J-B~SnWL_-D1L4qID#LNIg
zjhO@<)N4p`^f8H8P69C7djm=mToWH{ld)P*GRlXHGG34>be^zb=&i{an(V&^!>NJ{
z&Uv)ha@$jb4SEaJp`He~MN|*|?U$h@20ks5P%Z0+pG#_}&la)9AO$6mHpA4%FGEQ~
z9pS^=l_UMj`0Fd{){P0>%^W1dfb&3T@PQ8X^_{@eD}}x_T9T^g6{zl20&o3u2*${d
z#@Z4ZiQUk8wUqt}b(uhrS#j<NlFFxhB>CM#h;lfs`1-bC(WZzHQHWbY&yWQKV1vd&
z*XYabN&0i{A4{Fl$7xv9zRe*i$jeWe5Pi-`kXe;ZQIqOL@J5C5HA|q<xOsoDtD!6o
z;plW{FH3{AS-Xm2;5<N$@3!s(MZ!=}B!mIrkz>rQqEu^py;q3+HKezgLEns$mYP+(
zgI7GkqW)eOw;WKUhBqN~)2mPArRFXHeC}sy(p_$d%o~gISZq>BUc+++@}?QQY!hH*
zM{+=g@oRl^SlSn)M1o~Wvt8_UIB~nt_0<LwR9Ir}8w@-}sx1}(xNJsP$TP8d+4NTU
z;~Sm)42dDQ^>Y#d2ToOII%d})l{J4;cU2`bFPCi5_O8r)XY4(^$u!F@n!NlQ>6`Oa
zeO4_m3Y_PC@nc^TD}q1U-+8q{;IdZ|h=DIizdU#-yZy8ay!JB_3aR<-dZ#egGK4vH
z`1?76u>j-&pN$>uN{)#f=~od0Ec5qr?q2_E8?zGJh255C3(10<wai*qWt$L{iX2?g
zpg+A4NMKSjrRv}hXaj9c%UxH88ESjVU1GKt6koX;Rw42C5?YGCr~nDDqN~<Aymi5r
z`o}ZHvlF6HbB={DcI18D-BzTyPV>^D=dK01<@Ls9U1g(KW`_Q~mHxsBc{E*!$LU;B
zLSg}il+AwMgUW$>x9kXVFy!o2u$qUWn8T%O*RGjEez~w-xcf_<cZu)qO?Zwi>G5zb
zw9)<eVG5n|PEoaPKsi;e{0RM_9;Dtj0IY^3^Mew1AR&mfA1$O+yi8IjVA^;jUp^IT
zvxY9pC?CHTNZYp0>WxAnkd?my3K2^aLyRzM`JkKH%^RFB7k11kt9T@QHiKdx^ktg1
zoWiuK$xoIS=~WSl<mrr>wq=CzG$&kr?s)T;KBmGLSRu-Fmbdudjs;z?C&T2`sLui{
zp3yb}t+SA_EV|@wD|;R}Z^UD3<Gdq4s^9^$f95=tSaqPd>{pzIVD_5_q4SwtO`TBf
z2tZ4Mv4SB}-H0M!fVr~c`S=A~x0<?^)R5bpR9`-B!lX4V#8ln+(p5XoF(y7;MR)Dz
zG!5L*zWtwNB*YPl=KVNG)JFhru`5rZ85#k;uzEuE>*4(CC_!E6_Vbr$-vO{V<BF{b
zI+5YYN)q&~yKJ7z313Z7I<H8A74ET;YIY3{k5J~xD1=1)VO<N0FVQHJyBYpf2B%K|
zde|}Am1SUXLwzn6v!5^eRdSpK+mq`$Lngy@Az6i)b!JQiTOtX;j47M@4TCF<b~0e;
ziATOhqTSC&0Sz@rYo=K(Qzr<KmWzO52v56B<7oO*W2zI2sW}7RHK_WMXR5DP_}A5?
zXJjAM>T3`CWK%*g29)3l-w)}G7JL%=r$t2|?~mMJ$^_95g3Jkrz=hun#2)D5$#9HG
z%+};G2T+yyO0j9_b;DE`gValY+}EJ=@R04sYd-F)gi_In&Vp`|D>d7qglI@(``&X=
zIo^87$W8LynR6$RtAx^C;n`x+lxQ)@`RtX2W1+@>i_myoK%g>X(*#-Dz_&5PdD7Pn
zZFHPz>n*~#eRjnRka#N%x8AKJat<bn84x|o$glG%ZG6^kU{HU*?v*9SGKD6}=o4D!
zxMaHOp`T;--0<h#IO$K1d6x+|Yx<k#?*3ZD5rScft&-}`6QYwGh@e6`dH@V~RCyhL
zp{AyFLUG>bYq$`P>I;p|Lg+mC4s+IJVRC6gg})kS2_A*`<7@*kgGfn0gw1SCuZdFk
zN|CG5i~IHb6-(8K-%*IcFsF%LdA{qbyJ8~j&=wu{<zYt_q|qY|x<Wj-Gf(vMiT1j{
z5JPHm7?e6$em~(}hlumu2Fb&OJ+{MjwcdH*(spee0KEC&2=&Z%b-jVYZt2?$-%$B5
zfy8Q~`vGXh6Vbw8n-`U;<Vk!$Zu=Vzhd4U+h$xrO=M%TDb*Y1F@%qwW0J)3MlXFhd
zm>8#<_?b@_d2C^>)6|>Z7xuk|yFNdn<6r;@4DBh^v6nZkY*lxwU2#1f#b0s4-d|;P
zur+wNRoQgr!KkXD$H@h)g2Y{herr)j)XeKzba5?}ID^kGvL6gT3=LpZ&qwJ97EdjC
zEsbl(=Gb$LP#3r^<b>x-NyHl%dV4hMJhvux>Qz?as3`L<rb|2B5hdW}Pql=m(X6--
z5-M~!R#ri`?AoSSMNO!B>vdfjahWeRsvH&6p^b%NXC^6wC06Akd&=8ThV3bFvu@5f
z;agV>_(P6X=!PY)6*Txyw0?S#y(jCmJ74H0L)9p&$n-S%o^07S22=g-1ZJH-`*Y-1
zMT^|%i|%gU@8YjE&^0(Qi#Z?W=zB~pl%wdYQ(}QpZjQ`%ayVv@U_<C2$=^@M-yWH{
z7YMh!BRno2b;u(UdueG2n&?Dw%VbXS@bMME)(B%@1Rq7q$6}_oUE4vsxh5IVoU5T(
z{<ee)_>mtro|Yh%Ju5(89ykldso|VMb}fp0o?^E@(jxrBvJOBM7(ihpoN@?hL3e)b
zw(4ITt^wak1YF;kwo}Q`DD^USG>WyZKmPcQqF@jg?zQhGEvGr+=eTVaZ;uXg7&=Vi
zpK86%%*$&L{ILRyFRxh($^?V82>Z<=lkuEhva{Jy{6ss3L{g|0;>4tIW>txcAaITg
zF9iSXn4ofd{J?r${J<#qmUpV}R_<jDM^t%Md+E=?(bfc-;Udj5i?An)%fOYreSyFE
z>ht5p>qv<g2QXOrw%YT5AEFdDV~PU#qtM%<Hm^w%!Np+!h)hsaKp|}bYVk=K7i;nX
zT6x_;*2d>J>9`;Y^rl?+L72Hb>yrmaF8sxKzm*mbmH$L>JXixX(}!$ZXytfp$h5Dl
zcpa4D>{eL;fJ(_vJ$#zmQ4}dKca}QbnXduy^17(jRTUf>DE}8`l!sefRJoiC9Sfg(
z{90UgJ_rr3S5a<v$v&z`dY!L~!GOktF~Tk6y_xflUhVA^u$oR$E#8=xPCPqRRKp}F
zeaN!_umq(BE2H3+7J{Ty`%jVd3fIbx+t1}D!DFyVb^-{&B*YM>cm)#@>~-(Hr=N<o
zAmMlDGCff)D-EJ7i9(h=9eH)DfT%G8MP;3}CE%-@LNms^gQtyN`t~6GG2Bl!Gt!fY
zLi^kOYE*;`vO=GJGK&>b5X*}HYSX*T<xx2=DY1{uX=rcZ>mY9r3f2O^G6{E^pJqJJ
z>oQ$F=!L4+c7m87_-q7`uls>DAUutE`)n@NqUwe}kNP7rE(i?zdMx`CggE0_=c^E;
zPz$a-a1$6lCk;>Mr8I)x*=;2D2|H~Z?)xblgooPz#;=9}XLPzQmC<C5b|skpIBpI4
zzxPDp3cC3Q9*oenHn=4Y8IT|B{P{q^UtnM(wi#GLV(Wcl{gsuep@ZN?O%i!B)n!>w
z8LI`<A^;cge+|dI!?7%>QLzKXBFa=pF{yVCGH5`b0rR1cL0M0+jU}$j)(%wI`ZU&b
zsA=Yu!<OayD?IC(_9i%RCYl!cA(Uevjv!H?QR<zH(2OI&4;_D`R)o@~Q4mdk%Bzw0
z8p5L+>ZBm)gA9z*)+<i0dbTU<Zdt={F{syWjhq-8<mp>3|LRO*o<Om)yczwR`3<*r
ztuTmLbC3vviYuFnY27RVtze~Sp{93ATo(0QcpY(;E;&l@EOkO%j)#luQ-VpIxW>h6
zxD_B7e|PR!kyTETy{o*N(=%OC5RRdrmurHwiQJu<z1uGSdl<ASgG770vIj*1N48n+
zZAMPg8BPem)_QMc$bM%vkXLqE5E8IddI|ovzdIJ>|9Ekr#4J%IEN|O=+b$5O1HbOf
z?7!`wwhxVk(L|xPTYi390Qqx%t%&3$@WBWiOIW{}So6&-9}56-wN-j}oP}^!nL+}T
zQWqd3N)%6u#B~=ILLB)QZno=8AUXo@I0mJ#0Ej9?L0hI~H_{#x@gt$huu=ZqgQl2e
z*jk8;$Fl{;rMKr+Ck+6m^>n(f^|;P?s`4eM6K4V>si0~%a|eWpuNE>8pj|JV0}Fr)
z=m13F&pI_IAay}#Yrw(5!S#7q>A2yg$%`>j<a8msQ=_!^hyDx2sUS)8n33O}FcI3-
z7N_e#)6m{|BoejB)+wI^g_pdM`iK)Lh`0>WtM|U!z*Ddsh<FD>pXx-mjcKm83A`y7
zZ{I}vRK7TQ^A<%g68M6~wG;5^d(GBMTz3HzcC!PJk7FV{t`$Ivz&(@UdpP>({c)Wl
zU!5`5NpSpxK}KnG1Hg)j;`6|GHbMe^q$9))0K-8DS0D$vA;Sm|2B}xK72O@YpHm!s
zHN}&P82g4mLULlZez<Ad)`N`11==&vlToe;gCKEOoyQJ$1J$<IFui>cf!QI65~MvA
z!rvzLhv%)uf(F_!hYMST1A`NQke7?y^2?8Q3s*uIVGqsE)S(uxg}Fx34-Taf(t7iB
zCmdMXfpX=0;{)k_lmcI)N%)e4y<LE2zL_CZ;uS7bm|N_lXn*?wlt8rEHL5Um&-w}}
z+_4*D&h#zOXYR&vEiFUC<RX_emjl;}gULfjg^)0qH3q*tWxW$~Eh)jM{#wr~6Ath!
zhj?c+HyP7)UH9cg<`HczEptE^<${k?RD;@F%e5P8sG_pjgq%wCr6a3ip~C1CJkr?-
zu>pW*J)srh?9&HE(jH(Xose32ZguhzGnOz_{~Hr^Rx0%Z;65Br+q`+mnB~RD^x+&}
z03*bma+oW!xR_ezg;i-9KpLY9!nx_b^pZBC`tbLfDHi2-m<3Jrocpom*}*blj)`TG
zF$Bcw#Hj?90-+CfJlrj6UI+peykLv?r^4KfYb`<!(AojkT|%?iI?)Z_nRj>hYg$EE
z6g-@Myz<z*JEQnGbW2thxxWIvcex#$-1{NbrIp=aeeg(ul*A!=VhKRz?)Q#1Gb#`k
zax!Jg>Pwbq1IBj%U^mKgtDDh$nu!Rb`S4Pz8xut(EK>>B9>gBwthqu3Do$d^=)57x
zd7<dD*3lhFPz8Asx;-n9WWd2(UU%}L@LwsGY}Y_%Iz}NV+;$+_yJEE&hl4F5P%C3c
z8JAS8ia?`sn9l<Cy<N?kkVAs6<`q#15F@dyXtT$4bwXyVPDn^I!kT{Qu3w)TwBINI
z9|oi2JkEi506xa6kpAhNogtDRAVQj3ckki<79ta76<KyUrUIcQOgp@s*omgczVT#u
zps0t)4RK{FrJJLxq9K{r5xwyv*^#y1z-qiHG3!dC??IlTq+$jdkcV3GK0Qmc*D=T<
z0yefwJJ#x($g;rN8=y4tEi2qE5ONrK>$OYU#j<E%ci|&9>%pA-t9;QI9N5$rl9vHz
zr_Qu97XmQLzF(;~7kK_{)c!{mqLJKsa_Oi1f1Rc8NDvYkG>nYxzL9;NM8<87@HDRd
z$W<&=YF`XR0n)_SvabQ6qSeVD$FBwT&<Dl=cVk9GdyZ0Kxomu{bz}oPxOYj{Be5fB
zRVy*oy0+u0EX!T(6>csf#>PqbRhOWtio_D*;^0qt5<^lk=K$D|N#hWN{>K#leI!WJ
zHfRO8JC^zGK@&1X)16eZQW*#9UjSl4FK@`Sr_CZzP)*OvH6yJC@=}BGf!3zEBLu2*
zgt;p1h`3R4FWAm&c4zjjBGOT%_OAeA!?o*SamU5))k=aFY+AX=vV*58IQn5q63z&f
z(x$Y7>obAJ4S(NxCROimaD4&pS-h$G>l@APkgVHBo42oQE@gBS>10<QArim_mbxzc
z@NBr6(qpnqhliU+z*Lkbp(#9VKe3?!bXRgRzeHE!Ed5JMATdPRHeKcFe)d?d?$MFg
zxkW`C8U^O#8sDeotaI1gJuuc>&$Ui;leIX2p~bIL+qmtpg_4BiCi(|RP^~ox@)x@C
zWZwj7u6@LGe<8#z1l>(u^W7~8!OyqG;%4e*Hyka*N1Kd*LKQCIV&8Rq)!>|ZxUc~)
z%S<_n1{g>pSI}ujL${u%MWsxpPyW3!59{zv)Ik3SYb(cTL%q#Y*r+ToCC^W#93-gd
z3sedYB0pHf#kT*_LH6Sb$Lea|UNnl6-hCX}G9C*%rBE1nJ7JH)?m5NFoQ6kYZ&yD_
zSage49kpdyj4JIa{Hz0}V<;+jQDaqnqHt;q+SKi3m(*(_`Tn(@OyVf}qbuD8#!G!O
zxp@!B&d0xx|B;{97k)NYUHm)h#>9QoL7eQlwxv(vtO3KdtEhOEGgf*RP_5|Gv5)2E
zzRk8-<O!h#jGiSA%n@sFuoGCcT66jm5E2oS(XvZ=A}_+pv?^Rhr~3e^x@bE$^vM>F
zZC&g5VbborvF0RZGa_4JHBfDmJpTq}dAL$aO3Gki`T2?PmNLC6Z>?LCGpwb(iv^At
zQ{pnnmB{ei$QYAYZ?CMKAgyp;5C5J|40v&t+^FNi209reC&VTe;Pg~XA3vm<Z>GBF
z`utRH)j8UiL_|alT(3VSN&g(t-$U~{&rsyDX!(fB`c~*7Fd-5k#r`953aS9j@X$2}
z9lg>t(Vp!wu{z$I{Nu-Xv#To%yfSi=nKZoRP?i4u+qxC5SM^)EE&IyCM4U7LNc(8f
zq##qmeeF!Bpw5}@cS#mDEH^RW5X`s37@vy2Le=_SiFD}7(ygp;srdAAN{@|9mXMHe
zj86Tw^0B`s>*vtk#!$GfF7TpQS*jY~;?2-P>O&;9#BM@Ke6IK4Y3t-g$4N2!Cjl%i
zz-9_Ra>!f?tuYUgsLe9;^{FfEbr_k0$+q2CTVBIM^cY!_xzbG@8_KKMR+yC}?8d*s
zC=Y`c1+I_aooadqfdKQ-uaPorD0f+z7bmYq2wPyI%mk9oc90S&OMZ*?*u0~x7qw&R
z2#I!a(t07bZthExd~~0KFAStE*2uQRW7EA0nk7-6AFYoFU6|~AerpZ<yyggzJ58`I
z&2XqIjrIn)@vr$^7fxM<xptC5+B76ydd!drACA-xWb0N&-OD3$GWhx?NTIuHZF#=g
zkA|%YgaMnt!C!gN{w`^n<izZ4$J(Up&_78ULaYP$&^Y-oqHT9-{aKdjNdTQ(xB2|E
z@mJZ!sa_d$TgI}X`SVkiT5lPNcAaD=VEOi5NNC&sB->N+7Q7Z~P89(R*#oE+wc%EV
zK0QZTTF?#XB||~xtVMr?gf^sOLIQk5QbnhpNdP^MiIw&GtbdCMNa#tgE>6qCUC8Ac
z)njEVg**nQddOD3PxS!NZ20VvTAr+j^^WbYp3HL)nygJh;inmjP|e+icS;T4^}dFj
z*p~s*#cvO2<UTw(Fm=?+i=Nh#da^0}!1kwzEST(_k>nf222he~21=HulgRC9@$i<l
z!P<Zy@ZPl@{lx|n^yQ%}52oeJyToTZQn)uz2J!Jel#C4P9>NEkjdtbleVO7(U>oQC
zVgIj(^Pi^_UNmTUEtC(%=c`a0-L@IGp5+)}LP8S@o+O>^Q@8buKO!NHd9Nb2?+?GK
z5aCIXS3s}KzvFIxeepd!m$$fC$hUrH+i+6kBndRSAEkzF4Pg7Ddke;LD#ZWyA@lo-
zEyvjjGA-V>0ITbNTv753jOFFED}TBLv;DN5Z3VZm9DllZbr_3m;?e){gK&v_axzJ@
z<|+Ii?;W-qZ&Fjx{~QT{R~IpHOn&A|=Re&$Cj+e~gYx+H(`xI|k`^#!T{mRQ|8(zk
slw^{RQ*F1OC0m#N@ACb>uzYKznZyiclXB7v1n}qTrRx{7F6epxF9z#UW&i*H

literal 0
HcmV?d00001

diff --git a/docs/search.json b/docs/search.json
index ec43460..aa26c84 100644
--- a/docs/search.json
+++ b/docs/search.json
@@ -74,21 +74,21 @@
     "href": "start-data-def.html#the-birth-of-a-dataset",
     "title": "2  Defining “Data”",
     "section": "2.1 The birth of a dataset",
-    "text": "2.1 The birth of a dataset\nIn “The Art of Access”, David Cuillier and Charles N. Davis describe a process of tracking down the life and times of a dataset. Their purpose is to make sure they know how to request it from a government agency. The same idea applies to using data that we acquire elsewhere.\nAs reporters, we usually deal with data that was created in the process of doing something else – conducting an inspection or paying a parking ticket. These adminsitrative datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. In the sciences, those are often called “administrative records”.\nAnother type of data might be considered “digital trace” data, which often refers to social media posts, online publications and other items that are born in electronic form and posted publicly for anyone to view and use.\nFinally, there are datasets that are compiled or collected specifically for the purpose of studying something. They might be collected in the form of a survey or a poll, or they might be monitoring systems such as pollution or weather stations. In this instance, the information has intrinsic value AS information."
+    "text": "2.1 The birth of a dataset\nIn “The Art of Access”, David Cuillier and Charles N. Davis describe a process of tracking down the life and times of a dataset. Their purpose is to make sure they know how to request it from a government agency. The same idea applies to using data that we acquire elsewhere.\nAs reporters, we usually deal with data that was created in the process of doing something else such as conducting an inspection or paying a parking ticket. These datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. These are often called “administrative records”.\nAnother type of data might be considered “digital trace” data, which often refers to social media posts, online publications and other items that are born in electronic form and posted publicly for anyone to view and use.\nFinally, there are datasets that are compiled or collected specifically for the purpose of studying something. They might be collected in the form of a survey or a poll, or they might be monitoring systems such as pollution or weather stations. In this instance, the information has intrinsic value AS information."
   },
   {
     "objectID": "start-data-def.html#granular-and-aggregated-data",
     "href": "start-data-def.html#granular-and-aggregated-data",
     "title": "2  Defining “Data”",
     "section": "2.2 Granular and aggregated data",
-    "text": "2.2 Granular and aggregated data\nOne of the hardest concepts for a lot of new data journalists is the idea of granularity of your data source. There are a lot of ways to think about this: individual items in a list vs. figures in a table; original records vs. compilations; granular data vs. statistics.\nGenerally, an investigative reporter is interested in getting data that is as close as possible to the most granular information that exists, at least on computer files. Here’s an example , which might give you a little intuition about why it’s so important to think this way:\n\nExample: Death certificates\nWhen someone dies in the US, a standard death certificate is filled out by a series of officials - the attending physician, the institution where they died and even the funeral direcor.\n\n\n\ndeath certificate\n\n\nHere is a blank version of the standard US death certificate form – notice the detail and the detailed instructions on how it is supposed to be filled out.\nA good reporter could imagine many stories coming out of these little boxes. Limiting yourself to just to COVID-19-related stories: You could profile the local doctor who signed the most COVID-19-related death certificates in their city, or examine the number of deaths that had COVID as a contributing, but not underlying or immediate, cause of death. Or maybe you would want to map the deaths to find the block in your town most devastated by the virus.\n\nEarly in the pandemic, Coulter Jones and Jon Kamp of the Wall Street Journal examined the records from one of the few states that makes them public, and concluded that “Coronavirus Deaths were Likely Missed in Michigan, Death Certificates Suggest”\nBut you probably can’t do that. The reason is that, in most states, death certificates are not public records and are treated as secrets. 2. Instead, state and local governments provide limited statistics related to the deaths, usually by county, with no detail.\nHere’s an example from Arizona — note that we can only see statistics in the way the data source has decided we want to examine them, without access to the underlying information. There’s no way to look at age and race and gender combined for each county, just the generalized statistics for each category alone.\nHere are some of the typical (not universal) characteristics of granular vs. aggregated data:\n\n\n\n\n\n\n\nGranular\nAggregate\n\n\n\n\nIntended for some purpose other than your work\nIntended to be presented as is to the public\n\n\nMany rows (records), few columns (variables)\nMany columns (variables), few rows (records)\n\n\nRequires a good understanding of the source\nExplanatory notes usually come with the data\n\n\nEasy to cross-reference and compile\nOften impossible to repurpose\n\n\nHas few numeric columns\nMay be almost entirely numerical\n\n\nIs intended for use in a database\nIs intended for use in a spreadsheet\n\n\n\nWe often have to consider the trade-offs. Granular data with the detail we need - especially when it involves personally identifiable information like names and addresses - can take months or years of negotiation over public records requests, even when the law allows it. It’s often much easier to convince an agency to provide summarized or incomplete data. Don’t balk at using it if it works for you. But you lose flexibility and have to live with the statistics compiiled for you if you have to settle..\nFor our purposes, it’s important to remember that we can always create aggregated numbers like the ones shown in the Arizona COVID page out of individual items, but you can never de-aggregate statistics into more granular data like the boxes filled out in death certificates."
+    "text": "2.2 Granular and aggregated data\nOne of the hardest concepts for a lot of new data journalists is the idea of granularity of your data source. There are a lot of ways to think about this: individual items in a list vs. figures in a table; original records vs. compilations; granular data vs. statistics.\nGenerally, an investigative reporter is interested in getting data that is as close as possible to the most granular information that exists, at least on computer files. Here’s an example , which might give you a little intuition about why it’s so important to think this way:\n\nExample: Death certificates\nWhen someone dies in the US, a standard death certificate is filled out by a series of officials - the attending physician, the institution where they died and even the funeral direcor.\n\n\n\ndeath certificate\n\n\nHere is a blank version of the standard US death certificate form – notice the detail and the detailed instructions on how it is supposed to be filled out.\nA good reporter could imagine many stories coming out of these little boxes. Limiting yourself to just to COVID-19-related stories: You could profile the local doctor who signed the most COVID-19-related death certificates in their city, or examine the number of deaths that had COVID as a contributing, but not underlying or immediate, cause of death. Or maybe you would want to map the deaths to find the block in your town most devastated by the virus.\nEarly in the pandemic, Coulter Jones and Jon Kamp of the Wall Street Journal examined the records from one of the few states that makes them public, and concluded that “Coronavirus Deaths were Likely Missed in Michigan, Death Certificates Suggest”\n\nBut you probably can’t do that. The reason is that, in most states, death certificates are not public records and are treated as secrets. 2. Instead, state and local governments provide limited statistics related to the deaths, usually by county, with no detail.\nHere’s an example from Arizona — note that we can only see statistics in the way the data source has decided we want to examine them, without access to the underlying information. There’s no way to look at age and race and gender combined for each county, just the generalized statistics for each category alone.\nHere are some of the typical (not universal) characteristics of granular vs. aggregated data:\n\n\n\n\n\n\n\nGranular\nAggregate\n\n\n\n\nIntended for some purpose other than your work\nIntended to be presented as is to the public\n\n\nMany rows (records), few columns (variables)\nMany columns (variables), few rows (records)\n\n\nRequires a good understanding of the source\nExplanatory notes usually come with the data\n\n\nEasy to cross-reference and compile\nOften impossible to repurpose\n\n\nHas few numeric columns\nMay be almost entirely numerical\n\n\nIs intended for use in a database\nIs intended for use in a spreadsheet\n\n\n\nWe often have to consider the trade-offs. Granular data with the detail we need - especially when it involves personally identifiable information like names and addresses - can take months or years of negotiation over public records requests, even when the law allows it. It’s often much easier to convince an agency to provide summarized or incomplete data. Don’t balk at using it if it works for you. But you lose flexibility and have to live with the statistics compiiled for you if you have to settle..\nFor our purposes, it’s important to remember that we can always create aggregated numbers like the ones shown in the Arizona COVID page out of individual items, but you can never de-aggregate statistics into more granular data like the boxes filled out in death certificates."
   },
   {
-    "objectID": "start-data-def.html#nouns",
-    "href": "start-data-def.html#nouns",
+    "objectID": "start-data-def.html#unit-of-analysis-know-your-nouns",
+    "href": "start-data-def.html#unit-of-analysis-know-your-nouns",
     "title": "2  Defining “Data”",
-    "section": "2.3 Nouns",
-    "text": "2.3 Nouns\nThat brings us to one of the most important things you must find out about any data you begin to analyze: What noun describes each row? In statistics, these rows might be called observations or cases. In data science, they’re usually called records. Either way, every row must represent the same thing – a person, a place, a year, a water sample or a school. And you can’t really do anything with it until you figure out what that noun is.\nSometimes the only noun you can obtain is problematic. In 2015, we did a story at The New York Times called “More Deportation Follow Minor Crimes, Records Show” . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions.\nIn writing the piece, we had to work around a problem in our data: the agency refused to provide us anything that would help us distinguish one individual from another. All we knew was that each row represented one deportation, not a person. Without a column that uniquely identified people – say, name and date of birth, or some scrambled version of an their DHS number – we had no way to even estimate how often people were deported multiple times. If you read the story, you’ll see the very careful wording, except when we had reported out and spoken to people on the ground."
+    "section": "2.3 Unit of analysis: Know your nouns",
+    "text": "2.3 Unit of analysis: Know your nouns\nThat brings us to one of the most important things you must find out about any data you begin to analyze: What noun describes each row? In statistics, these rows might be called observations or cases. In data science, they’re usually called records. Either way, every row must represent the same thing – a person, a place, a year, a water sample or a school. And you can’t really do anything with it until you figure out what that noun is.\nSometimes the only level you can obtain creates problems. In 2015, we did a story at The New York Times called “More Deportation Follow Minor Crimes, Records Show” . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions.\nIn writing the piece, we had to work around a problem in our data: the agency refused to provide us anything that would help us distinguish one individual from another. All we knew was that each row represented one deportation, not a person. Without a column that uniquely identified people – say, name and date of birth, or some scrambled version of an their DHS number – we had no way to even estimate how often people were deported multiple times. If you read the story, you’ll see the very careful wording, except when we had reported out and spoken to people on the ground."
   },
   {
     "objectID": "start-data-def.html#further-reading",
@@ -116,14 +116,14 @@
     "href": "start-data-document.html#scripted-journalism",
     "title": "3  Documenting your work",
     "section": "3.1 Scripted journalism",
-    "text": "3.1 Scripted journalism\n \nChristian McDonald at the University of Texas has described “scripted journalism” as data reporting that is “repeatable, transparent and annotated”.\nWe will be accomplishing those goals through the use of Quarto documents, which combine writing with analysis and results. The next chapter gets you started in Quarto. But by the end of this book, you will know how to produce a site like McDonald’s Major League Soccer analysis on your own as a way to document your work.\nSome news organizations and editors talk instead about the “data diary”. It’s the same idea, but often done a little less formally. Sometimes you’ll do work that doesn’t fit into the scripted journalism idea, such as a quick spreadsheet sort or some hand-cleaning of data. When you do that, be sure to keep a running list of all of the steps you took. Unlike programming languages, there’s no reminder of your mouse clicks.\nI generally work on two levels: For stories, I have a Google or other document that lists all of the datasets I’ve attempted to use to answer my questions, where I tried to get them, and what experts I interview say about them. I also list stories and studies that used the data in these documents. But I almost exclusively work in programming languages to test, clean and analyze data. And to the extent possible, I try to share any results using Quarto documents so that I always know everyone is working from the same version.\nThink of the data work the same way you think about interview notes or transcripts and any other research for a story. You wouldn’t quote a court case without reading it and probably talking to some of the participants. You’d make sure you know where to find the documents and what people say about them. You will consult those documents during your fact-checking. All data work – even the most short-lived – should be documented in at least the same detail. Ideally, someone reading through your notes would be able to repeat your work and understand what it means."
+    "text": "3.1 Scripted journalism\n \nChristian McDonald at the University of Texas has described “scripted journalism” as data reporting that is “repeatable, transparent and annotated”.\nWe will be accomplishing those goals through the use of Quarto documents, which combine writing with analysis and results. By the end of this book, you will know how to produce a site like McDonald’s Major League Soccer analysis on your own as a way to document your work.\nSome news organizations and editors talk instead about the “data diary”. It’s the same idea, but accomplished by keeping the journalism equivalent of the lab notebooks common in the sciences. Sometimes you’ll do work that doesn’t fit into the scripted journalism idea, such as a quick spreadsheet sort or some hand-cleaning of data. When you do that, be sure to keep a running list of all of the steps you took. Unlike programming languages, there’s no reminder of your mouse clicks.\nI generally work on two levels: For stories, I have a Google or other document that lists all of the datasets I’ve attempted to use to answer my questions, where I tried to get them, and what experts I interview say about them. I also list stories and studies that used the data in these documents. But I almost exclusively work in programming languages to test, clean and analyze data. And to the extent possible, I try to share any results using Quarto documents so that I always know everyone is working from the same version.\nThink of the data work the same way you think about interview notes or transcripts and any other research for a story. You wouldn’t quote a court case without reading it and probably talking to some of the participants. You’d make sure you know where to find the documents and what people say about them. You will consult those documents during your fact-checking. All data work – even the most short-lived – should be documented in at least the same detail. Ideally, someone reading through your notes would be able to repeat your work and understand what it means."
   },
   {
     "objectID": "start-data-document.html#elements-of-your-documentation",
     "href": "start-data-document.html#elements-of-your-documentation",
     "title": "3  Documenting your work",
     "section": "3.2 Elements of your documentation",
-    "text": "3.2 Elements of your documentation\n\nData sourcing\nMake sure to write a document that details where you got your data, and any material you reviewed about it. This would include:\n\nThe source of YOUR data, and how you know it’s authentic. Be specific. And don’t pretend you got it from the original source when you found it elsewhere, such as in this textbook or in a Github repository.\nDescribe the original source of the data and how it is collected and released.1\nTake notes on other stories and studies that use this or similar data. Include interview notes, advice, warnings and findings along with stories that have already been done.\nIdentify alternative sources for this and similar or related datasets or documents.\nSpecifically write down where you have stored all of this and how you have organized your work. You want to make sure you can get back to the latest version easily, and that you have all of the supporting documents you need to check it.\n\n\n\nData documentation and flaws\nI usually put the details of data documentation in a data cleaning Quarto document, which includes the code and the links to my files in one place. But no matter where you do it:\n\nInclude links or copies of any original documentation such as a record layout, data dictionary2 or manual. If there isn’t one, consider making a data dictionary with what you’ve learned.\nDocument the ways you checked the integrity of the data. There are many ways it might be inaccurate. Try to reconcile the number of rows and any totals you can produce to match other reports created by the source, or other reports that have used it. On longer stories, you’ll also check for impossible combinations (10-year-olds with DUIs), missing data and invalid data such as corrupted date columns.\nRecord any questions (and answers as you get them) about the meaning of fields or the scope of the data.\nDocument decisions you’ve made about the scope or method of your analysis. For example, if you want to look at “serious” crimes, describe how and why you categorized each crime as “serious” or “not serious.” Some of these should be vetted by experts or should be verified by documenting industry standards. You might note whether you’ve agreed as a team on these decisions, and when.\nInclude a list of interviews conducted / questions asked of officials and what they said.\n\n\n\nProcessing notes\nSome projects require many steps to get to a dataset that can be analyzed. You may have had to scrape the data, combine it with other sources or fix some entries. Some common elements you should document:\n\nHand-made corrections. Try to list every one, but it’s ok if you describe HOW you did it, such as clustering and hand-entering using OpenRefine. Link to any spreadsheet, document or program you used. Just be sure to always work on a copy of the data.\nGeocoding (affixing geographic coordinates to addresses). Note how many were correct, how many missing, and what you did about it.\nA description of how you got messy data into a tabular form or a form suitable for analysis. For example, you may have had to strip headings or flip a spreadsheet on its head. Make sure to write down how you did that.\n\n\n\nThe good part: Your analysis\n\nEach question you asked of your data, and the steps you took to answer it. If you use programming notebooks, write it out in plain language before or after the query or statements.\nVetting of your answers: who has looked them over, commented on them\nWhy they might be wrong!"
+    "text": "3.2 Elements of your documentation\nYou may not document all of these for smaller stories, but try to keep track of as many of these elements as possible.\n\nData sourcing\nMake sure to write out, in detail, where you got your data and any material you reviewed about it. This includes:\n\nThe source of YOUR data, and how you know it’s authentic. Be specific. Don’t pretend you got it from the original source when you found it elsewhere, such as in this textbook or in a Github repository.\nDescribe the ORIGINAL source of the data and how it is collected and released.1\nTake notes on other stories and studies that use this or similar data. Include interview notes, advice, warnings and findings along with stories that have already been done.\nIdentify alternative sources for this and similar or related datasets or documents.\nSpecifically write down where you have stored all of this and how you have organized your work. You want to make sure you can get back to the latest version easily, and that you have all of the supporting documents you need to check it.\n\n\n\nData documentation and flaws\nI usually put the details of data documentation in a data cleaning Quarto document, which includes the code and the links to my files in one place. But no matter where you do it:\n\nInclude links or copies of any original documentation such as a record layout, data dictionary2 or manual. If there isn’t one, consider making a data dictionary with what you’ve learned.\nDocument the ways you checked the integrity of the data. There are many ways it might be inaccurate. Try to reconcile the number of rows and any totals you can produce to match other reports created by the source, or other reports that have used it. On longer stories, you’ll also check for impossible combinations (10-year-olds with DUIs), missing data and invalid data such as corrupted date columns.\nRecord any questions (and answers as you get them) about the meaning of fields or the scope of the data.\nDocument decisions you’ve made about the scope or method of your analysis. For example, if you want to look at “serious” crimes, describe how and why you categorized each crime as “serious” or “not serious.” Some of these should be vetted by experts or should be verified by documenting industry standards. You might note whether you’ve agreed as a team on these decisions, and when.\nInclude a list of interviews conducted / questions asked of officials and what they said.\n\n\n\nProcessing notes\nSome projects require many steps to get to a dataset that can be analyzed. You may have had to scrape the data, combine it with other sources or fix some entries. Some common elements you should document:\n\nHand-made corrections. Try to list every one, but it’s ok if you describe HOW you did it, such as clustering and hand-entering using OpenRefine. Link to any spreadsheet, document or program you used. Just be sure to always work on a copy of the data.\nGeocoding (affixing geographic coordinates to addresses). Note how many were correct, how many missing, and what you did about it.\nA description of how you got messy data into a tabular form or a form suitable for analysis. For example, you may have had to strip headings or flip a spreadsheet on its head. Make sure to write down how you did that if it’s not in a program that can be easily replicated.\n\n\n\nThe good part: Your analysis\n\nEach question you asked of your data, and the steps you took to answer it. If you use programming notebooks, write it out in plain language before or after the query or statements.\nVetting of your answers: who has looked them over, commented on them\nWhy they might be wrong!"
   },
   {
     "objectID": "start-data-document.html#further-reading",
@@ -539,11 +539,11 @@
     "text": "14.5 A note on NA, or missing data\n\n\n\n\n\n\nMissing values never match anything!\n\n\n\nA filter will never return any rows that are missing, or the special NA value, in the column you test – it is a void that is never equal or not equal to any other value.\n\n\nIn most datasets, there is at least a little missing data. In this case, the forgiveness amount never shows the value zero. Instead, it shows NA, or the R symbol for missing data.\nThis isn’t the same as the letters “NA”. It’s a special type of value that means “nothing”. It’s not zero, it’s not higher or lower than any other value, it doesn’t ever equal anything and it doesn’t ever NOT equal anything! It stands for a value that is truly unknown. In fact, an NA doesn’t even equal another NA, because we don’t know anything about either one.\nThis means you can’t filter for NA values directly and often have to take them into account.\nTo filter for NA values, you’ll use a special function — is.na() — designed to pick out these bad actors:\n\nppp_small |&gt; \n  filter ( is.na ( forgiveness_amount ))\n\nWe’ll come back to this later, but for now just remember that missing data is a scourge on your analysis, and it will almost always be a problem you have to solve."
   },
   {
-    "objectID": "r-verb-filter.html#your-turn",
-    "href": "r-verb-filter.html#your-turn",
+    "objectID": "r-verb-filter.html#footnotes",
+    "href": "r-verb-filter.html#footnotes",
     "title": "14  Verbs in depth: Select, arrange, filter",
-    "section": "14.6 Your turn",
-    "text": "14.6 Your turn\n\n\nOpen or re-open your project\nCreate a new Quarto document with the standard front matter (YAML) and libraries\nLoad the PPP data into a new object called ppp_orig (code below)\n\n\n```{r}\n\nppp_orig &lt;- readRDS( url ( \"https://cronkitedata.s3.amazonaws.com/rdata/ppp_az_loans.RDS\"))\n\n```\nNext, add a code chunk with the glimpse(ppp_orig) command to get a list of the columns in order.\n\nFiltering your data\nCreate code chunks that will filter for the following values. You can select just the columns you want to see in whatever sort order makes sense for you to troubleshoot. You can rewrite the list below into plain English sentences to introduce the code chunks.\nFind the following sets of loans, and try to think about why these might make for interesting or possible story ideas.\n\nAt least $1 million\nUnder $1,000\nUnder $1,000 for projects in Maricopa County\nLoans that were paid in full but not forgiven. In other words, the borrower decided to pay it back rather than ask for taxpayers to cover it. (These loans are shown as missing, or NA in the forgiveness_amount column)\nLoans in Phoenix, Scottsdale and Tempe\nLoans in Arizona that were for projects out of state.\nMcDonald’s franchises. (I removed the apostrophe.)\nAny loan to a Native American tribe or nation or reservation. Hint: It’s hard to do this because “NATION” could be part of the word “NATIONAL” and “TRIBE” could be part of the word “TRIBECA” and “RESERVATION” could be part of the word “PRESERVATION”. See how far you can get using your wildcards.\n\n\n\nYour questions\nCreate a new section in your document, and list three additional filters that might yield something interesting. Before each attempt you make, write a sentence that describes, in plain English, what you are trying to find. Then after running your attempt, write a sentence summing up what you see in the result. Was it what you expected? Anything interesting?"
+    "section": "",
+    "text": "That’s why I’ve removed all of the punctuation from your the borrower names and addresses in your data file, and turned all of the names into upper case.↩︎"
   },
   {
     "objectID": "r-verb-mutate.html#mutate-to-create-new-columns",
@@ -553,11 +553,18 @@
     "text": "15.1 mutate to create new columns\nUse the verb mutate whenever you want to create or change existing columns in your data.\nExamples of this include:\n\nComputing difference between numbers in two columns\nReplacing NA values with “Unknown” or zero.\nCollapsing or creating categories for more meaningful analysis\n\nYou will often use a combination of filtering and mutating to create a new data frame using the &lt;- assignment to use in future code chunks. That’s because they can get complex, and you don’t want to repeat code that you might have to change over and over.\n\n\n\n\n\n\nWarning\n\n\n\nThis chapter assumes you have gone through the filtering chapter. Many of the expressions and functions are the same, so if they don’t look familiar to you, consider reviewing that chapter (again) first.\n\n\nmutate uses the syntax:\n    mutate ( new_column_name = function (arguments))\nThat’s pretty abstract, so we’ll do this using examples."
   },
   {
-    "objectID": "r-verb-mutate.html#categorizing-text-using-conditional-statements",
-    "href": "r-verb-mutate.html#categorizing-text-using-conditional-statements",
+    "objectID": "r-verb-mutate.html#math-on-columns",
+    "href": "r-verb-mutate.html#math-on-columns",
+    "title": "15  Verbs in depth: New from old data with Mutate",
+    "section": "15.2 Math on columns",
+    "text": "15.2 Math on columns\nHere’s an example of computing the portion of the original amount that was forgiven.\nUse the single “=” sign to provide a name for the new column and create more than one new column using a comma between them:\n\nppp_orig |&gt;\n1    select ( borrower_name, borrower_city, amount, forgiveness_amount) |&gt;\n    mutate ( forgiven_pct = \n2               ( forgiveness_amount  /  amount )  * 100\n            )   |&gt;                   \n3  sample_n (50)\n\n\n1\n\nPick out a few columns to work with\n\n2\n\nCompute a new column using math\n\n3\n\nPrint out a random sample of 50 rows\n\n\n\n\n\n\n\n\n\n\nNote\n\n\n\nA value over 100 means that the government also waived fees and interest; a value under 100 means that less than all of the loan was forgiven.\n\n\n\n\n\n\n\n\n\n(I did some fancy formatting so you could read the results more clearly. We’ll come back to that in a later lesson.)\n\nConverting NA to 0\nIf you look through the pages, you’ll realize that there are a lot of rows with NA instead of a value. That’s because whenever you do anything with a missing value, the result is missing. Missing values infect everything they touch. You often have to deal with them before you can do anything else.\nWe would like to convert the forgiven amount from a missing value to zero, under the idea that if they have not filled it out, nothing has (yet) been forgiven. Of course, we’d have to check that with the SBA before publication.\nThere is a specific function used for that: replace_na(), with two arguments: The column you want to check for missing values, and the value you want to use instead. This isn’t limited to numbers – you can do the same thing with words, often replacing NA with “Unknown”.\n\nppp_forgiven_fixed &lt;- \n  ppp_orig |&gt;\n  mutate (amount_forgiven = replace_na(forgiveness_amount, 0))\n\n(Note that nothing came out in this code chunk because the result was saved into a new data frame)\n\n\nDetail and total with summary statistics\nYou can use mutate to put summary statistics next to your values so you can see whether or not they are similar to the average or median.1\n\nppp_forgiven_fixed |&gt; \n  mutate ( avg_forgiven = mean(amount_forgiven), \n           median_forgiven = median(amount_forgiven)) |&gt; \n  select (borrower_name, amount_forgiven, avg_forgiven, median_forgiven ) |&gt; \n  head()"
+  },
+  {
+    "objectID": "r-verb-mutate.html#creating-categories-for-easier-analysis",
+    "href": "r-verb-mutate.html#creating-categories-for-easier-analysis",
     "title": "15  Verbs in depth: New from old data with Mutate",
-    "section": "15.2 Categorizing text using conditional statements",
-    "text": "15.2 Categorizing text using conditional statements\nVery often, you’ll want to categorize entries in a database in order to make it simpler to count and sum the values in a meaningful way. For example, the business_type column has 24 different values, including NA . Here’s a list with the number of loans in each category:\n\n\n\n\n  \n\n\n\nOne way to work with these is to create new columns with yes-no indicators for certain types of businesses like non-profits or individuals vs. companies.\n\nTwo categories using if_else\nThe function to do this is if_else() , which tests a condition exactly the same way filter did, but then assigns a value based on whether it’s met or not. You list the condition first, then a comma , then what should happen if the condition is true, and then what should happen if it’s false.\nThe condition is done the same way you did it in a filter, but instead of picking out the rows that match, mutate acts on them separately from the rows that don’t match.\nThere’s no “maybe”, except for NA’s in the original value. Here’s the general form of what it looks like:\n\nnew_column_name = if_else ( test the old column for something as in a a filter,\n                         give it a value if it's true,\n                         give it another value if it's not true)\nSo here is a way to do this with the business_type using the same %in% operator you used in the filter lesson, saving it to new data frame in your Environment, then displaying the first 10 types of businesses using count()\n\nppp_category_indiv &lt;- \n  ppp_forgiven_fixed |&gt;\n  mutate ( is_individual = \n              if_else ( business_type %in% \n                          c(\"Independent Contractors\", \n                            \"Sole Proprietorship\", \n                            \"Self-Employed Individuals\", \n                            \"Single Member LLC\"), \n                 \"Individual\", \n                 \"Organization\")\n  )  \n\nNow print off a sampling of a some of the rows to take a look:\n\nppp_category_indiv |&gt; \n  select ( is_individual, business_type) |&gt; \n  sample_n (50) \n\n\n\n  \n\n\n\n\n\nTwo categories using fuzzy matching\nYou can also use the same str_like() function you used in filtering. Here, it sets whether or not the borrower was a non-profit. Remember, a % means “anything or nothing, and a _ means”any single character but it must be present”. In our case, the word is shown as “Non Profit” and “Non-Profit”. Once. you’ve set it to your new words, you can just filter on the new value:\n\nppp_category_nonprofit &lt;-\n  ppp_forgiven_fixed |&gt; \n  mutate ( is_nonprofit = \n             if_else ( str_like(business_type, \"%Non_Profit%\") , \n                       \"Is nonprofit\", \n                       \"Not nonprofit\"))  \n\nYou can check it using a new verb, count(), to check the number of items by category:\n\nppp_category_nonprofit |&gt; \n  filter ( is_nonprofit == \"Is nonprofit\") |&gt; \n  count ( is_nonprofit, business_type)\n\n\n\n  \n\n\n\n(The profit categorization is unclear for some of the original types, such as professional associations , tribal concerns and cooperatives.)\n\n\nMore than two categories using case_when()\nSometimes you will want more than one outcome, such as setting a value for “High”, “Medium” and “Low”. Instead of if_then, use the function case_when, which lets you string along conditions and their outcomes. The tilde (~) is used to show what happens if it’s true. At the end, you can add an argument .default=\"Other\" to say “everything else”.\n\noriginal data |&gt; \n  case_when ( first condition ~ what if it's true,\n              second condition ~ what if  it's true, \n              third condition  ~ what if it's true, \n              .default =  what to do with everything that's left\n              )"
+    "section": "15.3 Creating categories for easier analysis",
+    "text": "15.3 Creating categories for easier analysis\nVery often, you’ll want to categorize entries in a database in order to make it simpler to count and sum the values in a meaningful way. For example, the business_type column has 24 different values, including NA . Here’s a list with the number of loans in each category:\n\n\n\n\n  \n\n\n\n\nYes-No categories using if_else()\nOne way to work with these is to create new columns with yes-no indicators for certain types of businesses like non-profits or individuals vs. companies.\nThe function to do this is if_else() , which tests a condition exactly the same way filter did, but then assigns a value based on whether it’s met or not. You list the condition first, then a comma , then what should happen if the condition is true, and then what should happen if it’s false.\nThe condition is done the same way you did it in a filter, but instead of picking out the rows that match, mutate acts on them separately from the rows that don’t match.\nThere’s no “maybe”, except for NA’s in the original value. Here’s the general form of what it looks like:\n\nnew_column_name = if_else ( test the old column for something as in a a filter,\n                         give it a value if it's true,\n                         give it another value if it's not true)\nSo here is a way to do this with the business_type using the same %in% operator you used in the filter lesson, saving it to new data frame in your Environment, then displaying the first 10 types of businesses using count()\n\nExample 1: Two categories using exact conditions\n\nppp_category_indiv &lt;- \n  ppp_forgiven_fixed |&gt;\n  mutate ( is_individual = \n              if_else ( business_type %in% \n                          c(\"Independent Contractors\", \n                            \"Sole Proprietorship\", \n                            \"Self-Employed Individuals\", \n                            \"Single Member LLC\"), \n                 \"Individual\", \n                 \"Organization\")\n  )  \n\nNow print off a sampling of a some of the rows to take a look:\n\nppp_category_indiv |&gt; \n  select ( is_individual, business_type) |&gt; \n  sample_n (50) \n\n\n\n  \n\n\n\n\n\nExample 2: Two categories using wildcards\nYou can also use the same str_like() function you used in filtering. Here, it sets whether or not the borrower was a non-profit. Remember, a % means “anything or nothing, and a _ means”any single character but it must be present”. In our case, the word is shown as “Non Profit” and “Non-Profit”. Once. you’ve set it to your new words, you can just filter on the new value:\n\nppp_category_nonprofit &lt;-\n  ppp_forgiven_fixed |&gt; \n  mutate ( is_nonprofit = \n             if_else ( str_like(business_type, \"%Non_Profit%\") , \n                       \"Is nonprofit\", \n                       \"Not nonprofit\"))  \n\nYou can check it using a new verb, count(), to check the number of items by category:\n\nppp_category_nonprofit |&gt; \n  filter ( is_nonprofit == \"Is nonprofit\") |&gt; \n  count ( is_nonprofit, business_type)\n\n\n\n  \n\n\n\n(The profit categorization is unclear for some of the original types, such as professional associations , tribal concerns and cooperatives.)\n\n\n\nMore than two categories using case_when()\nSometimes you will want more than one outcome, such as setting a value for “High”, “Medium” and “Low”. Instead of if_then, use the function case_when, which lets you string along conditions and their outcomes. The tilde (~) is used to show what happens if it’s true. At the end, you can add an argument .default=\"Other\" to say “everything else”.\n\noriginal data |&gt; \n  case_when ( first condition ~ what if it's true,\n              second condition ~ what if  it's true, \n              third condition  ~ what if it's true, \n              .default =  what to do with everything that's left\n              )"
   },
   {
     "objectID": "r-verb-mutate.html#putting-it-all-together",
@@ -574,11 +581,53 @@
     "text": "If you skipped converting the NA values above, the answers will always come out as NA. ↩︎"
   },
   {
-    "objectID": "r-verb-groupby.html#summarize-and-group_by-basics",
-    "href": "r-verb-groupby.html#summarize-and-group_by-basics",
+    "objectID": "start-groupby.html#summarizing-with-groups",
+    "href": "start-groupby.html#summarizing-with-groups",
+    "title": "16  Summarizing detail into totals",
+    "section": "16.1 Summarizing with groups",
+    "text": "16.1 Summarizing with groups\nSummarizing a list of items in R is done using the verbs group_by and summarize. Think of grouping as answering the questions, “How many?” and “How much?”. They are particularly powerful when your question also has the words “the most” or the “the least” or “of each”. Some examples:\n\nWhich Zip Code had the most crimes?\nWhat month had the least total rainfall?\nHow much did each candidate raise last quarter?\nIn playing cards, how many of each suit do I have in my hand?\nOn average, are Cronkite students taller or shorter than in other schools?\n\n\nConfusing grouping with sorting or arranging\nMany reporters confuse this summarization with “sorting”. One reason is that this is how we express the concept in plain language: “I want to sort Skittles by color”.\nBut in data analysis, sorting and and grouping are very different things. Sorting, which is done in R via the arrange() verb, involves shuffling a data frame’s rows into some order based on the values in a column. Grouping is a way to aggregate and compute summary statistics such as a count (the number of items), sum (how much they add up to), or average for category. It means “make piles and compute statistics for each one.”\n\n\nWhen to use filter vs. summary\nSomething that trips up beginners is a desire to see details and totals at the same time, which is more difficult than it sounds.\nA filter is used to display your selected items as a list. You’ll get to see all of the detail and every column. As a convenience, R shows you how many items are in that filtered list (usually). That’s great when you want to just look at them, or get more information about them. For instance, if you had a list of crimes by ZIP code, you might just want to see the list in your neighborhood – where, exactly, were they? When did they happen? Was it at night or the morning? What crimes happened on which blocks?\nAggregates are used when you just want to see summaries – does my ZIP code have more crime than others? Are robberies more common than car theft in my Zip code, and how does that compare to others?\nIn practice, you’ll go back and forth between summary and detail. They’re both important, just different."
+  },
+  {
+    "objectID": "start-groupby.html#motivational-exercise",
+    "href": "start-groupby.html#motivational-exercise",
+    "title": "16  Summarizing detail into totals",
+    "section": "16.2 Motivational exercise",
+    "text": "16.2 Motivational exercise\nThis exercise helps you internalize the differences between filtering and aggregating (or grouping). It uses a list of fatal police shootings logged by the Washington Post, downloaded in late 2022. The column names are pretty self-explanatory, and we’ll only be looking at the ethnicity column. (Scroll to the right if you can’t see it.)\n\n\nCreate a Quarto document in your project\nAdd the usual setup chunk to load the tidyverse, using the library command. Copy and paste the chunk from a previous lesson if you need to.\nCopy and paste this into a code chunk to read the data.\n\n\n\nwaposhootings &lt;- \n  readRDS ( \n        url ( \n           \"https://cronkitedata.s3.amazonaws.com/rdata/waposhootings.RDS\"\n           )\n        )\n\nExamine the data:\n\nwaposhootings |&gt; \n  sample_n (50)\n\n\n\n  \n\n\n\n\nFiltering method\nUsing a pen and paper, write out something that looks like the table below. The total number comes from the number of rows in the data frame.\n\n\n\n\nEthnicity\n# of victims\n% of total\n\n\n\n\nBlack, non-Hispanic\n\n\n\n\nWhite, non-Hispanic\n\n\n\n\nHispanic\n\n\n\n\nTOTAL\n5,945\n100%\n\n\n\n\nCopy and paste this code to list all fatal shootings of Black, non-Hispanic victims. Look at the number of rows returned and enter it into your handmade table.\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"Black, non-Hispanic\")\n\n\nRun this code to get the number of White, non-Hispanic victims and write the answer down on your handmade table\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"White, non-Hispanic\")\n\n\nRun this filter to get the number of Hispanic victims, and write the answer down.\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"Hispanic\")\n\n\nCompute the percentages\n\n\nCompute the percent of total each group represents. Do this on a calculator (or just type the formula into Google). Remember, the percent of total formula is: group / total * 100\nWrite the answers into your handwritten chart.\nNow, rewrite your chart so that the highest percentage is on the first row, the second highest on the second row, and the third highest on the third row.\nRepeat your calculations, since you always have to do every hand calculation twice. They won’t add up to 100% because we left out people who had an unknown ethnicity or other ethnicity."
+  },
+  {
+    "objectID": "start-groupby.html#aggregating-for-an-easier-way",
+    "href": "start-groupby.html#aggregating-for-an-easier-way",
+    "title": "16  Summarizing detail into totals",
+    "section": "16.3 Aggregating for an easier way",
+    "text": "16.3 Aggregating for an easier way\nThat’s a lot of work. Imagine if you had to do this 50 times for 50 states, or even more for every Zip code? There has to be an easier way, and there is: Grouping and summarizing.\nWe’ll build the code one piece at a time so you can see what’s happening:\n\nCopy this code, one chunk at a time, into your Quarto document and run them to see what the answers are.\n\n\nGet the totals\n\n\nwaposhootings |&gt; \n  summarize ( num_shootings = n() ) \n\n\nAdd a “group” to summarize by ethnicity\n\n\nwaposhootings |&gt; \n  group_by (ethnicity) |&gt; \n  summarize (num_shootings = n() )\n\n\nArrange to get the highest number at the top\n\n\nwaposhootings |&gt; \n  group_by ( ethnicity) |&gt; \n  summarize (num_shootings = n() ) |&gt; \n  arrange ( desc (num_shootings ))\n\n\nCompute the percent\nNotice that the “mutate” line is exactly the same formula you typed into your calculator. The only difference is that we computed the total number of shootings instead of typing it. \nThis is one of the most common patterns you’ll use in data reporting, and we can get much more sophisticated with it. Don’t expect to understand it fully right away, but you should recognize the idea: If you find yourself filtering in order to count something, consider using grouping instead.\n\n\nwaposhootings |&gt; \n  group_by ( ethnicity) |&gt; \n  summarize ( num_shootings = n() ) |&gt; \n  mutate ( total = sum ( num_shootings ) , \n           percent = num_shootings /total * 100 ) |&gt; \n  arrange ( desc ( num_shootings))"
+  },
+  {
+    "objectID": "start-groupby.html#differences",
+    "href": "start-groupby.html#differences",
+    "title": "16  Summarizing detail into totals",
+    "section": "16.4 Differences",
+    "text": "16.4 Differences\nWhen you filtered, you could see the details of every row. That’s great if you want to explore it or find examples. But it’s not great for that sentence you need when you want to say which group is the largest or smallest, or compare the numbers across groups.\nWhen you grouped, you got statistics by ethnicity, but you couldn’t see the details.\nTypically, you’ll go back and forth between aggregating and filtering, using each method to get to different goals. Don’t fall in the trap of trying to force all of your questions into a filter, or all of your questions into aggregation. Instead, think about which method will give you the answers you want most efficiently."
+  },
+  {
+    "objectID": "start-groupby.html#your-questions",
+    "href": "start-groupby.html#your-questions",
+    "title": "16  Summarizing detail into totals",
+    "section": "16.5 Your questions:",
+    "text": "16.5 Your questions:\n\nOn your Quarto document, try to write out, in plain English, what you just did. See if you can articulate what the difference between filtering and aggregating is, in your own words."
+  },
+  {
+    "objectID": "r-verb-groupby.html#summarize-for-statistics",
+    "href": "r-verb-groupby.html#summarize-for-statistics",
+    "title": "17  Verbs in depth: Aggregating with groups",
+    "section": "17.1 summarize for statistics",
+    "text": "17.1 summarize for statistics\nsummarize computes summary statistics such as the number of rows in a data frame or the sum of dollar values. It removes the original columns completely, and only produces the summary statistics you compute within that statement. Using summarize alone produces a data frame with one row. It’s the equivalent of putting nothing in your pivot table in Excel other than the “Values” area.\nAnother way to think of summarize is that it collapses your list of items (loans, in our example) into a statistical report.\n\nThe dreaded NA\nYou saw in the mutate section that missing values are always a problem. Because they’re unknown, they can’t match anything else, they can’t be considered 0, and they can warp any answers you get. But there’s usually nothing you can do about missing data, so you have to tell the program exactly what to do about them.\nThere are two choices:\n\nLet them infect everything they touch, turning everything into NA. In this scenario, a total of the dollar values in a column would be NA if any of the values in that column is missing:\n\n\nppp_orig |&gt; \n  summarize ( total = sum(forgiveness_amount))\n\n\n\n  \n\n\n\n\nIgnore them in a computation completely, effectively removing that value from your calculation.\n\nThere’s no right answer, and it depends on what you’re doing. In some cases, you know that they stand for the value 0, and in others you don’t. We will usually ignore them by adding an argument to every summary function that could be infected by them : na.rm = TRUE , which means, “remove NA’s before you do anything.”.\n\nppp_orig |&gt; \n  summarize ( total = sum (forgiveness_amount, na.rm=TRUE))\n\n\n\n  \n\n\n\n\n\nSummary functions\nSome of the common functions you’ll use to summarize are :\n\nmean (column_name, na.rm=T) – for an average : Numbers only\nsum (column_name, na.rm = T): Numbers only\nn() – for “how many”, or “count”. Anything - this counts rows, not values\nn_distinct ( column_name) : The number of unique entries in the column. Use it to see how many categories there are in a column.\nmedian (column_name, na.rm=T): Numbers only\nmin (column_name, na.rm=T): Dates and numbers\nmax (column_name , na.rm=T): Dates and numbers\n\nWhen used on the whole data frame, it’s customary to just glimpse the output, since there’s only one row:\n\nppp_orig |&gt; \n  summarize ( n(), \n              mean (amount, na.rm=T), \n              mean (forgiveness_amount, na.rm=T), \n              min (date_approved, na.rm=T), \n              max (date_approved, na.rm= T), \n              n_distinct ( business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 6\n$ `n()`                                 &lt;int&gt; 169259\n$ `mean(amount, na.rm = T)`             &lt;dbl&gt; 73206.66\n$ `mean(forgiveness_amount, na.rm = T)` &lt;dbl&gt; 77050.41\n$ `min(date_approved, na.rm = T)`       &lt;date&gt; 2020-04-03\n$ `max(date_approved, na.rm = T)`       &lt;date&gt; 2021-06-29\n$ `n_distinct(business_type)`           &lt;int&gt; 24\n\n\nThis produced a data frame with 1 row and 5 columns. The column names are the same as the formulas that created them, which is difficult to work with. Create new column names using the name (in back-ticks if it’s got spaces or special characters) and assign them the values of the summaries using the = sign:\n\nppp_orig |&gt; \n  summarize ( number_of_rows =  n(), \n              mean_amount = mean (amount, na.rm=T),    \n              median_amount = median (amount, na.rm=T),\n              mean_forgiven = mean (forgiveness_amount, na.rm=T),   \n              first_loan = min (date_approved, na.rm=T),   \n              last_loan = max (date_approved, na.rm= T), \n              business_type_count = n_distinct(business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 7\n$ number_of_rows      &lt;int&gt; 169259\n$ mean_amount         &lt;dbl&gt; 73206.66\n$ median_amount       &lt;dbl&gt; 20800\n$ mean_forgiven       &lt;dbl&gt; 77050.41\n$ first_loan          &lt;date&gt; 2020-04-03\n$ last_loan           &lt;date&gt; 2021-06-29\n$ business_type_count &lt;int&gt; 24\n\n\nNote that the mean forgiven removes those with missing values for the forgiven amount, which is wrong! We need to turn them into zeroes first."
+  },
+  {
+    "objectID": "r-verb-groupby.html#grouping-for-lists",
+    "href": "r-verb-groupby.html#grouping-for-lists",
     "title": "17  Verbs in depth: Aggregating with groups",
-    "section": "17.1 summarize and group_by basics",
-    "text": "17.1 summarize and group_by basics\n\nsummarize\nsummarize1 computes summary statistics such as the number of rows in a data frame or the sum of dollar values. It removes the original columns completely, and only produces the summary statistics you compute within that statement. Using summarize alone produces a data frame with one row. It’s the equivalent of putting nothing in your pivot table in Excel other than the “Values” area.\nAnother way to think of summarize is that it collapses your list of items (loans, in our example) into a statistical report.\n\nThe dreaded NA\nYou saw in the mutate section that missing values are always a problem. Because they’re unknown, they can’t match anything else, they can’t be considered 0, and they can warp any answers you get. But there’s usually nothing you can do about missing data, so you have to tell the program exactly what to do about them.\nThere are two choices:\n\nLet them infect everything they touch, turning everything into NA. In this scenario, a total of the dollar values in a column would be NA if any of the values in that column is missing:\n\n\nppp_orig |&gt; \n  summarize ( sum(forgiveness_amount))\n\n\n\n  \n\n\n\n\nIgnore them in a computation completely, effectively removing that value from your calculation. Be careful with this - it means that two summary statistics (say,n() and sum()) could be using a different set of rows!\n\nThere’s no right answer, and it depends on what you’re doing. In some cases, you know that they stand for the value 0, and in others you don’t. We will usually ignore them by adding an argument to every summary function that could be infected by them : na.rm = TRUE , which means, “remove NA’s before you do anything.”.\n\n\n\nSummary functions\nSome of the common functions you’ll use to summarize are :\n\nmean (column_name, na.rm=T) – for an average : Numbers only\nsum (column_name, na.rm = T): Numbers only\nn() – for “how many”, or “count”. Anything - this counts rows, not values\nn_distinct ( column_name) : The number of unique entries in the column. Use it to see how many categories there are in a column.\nmedian (column_name, na.rm=T): Numbers only\nmin (column_name, na.rm=T): Dates and numbers\nmax (column_name , na.rm=T): Dates and numbers\n\nWhen used on the whole data frame, it’s customary to just glimpse the output, since there’s only one row:\n\nppp_orig |&gt; \n  summarize ( n(), \n              mean (amount, na.rm=T), \n              mean (forgiveness_amount, na.rm=T), \n              min (date_approved, na.rm=T), \n              max (date_approved, na.rm= T), \n              n_distinct ( business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 6\n$ `n()`                                 &lt;int&gt; 169259\n$ `mean(amount, na.rm = T)`             &lt;dbl&gt; 73206.66\n$ `mean(forgiveness_amount, na.rm = T)` &lt;dbl&gt; 77050.41\n$ `min(date_approved, na.rm = T)`       &lt;date&gt; 2020-04-03\n$ `max(date_approved, na.rm = T)`       &lt;date&gt; 2021-06-29\n$ `n_distinct(business_type)`           &lt;int&gt; 24\n\n\nThis produced a data frame with 1 row and 5 columns. The column names are the same as the formulas that created them, which is difficult to work with. Create new column names using the name (in back-ticks if it’s got spaces or special characters) and assign them the values of the summaries using the = sign:\n\nppp_orig |&gt; \n  summarize ( number_of_rows =  n(), \n              mean_amount = mean (amount, na.rm=T),    \n              median_amount = median (amount, na.rm=T),\n              mean_forgiven = mean (forgiveness_amount, na.rm=T),   # doesn't count the missing as zero!\n              first_loan = min (date_approved, na.rm=T),   \n              last_loan = max (date_approved, na.rm= T), \n              business_type_count = n_distinct(business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 7\n$ number_of_rows      &lt;int&gt; 169259\n$ mean_amount         &lt;dbl&gt; 73206.66\n$ median_amount       &lt;dbl&gt; 20800\n$ mean_forgiven       &lt;dbl&gt; 77050.41\n$ first_loan          &lt;date&gt; 2020-04-03\n$ last_loan           &lt;date&gt; 2021-06-29\n$ business_type_count &lt;int&gt; 24\n\n\n\n\nGrouping\nNow that you know how to summarize the whole data frame, you’ll want to start getting totals by category. This is the same thing as a pivot table – the column names that create the “groups” are the equivalent of the Rows area a spreadsheet pivot table:\n\n\nGrouping by one column\nIn the PPP data, the “draw” refers to which of the two programs was involved - the original one, or the one passed by Congress in late 2020.\nHere’s how we’d get some key statistics by draw:\n\nppp_orig |&gt;\n  group_by ( draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              `# of loans` = n() \n  )\n\n\n\n  \n\n\n\nHere are a couple of things to note about grouped output:\n\nThe only columns saved are the ones that are shown in either the group_by or summarize rows. All of the other original columns have been eliminated. You no longer have them to work with .\n\nThe names of the columns for the summary statistics are the ones defined before the “=” sign in the summarize statement.\nTRAP! Don’t ever name your summary columns the same thing as a group_by column. It will override those names, and your output will be unintelligible.\n\n\n\n\n\n\n\nNaming your columns\n\n\n\nNote that the name of the columns doesn’t always follow our standard. In this case, # of loans has a special character and spaces. In order to create or use it, you must enclose them in back-tics (`) or you’ll get an error.\n\n\n\n\n\n\n\n\nExpand your table to a new window\n\n\n\n\n\nIf you’re using visual mode, the table that gets produced can be really squished. To see more of it, pop it out into its own window using the leftmost button. 2\n\n\n\n\n\n\n\n\n\nGrouping by more than one column\nIf you wanted to know the numbers outstanding and forgiven by draw, you could add another column to the group by:\n\nppp_orig |&gt;\n  group_by ( loan_status, draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              loan_ct = n() \n  )\n\n\n\n  \n\n\n\nThe summarize verb here has created a data frame that has only the variables you identified in the group_by and summarize commands - everything else has been removed, since it’s a summary of your data.\n\n\n\nA shortcut : count()\nIf all you want to do is count or add by group, you can use the count() function as a shortcut. It does the exact same thing as a combination of group_by() and summarize( n() ) and arrange()` to get the number of items in each category, sorted by the most frequent to least:\n\nppp_orig |&gt; \n  count ( loan_status, draw, \n          sort=TRUE,\n          name = \"loans\")"
+    "section": "17.2 Grouping for lists",
+    "text": "17.2 Grouping for lists\nNow that you know how to summarize the whole data frame, you’ll want to start getting totals by category. This is the same thing as a pivot table in spreadsheets: the column names that create the “groups” are the equivalent of the Rows area a spreadsheet pivot table:\n\n\nGrouping by one column\nIn the PPP data, the “draw” refers to which of the two programs was involved - the original one created in April 2020, or the one with stricter criteria passed by Congress that December.\nHere’s how we’d get some key statistics by draw:\n\nppp_orig |&gt;\n  group_by ( draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              `# of loans` = n() \n  )\n\n\n\n  \n\n\n\nHere are a couple of things to note about grouped output:\n\nThe only columns saved are the ones that are shown in either the group_by or summarize rows. All of the other original columns have been eliminated. You no longer have them to work with .\nThe names of the columns for the summary statistics are the ones defined before the “=” sign in the summarize statement.\nTRAP! Don’t ever name your summary columns the same thing as a group_by column. It will override those names, and your output will be unintelligible.\n\n\n\n\n\n\n\nNaming your columns\n\n\n\nNote that the name of the columns doesn’t always follow our standard. In this case, # of loans has a special character and spaces. In order to create or use it, you must enclose them in back-tics (`) or you’ll get an error.\n\n\n\n\nGrouping by more than one column\nIf you wanted to know the numbers outstanding and forgiven by draw, you could add another column to the group by:\n\nppp_orig |&gt;\n  group_by ( loan_status, draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              loan_ct = n() \n  )\n\n\n\n  \n\n\n\n\n\nA shortcut : count()\nIf all you want to do is count or add by group, you can use the count() function as a shortcut. It does the exact same thing as a combination of group_by() and summarize( n() ) and arrange()` to get the number of items in each category, sorted by the most frequent to least:\n\nppp_orig |&gt; \n  count ( loan_status, draw, \n          sort=TRUE,\n          name = \"loan_ct\")\n\n\n\n  \n\n\n\n\n\n\n\n\n\nNew ways to summarize\n\n\n\nA new version of the tidyverse, in 2023, has changed some of the options for the summarize() verb that make grouping unnecessary much of the time, but it’s confusing. I’m skipping it for now."
   },
   {
     "objectID": "r-verb-groupby.html#using-and-converting-groups",
@@ -615,6 +664,13 @@
     "section": "18.1 Join basics and relational databases",
     "text": "18.1 Join basics and relational databases\nHere is a great explainer on joins made by a previous MAIJ student, Andy Blye. Be sure to watch it:\n\njoining in computer programming matches columns in one table to another, where the values within one or more columns match exactly. Here’s an example from Jenny Bryan’s Stat 545 course textbook:\n\n\n\n\nsuperheroes\n\n\n\n\nRelational databases\nMost large data systems created since the 1980s are called “relational databases”, which means that each unit (person, ticket, vehicle) is stored in a different table1, and common columns link them together. They also routinely use codes or shorthand to store data, and provide another table to “look up” the details.\nFor example, your student ID is stored in one place in the university system with your name, address, email address, etc. When you sign up for courses, the database looks up that information to attach to your schedule. You only have to change it in one place, and it is automatically sent out through all of the interactions you have with the university.\nSimilarly, the course number, section and term is all that needs to be stored in the system for any semester. Those data points then populate the name of the course, the students registered for it, and holidays.\nThe system is created this way because it’s more efficient and reliable. Important information is stored only once, and can then be applied to millions of rows.\nExamples of relational databases include:\n\nCampaign finance systems, where donors are stored in one table and candidates in another, linked through a candidate or political action committee id.\nInspection records, such as those for restaurants, hospitals, housing code violations and workplace safety, which typically have at least three tables: The establishment (like a restaurant or a workplace), an inspection (an event on a date), and a violation (something that they found). Each table has its own ID, which is used whenever they are linked together.\nA court docket data system, which usually has many types of rows: A master case table links to information on charges, defendants, lawyers, sentences and court hearings.\n\nThis is similar to, but stricter than, the tidy data principles of separating different kinds of information into separate data frames."
   },
+  {
+    "objectID": "r-verb-join.html#join-syntax",
+    "href": "r-verb-join.html#join-syntax",
+    "title": "18  Verbs in depth: Matchmaking with joins",
+    "section": "18.2 join syntax",
+    "text": "18.2 join syntax\n\n\n\n\n\n\nTip\n\n\n\nThere are actually pretty good explanations of the concept of joining and the variations of it in R in the documentation.\n\n\nThere are several kinds of joins, but the syntax is similar across them.\n  old_table |&gt;\n     inner_join (new_table , \n     join_by = (name of old_table column == name of new_table column) )\n     \n\n\n\n\n\n\nTip\n\n\n\nTry to piece together what this is saying, similar to the filter conditions: If the two columns are equal (double-equal signs), put them together.\n\n\n\nTypes of joins\n\nAn inner_join means that the value(s) in the common columns must match in BOTH tables – it will eliminate any row without a match\nA left_join or right_join keeps everything from one table, and only the information that matches from the other. Those columns will contain NA wherever the match fails.\nA full_join puts together both tables no matter whether there is any match. It’s pretty rare to use this."
+  },
   {
     "objectID": "r-verb-join.html#matchmaking-with-joins",
     "href": "r-verb-join.html#matchmaking-with-joins",
@@ -660,191 +716,191 @@
   {
     "objectID": "advanced-regex.html#types-of-regular-expression-patterns",
     "href": "advanced-regex.html#types-of-regular-expression-patterns",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.1 Types of regular expression patterns",
-    "text": "18.1 Types of regular expression patterns\n\nLiteral strings\nThese are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.\nYou can tell the regex that you want to find your pattern at the beginning or end of a line:\n  ^   = \"Find only at the beginning of a line\"\n  $   = \"Find only at the end of a line\"\n\n\nWild cards\nA wild card is a character you use to indicate the word “anything”. Here are some ways to use wild cards in regex:\n    .      = \"any single character of any type\"\n    .?     = \"a possible single character of any type (but it might not exist)\"\n    .*     = \"anything or nothing of any length\"\n    .+     = \"anything one or more times\"\n    .{1,3} = \"anything running between 1 and 3 characters long\"\nRegular expressions also have wild cards of specific types. In R, they are “escaped” using two backslashes. In other languages and in the example https://regex101.com site they only use one backslash:\n      \\\\d   = \"Any digit\"\n      \\\\w   = \"Any word character\"\n      \\\\s   = \"Any whitespace (tab, space, etc.)\"\n      \\\\b   = \"Any word boundary\" (period, comma, space, etc.)\nWhen you upper-case them, it’s the opposite:\n      \\\\D = \"Anything but a digit\"\n\n\nCharacter classes\nSometimes you want to tell the regex what characters it is allowed to accept. For example, say you don’t know whether there is an alternative spelling for a name – you can tell the regex to either ignore a character, or take one of several.\nIn R, we saw that there were alternative spellings for words like “summarize” – the British and the American spellings. You could, for example, use this pattern to pick up either spelling:\n        summari[sz]e\nThe bracket tells the regex that it’s allowed to take either one of those characters. You can also use ranges:\n  [a-zA-Z0-9]\nmeans that any lower case, upper case or numeric character is allowed.\n\n\nEscaping\nBecause they’re already being used for special purposes, some characters have to be “escaped” before you can search for them. Notably, they are parentheses (), periods, backslashes, dollar signs, question marks, dashes and carets.\nThis means that to find a period or question mark, you have to use the pattern\n    \\\\. or\n    \\\\?\nIn the Regex101, this is the biggest difference among the flavors of regex – Python generally requires the least amount of escaping.\n\n\nMatch groups\nUse parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with lists."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.1 Types of regular expression patterns",
+    "text": "19.1 Types of regular expression patterns\n\nLiteral strings\nThese are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.\nYou can tell the regex that you want to find your pattern at the beginning or end of a line:\n  ^   = \"Find only at the beginning of a line\"\n  $   = \"Find only at the end of a line\"\n\n\nWild cards\nA wild card is a character you use to indicate the word “anything”. Here are some ways to use wild cards in regex:\n    .      = \"any single character of any type\"\n    .?     = \"a possible single character of any type (but it might not exist)\"\n    .*     = \"anything or nothing of any length\"\n    .+     = \"anything one or more times\"\n    .{1,3} = \"anything running between 1 and 3 characters long\"\nRegular expressions also have wild cards of specific types. In R, they are “escaped” using two backslashes. In other languages and in the example https://regex101.com site they only use one backslash:\n      \\\\d   = \"Any digit\"\n      \\\\w   = \"Any word character\"\n      \\\\s   = \"Any whitespace (tab, space, etc.)\"\n      \\\\b   = \"Any word boundary\" (period, comma, space, etc.)\nWhen you upper-case them, it’s the opposite:\n      \\\\D = \"Anything but a digit\"\n\n\nCharacter classes\nSometimes you want to tell the regex what characters it is allowed to accept. For example, say you don’t know whether there is an alternative spelling for a name – you can tell the regex to either ignore a character, or take one of several.\nIn R, we saw that there were alternative spellings for words like “summarize” – the British and the American spellings. You could, for example, use this pattern to pick up either spelling:\n        summari[sz]e\nThe bracket tells the regex that it’s allowed to take either one of those characters. You can also use ranges:\n  [a-zA-Z0-9]\nmeans that any lower case, upper case or numeric character is allowed.\n\n\nEscaping\nBecause they’re already being used for special purposes, some characters have to be “escaped” before you can search for them. Notably, they are parentheses (), periods, backslashes, dollar signs, question marks, dashes and carets.\nThis means that to find a period or question mark, you have to use the pattern\n    \\\\. or\n    \\\\?\nIn the Regex101, this is the biggest difference among the flavors of regex – Python generally requires the least amount of escaping.\n\n\nMatch groups\nUse parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with lists."
   },
   {
     "objectID": "advanced-regex.html#sample-data",
     "href": "advanced-regex.html#sample-data",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.2 Sample data",
-    "text": "18.2 Sample data\nHere are three small text files that you can copy and paste from your browser into the regex101.com site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.\n\nA list of phone numbers in different formats\nA list of dates that you need to convert into a different form.\nA list of addresses that are in multiple lines, and you need to pull out the pieces. (Courtesy of IRE)\nA small chunk of the H2B visa applications from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.2 Sample data",
+    "text": "19.2 Sample data\nHere are three small text files that you can copy and paste from your browser into the regex101.com site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.\n\nA list of phone numbers in different formats\nA list of dates that you need to convert into a different form.\nA list of addresses that are in multiple lines, and you need to pull out the pieces. (Courtesy of IRE)\nA small chunk of the H2B visa applications from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format."
   },
   {
     "objectID": "advanced-regex.html#testing-and-examining-patterns-with-regex101",
     "href": "advanced-regex.html#testing-and-examining-patterns-with-regex101",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.3 Testing and examining patterns with Regex101",
-    "text": "18.3 Testing and examining patterns with Regex101\nRegex 101 is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\\` in R.\n\nLooking for specific words or characters\nThe easiest regex is one that has just the characters you’re looking for when you know that they are the right case. They’re called literals because you are literally looking for those letters or characters."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.3 Testing and examining patterns with Regex101",
+    "text": "19.3 Testing and examining patterns with Regex101\nRegex 101 is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\\` in R.\n\nLooking for specific words or characters\nThe easiest regex is one that has just the characters you’re looking for when you know that they are the right case. They’re called literals because you are literally looking for those letters or characters."
   },
   {
     "objectID": "advanced-regex.html#practice-1-extract-date-parts",
     "href": "advanced-regex.html#practice-1-extract-date-parts",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.4 Practice #1: Extract date parts",
-    "text": "18.4 Practice #1: Extract date parts\nIn Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.1\nWe want to turn dates that look like this:\n  1/24/2018\ninto something that looks like this:\n 2008-1-24\nCopy and paste these numbers into the regex 101 window:\n9/7/2017\n9/11/1998\n9/11/2017\n9/19/2018\n9/15/2017\n10/13/2019\n11/3/2017\nFirst, you can use any digit using the pattern “. Try to do it in pieces. First, see if you can find one or two digits at the beginning of the line.\n  ^\\d{1,2}\nTry coming up with the rest of it on your own before you type in the answer:\n  ^\\d{1,2}.\\d{1,2}.\\d{4}\n(This works because regular expressions normally are “greedy”. That is, if you tell it “one or two digits”, it will always take two if they exist.)\nPut parentheses around any pieces that you want to use for later:\n\n\n\n\n\nNow each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.4 Practice #1: Extract date parts",
+    "text": "19.4 Practice #1: Extract date parts\nIn Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.1\nWe want to turn dates that look like this:\n  1/24/2018\ninto something that looks like this:\n 2008-1-24\nCopy and paste these numbers into the regex 101 window:\n9/7/2017\n9/11/1998\n9/11/2017\n9/19/2018\n9/15/2017\n10/13/2019\n11/3/2017\nFirst, you can use any digit using the pattern “. Try to do it in pieces. First, see if you can find one or two digits at the beginning of the line.\n  ^\\d{1,2}\nTry coming up with the rest of it on your own before you type in the answer:\n  ^\\d{1,2}.\\d{1,2}.\\d{4}\n(This works because regular expressions normally are “greedy”. That is, if you tell it “one or two digits”, it will always take two if they exist.)\nPut parentheses around any pieces that you want to use for later:\n\n\n\n\n\nNow each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces."
   },
   {
     "objectID": "advanced-regex.html#practice-2-extract-pieces-of-phone-numbers",
     "href": "advanced-regex.html#practice-2-extract-pieces-of-phone-numbers",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.5 Practice #2: Extract pieces of phone numbers",
-    "text": "18.5 Practice #2: Extract pieces of phone numbers\nHere are some phone numbers in different formats:\n    623-374-1167\n    760.352.5212\n    831-676-3833\n    (831)-676-3833\n    623-374-1167 ext 203\n    831-775-0370\n    (602)-955-0222  x20\n    928-627-8080\n    831-784-1453\nThis is a little more complicated than it looks, so try piecing together what this one says:\n      (\\d{3})[-.\\)]+(\\d{3})[-.]+(\\d{4})\n(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)\nAnything within parentheses will be “captured” in a block."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.5 Practice #2: Extract pieces of phone numbers",
+    "text": "19.5 Practice #2: Extract pieces of phone numbers\nHere are some phone numbers in different formats:\n    623-374-1167\n    760.352.5212\n    831-676-3833\n    (831)-676-3833\n    623-374-1167 ext 203\n    831-775-0370\n    (602)-955-0222  x20\n    928-627-8080\n    831-784-1453\nThis is a little more complicated than it looks, so try piecing together what this one says:\n      (\\d{3})[-.\\)]+(\\d{3})[-.]+(\\d{4})\n(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)\nAnything within parentheses will be “captured” in a block."
   },
   {
     "objectID": "advanced-regex.html#practice-3-extract-address-pieces",
     "href": "advanced-regex.html#practice-3-extract-address-pieces",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.6 Practice #3: Extract address pieces",
-    "text": "18.6 Practice #3: Extract address pieces\nHere are a few lines of the data from Prof. McDonald’s tutorial, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)\n    \"10111 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.370945933000485, -97.6925542359997)\"\n    \"3636 N FM 620 RD\n    AUSTIN, TX 78734\n    (30.377873241000486, -97.9523496219997)\"\n    \"9919 SERVICE AVE\n    AUSTIN, TX 78743\n    (30.205028616000448, -97.6625588019997)\"\n    \"10601 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.37476574700048, -97.6903937089997)\"\n    \"801 E WILLIAM CANNON DR Unit 205\n    AUSTIN, TX 78745\n    (30.190914575000477, -97.77193838799968)\"\n    \"4408 LONG CHAMP DR\n    AUSTIN, TX 78746\n    (30.340981111000474, -97.7983147919997)\"\n    \"625 W BEN WHITE BLVD EB\n    AUSTIN, TX 78745\n    (30.206884239000487, -97.7956469989997)\"\n    \"3914 N LAMAR BLVD\n    AUSTIN, TX 78756\n    (30.307477098000447, -97.74169675199965)\"\n    \"15201 FALCON HEAD BLVD\n    BEE CAVE, TX 78738\n    (30.32068282700044, -97.96890311999965)\"\n    \"11905 FM 2244 RD Unit 100\n    BEE CAVE, TX 78738\n    (30.308363203000454, -97.92393357799966)\"\n    \"3801 JUNIPER TRCE\n    BEE CAVE, TX 78738\n    (30.308247975000484, -97.93511531999968)\"\n    \"12800 GALLERIA CIR Unit 101\n    BEE CAVE, TX 78738\n    (30.307996778000472, -97.94065088199966)\"\n    \"12400 W SH 71 Unit 510\n    BEE CAVE, TX 78733\n    (30.330682136000462, -97.86979886299969)\"\n    \"716 W 6TH ST\n    AUSTIN, TX 78701\n    (30.27019732500048, -97.75036306299967)\"\n    \"3003 BEE CAVES RD\n    ROLLINGWOOD, TX 78746\n    (30.271592738000436, -97.79583786499967)\""
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.6 Practice #3: Extract address pieces",
+    "text": "19.6 Practice #3: Extract address pieces\nHere are a few lines of the data from Prof. McDonald’s tutorial, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)\n    \"10111 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.370945933000485, -97.6925542359997)\"\n    \"3636 N FM 620 RD\n    AUSTIN, TX 78734\n    (30.377873241000486, -97.9523496219997)\"\n    \"9919 SERVICE AVE\n    AUSTIN, TX 78743\n    (30.205028616000448, -97.6625588019997)\"\n    \"10601 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.37476574700048, -97.6903937089997)\"\n    \"801 E WILLIAM CANNON DR Unit 205\n    AUSTIN, TX 78745\n    (30.190914575000477, -97.77193838799968)\"\n    \"4408 LONG CHAMP DR\n    AUSTIN, TX 78746\n    (30.340981111000474, -97.7983147919997)\"\n    \"625 W BEN WHITE BLVD EB\n    AUSTIN, TX 78745\n    (30.206884239000487, -97.7956469989997)\"\n    \"3914 N LAMAR BLVD\n    AUSTIN, TX 78756\n    (30.307477098000447, -97.74169675199965)\"\n    \"15201 FALCON HEAD BLVD\n    BEE CAVE, TX 78738\n    (30.32068282700044, -97.96890311999965)\"\n    \"11905 FM 2244 RD Unit 100\n    BEE CAVE, TX 78738\n    (30.308363203000454, -97.92393357799966)\"\n    \"3801 JUNIPER TRCE\n    BEE CAVE, TX 78738\n    (30.308247975000484, -97.93511531999968)\"\n    \"12800 GALLERIA CIR Unit 101\n    BEE CAVE, TX 78738\n    (30.307996778000472, -97.94065088199966)\"\n    \"12400 W SH 71 Unit 510\n    BEE CAVE, TX 78733\n    (30.330682136000462, -97.86979886299969)\"\n    \"716 W 6TH ST\n    AUSTIN, TX 78701\n    (30.27019732500048, -97.75036306299967)\"\n    \"3003 BEE CAVES RD\n    ROLLINGWOOD, TX 78746\n    (30.271592738000436, -97.79583786499967)\""
   },
   {
     "objectID": "advanced-regex.html#on-your-own",
     "href": "advanced-regex.html#on-your-own",
-    "title": "18  Regular Expressions for pattern matching",
-    "section": "18.7 On your own",
-    "text": "18.7 On your own\nThis is a small list of H2A visa applications, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.\nYou might try:\n\nFinding all of the LLC’s in the list (limited liability companies) of names. (You should turn on the case-insensitive flag in Regex 101 or set that flag in your program if you do.)\nSee how far you can get in standardizing the addresses.\nSplit the city, state and zip code of the worksite.\nFind all of the jobs related to field crops such as lettuce or celery."
+    "title": "19  Regular Expressions for pattern matching",
+    "section": "19.7 On your own",
+    "text": "19.7 On your own\nThis is a small list of H2A visa applications, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.\nYou might try:\n\nFinding all of the LLC’s in the list (limited liability companies) of names. (You should turn on the case-insensitive flag in Regex 101 or set that flag in your program if you do.)\nSee how far you can get in standardizing the addresses.\nSplit the city, state and zip code of the worksite.\nFind all of the jobs related to field crops such as lettuce or celery."
   },
   {
     "objectID": "advanced-regex.html#footnotes",
     "href": "advanced-regex.html#footnotes",
-    "title": "18  Regular Expressions for pattern matching",
+    "title": "19  Regular Expressions for pattern matching",
     "section": "",
     "text": "Each language implements regular expressions slightly differently – when you begin to learn more languages, this will be one of the first things you’ll need to look up.↩︎"
   },
   {
     "objectID": "advanced-regex-r.html#resouces-for-using-regular-expressions-in-r",
     "href": "advanced-regex-r.html#resouces-for-using-regular-expressions-in-r",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.1 Resouces for using regular expressions in R",
-    "text": "19.1 Resouces for using regular expressions in R\n\nThe tidyverse cheat sheet on the package stringr (which is party of tidyverse) has the outlines of each function that uses these regular expression. It’s a little hard to read, but has all of the shortcuts on the second page. Look at the part that says, “type this”, which translates to “to mean this” as in Regex101.\nNathan Collins, a student in the 2022 MAIJ cohort, [did an explainer on using str_detect() on the ACLED])(https://npcdata.github.io/projects/collins_explainer.html) data that many of you are using for your projects\n\n\n\n\n\n\n\nImportant\n\n\n\nIn R, regular expressions are a little different than in the Python and other languages listed in regex101 .\n\nThe expressions are still case sensitive! In real life, it’s often easier to use a combination of functions, first to turn the column into lower case, and then to use the regular expression on that. It’s much easier than trying to match which case it is.\nInstead of one \\ backslash (\\), there are two (\\\\) to “escape” special characters like parentheses or to use a special type of character such as a digit."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.1 Resouces for using regular expressions in R",
+    "text": "20.1 Resouces for using regular expressions in R\n\nThe tidyverse cheat sheet on the package stringr (which is party of tidyverse) has the outlines of each function that uses these regular expression. It’s a little hard to read, but has all of the shortcuts on the second page. Look at the part that says, “type this”, which translates to “to mean this” as in Regex101.\nNathan Collins, a student in the 2022 MAIJ cohort, [did an explainer on using str_detect() on the ACLED])(https://npcdata.github.io/projects/collins_explainer.html) data that many of you are using for your projects\n\n\n\n\n\n\n\nImportant\n\n\n\nIn R, regular expressions are a little different than in the Python and other languages listed in regex101 .\n\nThe expressions are still case sensitive! In real life, it’s often easier to use a combination of functions, first to turn the column into lower case, and then to use the regular expression on that. It’s much easier than trying to match which case it is.\nInstead of one \\ backslash (\\), there are two (\\\\) to “escape” special characters like parentheses or to use a special type of character such as a digit."
   },
   {
     "objectID": "advanced-regex-r.html#sample-data",
     "href": "advanced-regex-r.html#sample-data",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.2 Sample data",
-    "text": "19.2 Sample data\nFor this example, we’ll use a small set of Tweets from Donald Trump, representing the last few months he was on on the platform. It runs from Nov. 1, 2020 to Jan. 8, 2021, when Twitter banned him. Retweets were removed, but there are a number of tweets that are just links to other content.\nHere is a random sample of the 1,112 tweets:\n\n\n\n\n\n\n\nUsing tweets as a source for data is a little dated, but it’s a good example of trying to pick out text from a set of short items.\n\nString functions for regular expressions\n\nstr_detect( col_name, expression) looks for the presence of the expression, and returns TRUE or FALSE.\nstr_extract(col_name, expression) does the same thing, except it returns whatever the first thing that matched your expression. It includes wild cards, so it could just return the whole phrase.\nstr_match(col_name, expression) is more complicated - it lets you use pieces of the expression, but it returns a list, which is hard to work with. We’ll skip that for now.\n\nSome other useful string functions are:\n\nTo convert the case, use str_to_lower(colname), str_to_upper(colname), or str_to_title()\n\nTo smush together phrases, use str_c() or paste() or paste0.\n\n\n\nConvert everything to lowercase\nOne strategy for using regular expressions is to convert everything to lower case before trying to match text. In R, the regular expression is, by default, case-sensitive and it’s difficult to predict how it might show up in free text like this. 1\n\ntrump_tweets &lt;-\n  trump_tweets |&gt; \n  mutate ( tweet_lowercase = str_to_lower(text))"
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.2 Sample data",
+    "text": "20.2 Sample data\nFor this example, we’ll use a small set of Tweets from Donald Trump, representing the last few months he was on on the platform. It runs from Nov. 1, 2020 to Jan. 8, 2021, when Twitter banned him. Retweets were removed, but there are a number of tweets that are just links to other content.\nHere is a random sample of the 1,112 tweets:\n\n\n\n\n\n\n\nUsing tweets as a source for data is a little dated, but it’s a good example of trying to pick out text from a set of short items.\n\nString functions for regular expressions\n\nstr_detect( col_name, expression) looks for the presence of the expression, and returns TRUE or FALSE.\nstr_extract(col_name, expression) does the same thing, except it returns whatever the first thing that matched your expression. It includes wild cards, so it could just return the whole phrase.\nstr_match(col_name, expression) is more complicated - it lets you use pieces of the expression, but it returns a list, which is hard to work with. We’ll skip that for now.\n\nSome other useful string functions are:\n\nTo convert the case, use str_to_lower(colname), str_to_upper(colname), or str_to_title()\n\nTo smush together phrases, use str_c() or paste() or paste0.\n\n\n\nConvert everything to lowercase\nOne strategy for using regular expressions is to convert everything to lower case before trying to match text. In R, the regular expression is, by default, case-sensitive and it’s difficult to predict how it might show up in free text like this. 1\n\ntrump_tweets &lt;-\n  trump_tweets |&gt; \n  mutate ( tweet_lowercase = str_to_lower(text))"
   },
   {
     "objectID": "advanced-regex-r.html#using-str_detect",
     "href": "advanced-regex-r.html#using-str_detect",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.3 Using str_detect",
-    "text": "19.3 Using str_detect\nThe function str_detect() results in either TRUE or FALSE. It’s very useful in two situations:\n\nAs a filter for rows in your data.\nAs a way to assign values in an if_else() or case_when() expression, when creating a new column using mutate().\n\n\nExample 1: Filter for “biden”\n\ntrump_tweets |&gt; \n  select ( tweet_date, tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"biden\"))\n\n\n\n  \n\n\n\nBut that leaves out any time he’s called “sleepy joe”. Instead, look for either “sleepy joe” or “joe biden”:\n\ntrump_tweets |&gt; \n  select ( tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"(biden|sleepy joe)\")) \n\n\n\n  \n\n\n\nNow, use either the reactable() or the DT() library to make a table that will show you the entire text:\n\ntrump_tweets |&gt; \n  select (  tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"(biden|sleepy joe)\")) |&gt;  \n  reactable( striped=TRUE, \n             searchable=TRUE) \n\n\n\n\n\n\n(When you have multi-word phrases, be sure to put the whole thing in parentheses. Otherwise, it doesn’t know how to interpret it: It could be (biden|sleepy) joe, which wouldn’t match much! )\n\n\nKeywords\nAnother way to look at these tweets is to look for signs of populism – words like “globalist” or “marxist”. Here’s one way to find a lot of them:\nThere are two or or three parts to a regular expression:\n\nThe pattern you are trying to find.\nAn optional replacement, which could use part of what you’ve found in the original seeking phase.\nOptions - the big one in R is negate=TRUE, which means to match everything EXCEPT what is found.\n\nIn practice, you’ll usually save the pieces of each pattern you’ve found into a variable, then put them back together differently.\nHere are two other good tutorials on regular expressions:\n\nFrom Justin Meyer at a recent IRE conference that can serve as a guide.\n\nFrom Prof. Christian McDonald of UT-Austin\nIf you’re using R, you can use a regex using the stringr package (part of the tidyverse) using the functions str_detect , str_extract and their cousins. They look like this:\n\n    str_detect(var_name, regex(\"pattern\")) \n\nThe vignette on the stringr package in the Tidyverse is reasonably helpful."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.3 Using str_detect",
+    "text": "20.3 Using str_detect\nThe function str_detect() results in either TRUE or FALSE. It’s very useful in two situations:\n\nAs a filter for rows in your data.\nAs a way to assign values in an if_else() or case_when() expression, when creating a new column using mutate().\n\n\nExample 1: Filter for “biden”\n\ntrump_tweets |&gt; \n  select ( tweet_date, tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"biden\"))\n\n\n\n  \n\n\n\nBut that leaves out any time he’s called “sleepy joe”. Instead, look for either “sleepy joe” or “joe biden”:\n\ntrump_tweets |&gt; \n  select ( tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"(biden|sleepy joe)\")) \n\n\n\n  \n\n\n\nNow, use either the reactable() or the DT() library to make a table that will show you the entire text:\n\ntrump_tweets |&gt; \n  select (  tweet_lowercase) |&gt; \n  filter (  str_detect ( tweet_lowercase, \"(biden|sleepy joe)\")) |&gt;  \n  reactable( striped=TRUE, \n             searchable=TRUE) \n\n\n\n\n\n\n(When you have multi-word phrases, be sure to put the whole thing in parentheses. Otherwise, it doesn’t know how to interpret it: It could be (biden|sleepy) joe, which wouldn’t match much! )\n\n\nKeywords\nAnother way to look at these tweets is to look for signs of populism – words like “globalist” or “marxist”. Here’s one way to find a lot of them:\nThere are two or or three parts to a regular expression:\n\nThe pattern you are trying to find.\nAn optional replacement, which could use part of what you’ve found in the original seeking phase.\nOptions - the big one in R is negate=TRUE, which means to match everything EXCEPT what is found.\n\nIn practice, you’ll usually save the pieces of each pattern you’ve found into a variable, then put them back together differently.\nHere are two other good tutorials on regular expressions:\n\nFrom Justin Meyer at a recent IRE conference that can serve as a guide.\n\nFrom Prof. Christian McDonald of UT-Austin\nIf you’re using R, you can use a regex using the stringr package (part of the tidyverse) using the functions str_detect , str_extract and their cousins. They look like this:\n\n    str_detect(var_name, regex(\"pattern\")) \n\nThe vignette on the stringr package in the Tidyverse is reasonably helpful."
   },
   {
     "objectID": "advanced-regex-r.html#types-of-regular-expression-patterns",
     "href": "advanced-regex-r.html#types-of-regular-expression-patterns",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.4 Types of regular expression patterns",
-    "text": "19.4 Types of regular expression patterns\n\nLiteral strings\nThese are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.\nYou can tell the regex that you want to find your pattern at the beginning or end of a line:\n  ^   = \"Find only at the beginning of a line\"\n  $   = \"Find only at the end of a line\"\n\n\nWild cards\nA wild card is a character you use to indicate the word “anything”. Here are some ways to use wild cards in regex:\n    .      = \"any single character of any type\"\n    .?     = \"a possible single character of any type (but it might not exist)\"\n    .*     = \"anything or nothing of any length\"\n    .+     = \"anything one or more times\"\n    .{1,3} = \"anything running between 1 and 3 characters long\"\nRegular expressions also have wild cards of specific types. In R, they are “escaped” using two backslashes. In other languages and in the example https://regex101.com site they only use one backslash:\n      \\\\d   = \"Any digit\"\n      \\\\w   = \"Any word character\"\n      \\\\s   = \"Any whitespace (tab, space, etc.)\"\n      \\\\b   = \"Any word boundary\" (period, comma, space, etc.)\nWhen you upper-case them, it’s the opposite:\n      \\\\D = \"Anything but a digit\"\n\n\nCharacter classes\nSometimes you want to tell the regex what characters it is allowed to accept. For example, say you don’t know whether there is an alternative spelling for a name – you can tell the regex to either ignore a character, or take one of several.\nIn R, we saw that there were alternative spellings for words like “summarize” – the British and the American spellings. You could, for example, use this pattern to pick up either spelling:\n        summari[sz]e\nThe bracket tells the regex that it’s allowed to take either one of those characters. You can also use ranges:\n  [a-zA-Z0-9]\nmeans that any lower case, upper case or numeric character is allowed.\n\n\nEscaping\nBecause they’re already being used for special purposes, some characters have to be “escaped” before you can search for them. Notably, they are parentheses (), periods, backslashes, dollar signs, question marks, dashes and carets.\nThis means that to find a period or question mark, you have to use the pattern\n    \\\\. or\n    \\\\?\nIn the Regex101, this is the biggest difference among the flavors of regex – Python generally requires the least amount of escaping.\n\n\nMatch groups\nUse parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with lists."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.4 Types of regular expression patterns",
+    "text": "20.4 Types of regular expression patterns\n\nLiteral strings\nThese are just letters, like “abc” or “Mary”. They are case-sensitive and no different than using text in a filter.\nYou can tell the regex that you want to find your pattern at the beginning or end of a line:\n  ^   = \"Find only at the beginning of a line\"\n  $   = \"Find only at the end of a line\"\n\n\nWild cards\nA wild card is a character you use to indicate the word “anything”. Here are some ways to use wild cards in regex:\n    .      = \"any single character of any type\"\n    .?     = \"a possible single character of any type (but it might not exist)\"\n    .*     = \"anything or nothing of any length\"\n    .+     = \"anything one or more times\"\n    .{1,3} = \"anything running between 1 and 3 characters long\"\nRegular expressions also have wild cards of specific types. In R, they are “escaped” using two backslashes. In other languages and in the example https://regex101.com site they only use one backslash:\n      \\\\d   = \"Any digit\"\n      \\\\w   = \"Any word character\"\n      \\\\s   = \"Any whitespace (tab, space, etc.)\"\n      \\\\b   = \"Any word boundary\" (period, comma, space, etc.)\nWhen you upper-case them, it’s the opposite:\n      \\\\D = \"Anything but a digit\"\n\n\nCharacter classes\nSometimes you want to tell the regex what characters it is allowed to accept. For example, say you don’t know whether there is an alternative spelling for a name – you can tell the regex to either ignore a character, or take one of several.\nIn R, we saw that there were alternative spellings for words like “summarize” – the British and the American spellings. You could, for example, use this pattern to pick up either spelling:\n        summari[sz]e\nThe bracket tells the regex that it’s allowed to take either one of those characters. You can also use ranges:\n  [a-zA-Z0-9]\nmeans that any lower case, upper case or numeric character is allowed.\n\n\nEscaping\nBecause they’re already being used for special purposes, some characters have to be “escaped” before you can search for them. Notably, they are parentheses (), periods, backslashes, dollar signs, question marks, dashes and carets.\nThis means that to find a period or question mark, you have to use the pattern\n    \\\\. or\n    \\\\?\nIn the Regex101, this is the biggest difference among the flavors of regex – Python generally requires the least amount of escaping.\n\n\nMatch groups\nUse parentheses within a pattern to pick out pieces, which you can then use over again. The end of this chapter shows how to do this in R, which is a little complicated because we haven’t done much with lists."
   },
   {
     "objectID": "advanced-regex-r.html#sample-data-1",
     "href": "advanced-regex-r.html#sample-data-1",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.5 Sample data",
-    "text": "19.5 Sample data\nHere are three small text files that you can copy and paste from your browser into the regex101.com site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.\n\nA list of phone numbers in different formats\nA list of dates that you need to convert into a different form.\nA list of addresses that are in multiple lines, and you need to pull out the pieces. (Courtesy of IRE)\nA small chunk of the H2B visa applications from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.5 Sample data",
+    "text": "20.5 Sample data\nHere are three small text files that you can copy and paste from your browser into the regex101.com site. It’s a site that lets you test out regular expressions, while explaining to you what’s happening with them.\n\nA list of phone numbers in different formats\nA list of dates that you need to convert into a different form.\nA list of addresses that are in multiple lines, and you need to pull out the pieces. (Courtesy of IRE)\nA small chunk of the H2B visa applications from Arizona companies or worksites that has been kind of messed up for this demonstration, in tab-delimited format."
   },
   {
     "objectID": "advanced-regex-r.html#testing-and-examining-patterns-with-regex101",
     "href": "advanced-regex-r.html#testing-and-examining-patterns-with-regex101",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.6 Testing and examining patterns with Regex101",
-    "text": "19.6 Testing and examining patterns with Regex101\nRegex 101 is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\\` in R.\n\nLooking for specific words or characters\nThe easiest regex is one that has just the characters you’re looking for when you know that they are the right case. They’re called literals because you are literally looking for those letters or characters."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.6 Testing and examining patterns with Regex101",
+    "text": "20.6 Testing and examining patterns with Regex101\nRegex 101 is a website that lets you copy part of your data into a box, then test different patterns to see how they get applied. Regular expressions are very difficult to write and to read, but Regex101 lets you do it a little piece at a time. Just remember that every time you use ‘' in regex101, you will need’\\` in R.\n\nLooking for specific words or characters\nThe easiest regex is one that has just the characters you’re looking for when you know that they are the right case. They’re called literals because you are literally looking for those letters or characters."
   },
   {
     "objectID": "advanced-regex-r.html#practice-1-extract-date-parts",
     "href": "advanced-regex-r.html#practice-1-extract-date-parts",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.7 Practice #1: Extract date parts",
-    "text": "19.7 Practice #1: Extract date parts\nIn Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.2\nWe want to turn dates that look like this:\n  1/24/2018\ninto something that looks like this:\n 2008-1-24\nCopy and paste these numbers into the regex 101 window:\n9/7/2017\n9/11/1998\n9/11/2017\n9/19/2018\n9/15/2017\n10/13/2019\n11/3/2017\nFirst, you can use any digit using the pattern “. Try to do it in pieces. First, see if you can find one or two digits at the beginning of the line.\n  ^\\d{1,2}\nTry coming up with the rest of it on your own before you type in the answer:\n  ^\\d{1,2}.\\d{1,2}.\\d{4}\n(This works because regular expressions normally are “greedy”. That is, if you tell it “one or two digits”, it will always take two if they exist.)\nPut parentheses around any pieces that you want to use for later:\n\n\n\n\n\nNow each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.7 Practice #1: Extract date parts",
+    "text": "20.7 Practice #1: Extract date parts\nIn Regex 101, change the “Flavor” to “Python” – otherwise, you have to escape more of the characters.2\nWe want to turn dates that look like this:\n  1/24/2018\ninto something that looks like this:\n 2008-1-24\nCopy and paste these numbers into the regex 101 window:\n9/7/2017\n9/11/1998\n9/11/2017\n9/19/2018\n9/15/2017\n10/13/2019\n11/3/2017\nFirst, you can use any digit using the pattern “. Try to do it in pieces. First, see if you can find one or two digits at the beginning of the line.\n  ^\\d{1,2}\nTry coming up with the rest of it on your own before you type in the answer:\n  ^\\d{1,2}.\\d{1,2}.\\d{4}\n(This works because regular expressions normally are “greedy”. That is, if you tell it “one or two digits”, it will always take two if they exist.)\nPut parentheses around any pieces that you want to use for later:\n\n\n\n\n\nNow each piece has its section, numbered 0 for the whole match, and then 1-3 for the pieces."
   },
   {
     "objectID": "advanced-regex-r.html#practice-2-extract-pieces-of-phone-numbers",
     "href": "advanced-regex-r.html#practice-2-extract-pieces-of-phone-numbers",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.8 Practice #2: Extract pieces of phone numbers",
-    "text": "19.8 Practice #2: Extract pieces of phone numbers\nHere are some phone numbers in different formats:\n    623-374-1167\n    760.352.5212\n    831-676-3833\n    (831)-676-3833\n    623-374-1167 ext 203\n    831-775-0370\n    (602)-955-0222  x20\n    928-627-8080\n    831-784-1453\nThis is a little more complicated than it looks, so try piecing together what this one says:\n      (\\d{3})[-.\\)]+(\\d{3})[-.]+(\\d{4})\n(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)\nAnything within parentheses will be “captured” in a block."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.8 Practice #2: Extract pieces of phone numbers",
+    "text": "20.8 Practice #2: Extract pieces of phone numbers\nHere are some phone numbers in different formats:\n    623-374-1167\n    760.352.5212\n    831-676-3833\n    (831)-676-3833\n    623-374-1167 ext 203\n    831-775-0370\n    (602)-955-0222  x20\n    928-627-8080\n    831-784-1453\nThis is a little more complicated than it looks, so try piecing together what this one says:\n      (\\d{3})[-.\\)]+(\\d{3})[-.]+(\\d{4})\n(This won’t work in the “substitute” area – it would be easier to create a new variable with the results than to replace the originals.)\nAnything within parentheses will be “captured” in a block."
   },
   {
     "objectID": "advanced-regex-r.html#practice-3-extract-address-pieces",
     "href": "advanced-regex-r.html#practice-3-extract-address-pieces",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.9 Practice #3: Extract address pieces",
-    "text": "19.9 Practice #3: Extract address pieces\nHere are a few lines of the data from Prof. McDonald’s tutorial, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)\n    \"10111 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.370945933000485, -97.6925542359997)\"\n    \"3636 N FM 620 RD\n    AUSTIN, TX 78734\n    (30.377873241000486, -97.9523496219997)\"\n    \"9919 SERVICE AVE\n    AUSTIN, TX 78743\n    (30.205028616000448, -97.6625588019997)\"\n    \"10601 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.37476574700048, -97.6903937089997)\"\n    \"801 E WILLIAM CANNON DR Unit 205\n    AUSTIN, TX 78745\n    (30.190914575000477, -97.77193838799968)\"\n    \"4408 LONG CHAMP DR\n    AUSTIN, TX 78746\n    (30.340981111000474, -97.7983147919997)\"\n    \"625 W BEN WHITE BLVD EB\n    AUSTIN, TX 78745\n    (30.206884239000487, -97.7956469989997)\"\n    \"3914 N LAMAR BLVD\n    AUSTIN, TX 78756\n    (30.307477098000447, -97.74169675199965)\"\n    \"15201 FALCON HEAD BLVD\n    BEE CAVE, TX 78738\n    (30.32068282700044, -97.96890311999965)\"\n    \"11905 FM 2244 RD Unit 100\n    BEE CAVE, TX 78738\n    (30.308363203000454, -97.92393357799966)\"\n    \"3801 JUNIPER TRCE\n    BEE CAVE, TX 78738\n    (30.308247975000484, -97.93511531999968)\"\n    \"12800 GALLERIA CIR Unit 101\n    BEE CAVE, TX 78738\n    (30.307996778000472, -97.94065088199966)\"\n    \"12400 W SH 71 Unit 510\n    BEE CAVE, TX 78733\n    (30.330682136000462, -97.86979886299969)\"\n    \"716 W 6TH ST\n    AUSTIN, TX 78701\n    (30.27019732500048, -97.75036306299967)\"\n    \"3003 BEE CAVES RD\n    ROLLINGWOOD, TX 78746\n    (30.271592738000436, -97.79583786499967)\""
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.9 Practice #3: Extract address pieces",
+    "text": "20.9 Practice #3: Extract address pieces\nHere are a few lines of the data from Prof. McDonald’s tutorial, which you can copy and paste to go his exercise. (He uses the Javascript version of regular expressions, but for our purposes in this exercise, it doesn’t matter which one you use. If you choose Python, you’ll have one extra step, of putting a slash () before the quotes. The colors work a little better if you leave it on the default PHP method.)\n    \"10111 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.370945933000485, -97.6925542359997)\"\n    \"3636 N FM 620 RD\n    AUSTIN, TX 78734\n    (30.377873241000486, -97.9523496219997)\"\n    \"9919 SERVICE AVE\n    AUSTIN, TX 78743\n    (30.205028616000448, -97.6625588019997)\"\n    \"10601 N LAMAR BLVD\n    AUSTIN, TX 78753\n    (30.37476574700048, -97.6903937089997)\"\n    \"801 E WILLIAM CANNON DR Unit 205\n    AUSTIN, TX 78745\n    (30.190914575000477, -97.77193838799968)\"\n    \"4408 LONG CHAMP DR\n    AUSTIN, TX 78746\n    (30.340981111000474, -97.7983147919997)\"\n    \"625 W BEN WHITE BLVD EB\n    AUSTIN, TX 78745\n    (30.206884239000487, -97.7956469989997)\"\n    \"3914 N LAMAR BLVD\n    AUSTIN, TX 78756\n    (30.307477098000447, -97.74169675199965)\"\n    \"15201 FALCON HEAD BLVD\n    BEE CAVE, TX 78738\n    (30.32068282700044, -97.96890311999965)\"\n    \"11905 FM 2244 RD Unit 100\n    BEE CAVE, TX 78738\n    (30.308363203000454, -97.92393357799966)\"\n    \"3801 JUNIPER TRCE\n    BEE CAVE, TX 78738\n    (30.308247975000484, -97.93511531999968)\"\n    \"12800 GALLERIA CIR Unit 101\n    BEE CAVE, TX 78738\n    (30.307996778000472, -97.94065088199966)\"\n    \"12400 W SH 71 Unit 510\n    BEE CAVE, TX 78733\n    (30.330682136000462, -97.86979886299969)\"\n    \"716 W 6TH ST\n    AUSTIN, TX 78701\n    (30.27019732500048, -97.75036306299967)\"\n    \"3003 BEE CAVES RD\n    ROLLINGWOOD, TX 78746\n    (30.271592738000436, -97.79583786499967)\""
   },
   {
     "objectID": "advanced-regex-r.html#on-your-own",
     "href": "advanced-regex-r.html#on-your-own",
-    "title": "19  Inexact matching and regular expressions in R",
-    "section": "19.10 On your own",
-    "text": "19.10 On your own\nThis is a small list of H2A visa applications, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.\nYou might try:\n\nFinding all of the LLC’s in the list (limited liability companies) of names. (You should turn on the case-insensitive flag in Regex 101 or set that flag in your program if you do.)\nSee how far you can get in standardizing the addresses.\nSplit the city, state and zip code of the worksite.\nFind all of the jobs related to field crops such as lettuce or celery."
+    "title": "20  Inexact matching and regular expressions in R",
+    "section": "20.10 On your own",
+    "text": "20.10 On your own\nThis is a small list of H2A visa applications, which are requests for agricultural and seasonal workers, from companies or worksites in Arizona. Try importing it into Excel, then copying some of the cells to practice your regular expression skills.\nYou might try:\n\nFinding all of the LLC’s in the list (limited liability companies) of names. (You should turn on the case-insensitive flag in Regex 101 or set that flag in your program if you do.)\nSee how far you can get in standardizing the addresses.\nSplit the city, state and zip code of the worksite.\nFind all of the jobs related to field crops such as lettuce or celery."
   },
   {
     "objectID": "advanced-regex-r.html#footnotes",
     "href": "advanced-regex-r.html#footnotes",
-    "title": "19  Inexact matching and regular expressions in R",
+    "title": "20  Inexact matching and regular expressions in R",
     "section": "",
     "text": "Yes, we might care about the case, isolating places where Trump uses all-caps. But for now let’s deal with the substance of the tweets.↩︎\nEach language implements regular expressions slightly differently – when you begin to learn more languages, this will be one of the first things you’ll need to look up.↩︎"
   },
   {
     "objectID": "r-recipes.html#re-using-answers-from-a-previous-step",
     "href": "r-recipes.html#re-using-answers-from-a-previous-step",
-    "title": "20  Recipes",
-    "section": "20.1 Re-using answers from a previous step",
-    "text": "20.1 Re-using answers from a previous step\nAssign your code chunk to a new variable in the environment to use it again later. This is usefule when:\n\nYou have a complex filter that you want to apply to future steps.\nYou are joining data frames with columns having the same names: select just the columns you need and re-name any of the ones that are the same, except for the join one.\nYou have fixed some column types or adjusted their values.\n\nThe assignment operator is &lt;-, which means “pour the answer into this variable name”\nThese can all go into one code chunk if you’d like:\n\nselect_ppp &lt;- \n  ppp_orig |&gt;\n  filter ( project_state == \"AZ\" & !is.na (project_county) ) |&gt;\n  mutate ( forgiveness_amount = if_else (is.na(forgiveness_amount),\n                                         0, \n                                         forgiveness_amount)) |&gt;\n  select ( borrower_name, borrower_city, \n           naics_code, \n           amount, forgiveness_amount,\n           date_approved, forgiveness_date)\n\n\n\nsectors &lt;- \n  naics_codes |&gt;\n  select ( naics_code, sector_code, sector_desc)\n\n\nselect_ppp |&gt;\n  left_join ( sectors, by=\"naics_code\")  |&gt;\n  select ( borrower_name, forgiveness_amount, sector_desc) |&gt;\n  head(5)"
+    "title": "21  Recipes",
+    "section": "21.1 Re-using answers from a previous step",
+    "text": "21.1 Re-using answers from a previous step\nAssign your code chunk to a new variable in the environment to use it again later. This is usefule when:\n\nYou have a complex filter that you want to apply to future steps.\nYou are joining data frames with columns having the same names: select just the columns you need and re-name any of the ones that are the same, except for the join one.\nYou have fixed some column types or adjusted their values.\n\nThe assignment operator is &lt;-, which means “pour the answer into this variable name”\nThese can all go into one code chunk if you’d like:\n\nselect_ppp &lt;- \n  ppp_orig |&gt;\n  filter ( project_state == \"AZ\" & !is.na (project_county) ) |&gt;\n  mutate ( forgiveness_amount = if_else (is.na(forgiveness_amount),\n                                         0, \n                                         forgiveness_amount)) |&gt;\n  select ( borrower_name, borrower_city, \n           naics_code, \n           amount, forgiveness_amount,\n           date_approved, forgiveness_date)\n\n\n\nsectors &lt;- \n  naics_codes |&gt;\n  select ( naics_code, sector_code, sector_desc)\n\n\nselect_ppp |&gt;\n  left_join ( sectors, by=\"naics_code\")  |&gt;\n  select ( borrower_name, forgiveness_amount, sector_desc) |&gt;\n  head(5)"
   },
   {
     "objectID": "r-recipes.html#importing-reading-data",
     "href": "r-recipes.html#importing-reading-data",
-    "title": "20  Recipes",
-    "section": "20.2 Importing / reading data",
-    "text": "20.2 Importing / reading data\n\nIf you are referring to a file on your computer or on the web, then it must be in quotes.\nUse the proper library and function to read the data.\nCheck the output for proper treatment of text, dates and numbers, especially with Excel and text imports\n\n\nReading R files\nTwo types of data files: RDS files contain one data frame, which you have to assign to a new object. Rda files have multiple objects that are already named. Use readRDS ( url ( ....) ) to read data stored on the internet instead of in your project.\nmydata &lt;- readRDS ( \"mydata_file.RDS\")\nload ( \"mydata_file.Rda\")\n\n\n\nReading text (csv) files\nUse read_csv(), read_tsv(), read_delimited() and various fixed file formats. Match the type of your file with the proper function. These require that the tidyverse or readr libraries are already loaded.\n\n\nReading Excel files\nlibrary (readxl)          # put this with the rest of your libraries.\nmydata &lt;- read_excel(\"mydata_file.xlsx\", \n                      sheet = \"sheet_name\")  \nThere is no url() option for reading Excel files. They must be stored in your project. One approach is to download the file in your code:\ndownload.file ( \"https://address-of-file.xlsx\", \"myfilename.xlsx\")\nmydata &lt;- read_excel (\"myfilename.xlsx\")\n\n\nImport options\nThere are options in most of the importing functions that tell you how to treat each column and let you rename at the same time. One approach is :\nread_csv ( \"mydata_file.csv\", \n           coltypes = col(.default = \"c\")\n           )\n           \nTo keep everything as text, which you can convert later."
+    "title": "21  Recipes",
+    "section": "21.2 Importing / reading data",
+    "text": "21.2 Importing / reading data\n\nIf you are referring to a file on your computer or on the web, then it must be in quotes.\nUse the proper library and function to read the data.\nCheck the output for proper treatment of text, dates and numbers, especially with Excel and text imports\n\n\nReading R files\nTwo types of data files: RDS files contain one data frame, which you have to assign to a new object. Rda files have multiple objects that are already named. Use readRDS ( url ( ....) ) to read data stored on the internet instead of in your project.\nmydata &lt;- readRDS ( \"mydata_file.RDS\")\nload ( \"mydata_file.Rda\")\n\n\n\nReading text (csv) files\nUse read_csv(), read_tsv(), read_delimited() and various fixed file formats. Match the type of your file with the proper function. These require that the tidyverse or readr libraries are already loaded.\n\n\nReading Excel files\nlibrary (readxl)          # put this with the rest of your libraries.\nmydata &lt;- read_excel(\"mydata_file.xlsx\", \n                      sheet = \"sheet_name\")  \nThere is no url() option for reading Excel files. They must be stored in your project. One approach is to download the file in your code:\ndownload.file ( \"https://address-of-file.xlsx\", \"myfilename.xlsx\")\nmydata &lt;- read_excel (\"myfilename.xlsx\")\n\n\nImport options\nThere are options in most of the importing functions that tell you how to treat each column and let you rename at the same time. One approach is :\nread_csv ( \"mydata_file.csv\", \n           coltypes = col(.default = \"c\")\n           )\n           \nTo keep everything as text, which you can convert later."
   },
   {
     "objectID": "r-recipes.html#filtering",
     "href": "r-recipes.html#filtering",
-    "title": "20  Recipes",
-    "section": "20.3 Filtering",
-    "text": "20.3 Filtering\n\nOne condition that’s a number :\n\n  filter ( amount &gt; 1000 )\n\nA number between two values\n\n  filter ( between ( amount, 0, 1000))\n\nAn exact phrase or word\n\n  filter ( project_county == \"MARICOPA\")\n\nOne of several possible entries in one column (exactly)\n\n    filter ( project_county %in% c(\"MARICOPA\", \"PIMA\", \"PINAL\" ))\n\nEverything except missing values\n\n    filter ( ! is.na (project_county) )\n\nBetween two dates . Be sure it’s really a date in the data by glimpsing your data frame. If not, turn it into a date first.\n\n    filter ( approval_date &gt;= \"2021-01-24\" & \n           approval_date &lt;= \"2021-01-31\")\n\nPhrases, words or letters at the beginning of a column\n\n    filter ( str_like (borrower_type , \"Non-Profit%\"))\n\nPhrases, letters or words at the end of a column\n\n    filter ( str_like  (borrower_type, \"Corporation%\"))\nAll of these examples can be used in a mutate statement to create flags or new values if the conditions are met."
+    "title": "21  Recipes",
+    "section": "21.3 Filtering",
+    "text": "21.3 Filtering\n\nOne condition that’s a number :\n\n  filter ( amount &gt; 1000 )\n\nA number between two values\n\n  filter ( between ( amount, 0, 1000))\n\nAn exact phrase or word\n\n  filter ( project_county == \"MARICOPA\")\n\nOne of several possible entries in one column (exactly)\n\n    filter ( project_county %in% c(\"MARICOPA\", \"PIMA\", \"PINAL\" ))\n\nEverything except missing values\n\n    filter ( ! is.na (project_county) )\n\nBetween two dates . Be sure it’s really a date in the data by glimpsing your data frame. If not, turn it into a date first.\n\n    filter ( approval_date &gt;= \"2021-01-24\" & \n           approval_date &lt;= \"2021-01-31\")\n\nPhrases, words or letters at the beginning of a column\n\n    filter ( str_like (borrower_type , \"Non-Profit%\"))\n\nPhrases, letters or words at the end of a column\n\n    filter ( str_like  (borrower_type, \"Corporation%\"))\nAll of these examples can be used in a mutate statement to create flags or new values if the conditions are met."
   },
   {
     "objectID": "r-recipes.html#aggregate-count-sum-rows-by-category",
     "href": "r-recipes.html#aggregate-count-sum-rows-by-category",
-    "title": "20  Recipes",
-    "section": "20.4 Aggregate (count, sum rows by category)",
-    "text": "20.4 Aggregate (count, sum rows by category)\n\nCounting (How many?)\n\n    group_by (project_county) |&gt;\n    summarize (loans = n() )  |&gt;\n    arrange ( desc ( loans ))\nMake sure you don’t name the new column containing the count the same thing as a group_by() column.\n\nShortcut for counting:\n\n     count( project_county, sort=TRUE, name=\"loans\")\n\nCounting unique entries\n\nSometimes you want to know how many items of a type, not how many rows, are included in a category.\n        group_by ( project_county) |&gt;\n        summarize ( number_of_loans = n(), \n                    number_of_industries = n_distinct ( naics_code )\n                    )\n\nSumming (how much?)\n\n    group_by (project_county) |&gt;\n    summarize ( total_amount = sum (amount, na.rm=T))\n\nShortcut for summing:\n\n    count ( project_county, weight=amount, name=\"total_amount\")"
+    "title": "21  Recipes",
+    "section": "21.4 Aggregate (count, sum rows by category)",
+    "text": "21.4 Aggregate (count, sum rows by category)\n\nCounting (How many?)\n\n    group_by (project_county) |&gt;\n    summarize (loans = n() )  |&gt;\n    arrange ( desc ( loans ))\nMake sure you don’t name the new column containing the count the same thing as a group_by() column.\n\nShortcut for counting:\n\n     count( project_county, sort=TRUE, name=\"loans\")\n\nCounting unique entries\n\nSometimes you want to know how many items of a type, not how many rows, are included in a category.\n        group_by ( project_county) |&gt;\n        summarize ( number_of_loans = n(), \n                    number_of_industries = n_distinct ( naics_code )\n                    )\n\nSumming (how much?)\n\n    group_by (project_county) |&gt;\n    summarize ( total_amount = sum (amount, na.rm=T))\n\nShortcut for summing:\n\n    count ( project_county, weight=amount, name=\"total_amount\")"
   },
   {
     "objectID": "r-recipes.html#recoding-categories",
     "href": "r-recipes.html#recoding-categories",
-    "title": "20  Recipes",
-    "section": "20.5 Recoding categories",
-    "text": "20.5 Recoding categories\nYou’ll often want fewer categories, or numbers in categories, that you want to use instead of the original values. This is done in a mutate statement. Don’t forget to save the output to a new data frame (&lt;-), or you won’t have access to it later on.\n\nCreate yes-no categories . This is really “Yes”, “No” or NA, where there is an NA to begin with.\n\n\n    mutate ( corp_yn = if_else  \n                      (str_detect (borrower_type, \n                                   \"Corporation\"), \n                       \"Yes\", \n                       \"No\")\n       )\n\n\nRecode into more than two categories using case_when()\n\nThis example introduces str_detect(), which uses regular expressions. We’ll go over that later in the book.\n  mutate (  new_business_type = \n                case_when (  str_like(business_type, \"%non_profit%\") ~ \"Non-profit\",    #1st category\n\n                             business_type %in% \n                               c(\"Independent Contractors\", \n                                \"Sole Proprietorship\", \n                                \"Self-Employed Individuals\", \n                                  \"Single Member LLC\")              ~ \"Individual\",    # 2nd category\n                             \n                             business_type == \"Tribal Concerns\"     ~ \"Tribal concerns\",    #3rd category\n                             \n                             str_detect (business_type, \"LLC|Company|Corporation|Partnership\") ~ \"Companies\",  #4th category\n                             \n                             TRUE ~ \"Other\")         #catchall 5th category\n            )\n\nRecode numbers into categories\n\nIt’s often useful to give them numeric codes in front so they sort properly:\n\n        mutate ( new_type = \n                 case_when  (\n                    amount &lt;= 1000 ~ \"00-Very low\", \n                    amount &lt;= 10000  ~ \"01-Low\", \n                    amount &lt;= 100000 ~ \"03-Medium\", \n                    amount &gt; 100000 ~ \"04-High\")\n              )\n\nThis works because the first one that it finds will be used, so a value of exactly 1,000 would be “Very low”, but a value of 1,001 would be “Low”."
+    "title": "21  Recipes",
+    "section": "21.5 Recoding categories",
+    "text": "21.5 Recoding categories\nYou’ll often want fewer categories, or numbers in categories, that you want to use instead of the original values. This is done in a mutate statement. Don’t forget to save the output to a new data frame (&lt;-), or you won’t have access to it later on.\n\nCreate yes-no categories . This is really “Yes”, “No” or NA, where there is an NA to begin with.\n\n\n    mutate ( corp_yn = if_else  \n                      (str_detect (borrower_type, \n                                   \"Corporation\"), \n                       \"Yes\", \n                       \"No\")\n       )\n\n\nRecode into more than two categories using case_when()\n\nThis example introduces str_detect(), which uses regular expressions. We’ll go over that later in the book.\n  mutate (  new_business_type = \n                case_when (  str_like(business_type, \"%non_profit%\") ~ \"Non-profit\",    #1st category\n\n                             business_type %in% \n                               c(\"Independent Contractors\", \n                                \"Sole Proprietorship\", \n                                \"Self-Employed Individuals\", \n                                  \"Single Member LLC\")              ~ \"Individual\",    # 2nd category\n                             \n                             business_type == \"Tribal Concerns\"     ~ \"Tribal concerns\",    #3rd category\n                             \n                             str_detect (business_type, \"LLC|Company|Corporation|Partnership\") ~ \"Companies\",  #4th category\n                             \n                             TRUE ~ \"Other\")         #catchall 5th category\n            )\n\nRecode numbers into categories\n\nIt’s often useful to give them numeric codes in front so they sort properly:\n\n        mutate ( new_type = \n                 case_when  (\n                    amount &lt;= 1000 ~ \"00-Very low\", \n                    amount &lt;= 10000  ~ \"01-Low\", \n                    amount &lt;= 100000 ~ \"03-Medium\", \n                    amount &gt; 100000 ~ \"04-High\")\n              )\n\nThis works because the first one that it finds will be used, so a value of exactly 1,000 would be “Very low”, but a value of 1,001 would be “Low”."
   },
   {
     "objectID": "r-recipes.html#working-with-grouped-data-for-subtotals-changes-percents",
     "href": "r-recipes.html#working-with-grouped-data-for-subtotals-changes-percents",
-    "title": "20  Recipes",
-    "section": "20.6 Working with grouped data for subtotals, changes, percents",
-    "text": "20.6 Working with grouped data for subtotals, changes, percents\n\nPercent of total by group\n\n  group_by ( project_county, sector_desc ) |&gt;\n  summarize ( loans = n() , .groups=\"drop_last\") |&gt;\n  mutate ( pct_in_county = loans / sum(loans))\n\nDisplay results as in spreadsheet form\n\nTo see the items across the top, use pivot_wider.\n  group_by ( project_county, new_type) |&gt;\n  summarize ( loan_count = n() ) |&gt;\n  pivot_wider ( names_from = new_type, \n                id_cols = c( project_county), \n                values_from = loan_count)\nYou can add an argument after values_from if you know that any missing values are zero, by using values_fill=0\nYou usually only choose one column to show down the side, one column to spread across the top, and one column to display the value.\n\nComplex example: Get the change by year within each county.\nannual_ppp &lt;- mutate ( approve_year = year (approval_date ) ) |&gt; group_by ( project_county, approve_year) |&gt; summarize ( loans = n() , amount = sum( amount, na.rm=T), .groups=“drop_last”)\n\nNext, compare them within groups. New functions introduced: complete(), which fills in missing information in a sequence. For example, if there were missing years by county, it would create a row to fill it in. lag() refers to the previous item in a group. First, make a new, summarized data frame with just the columns you nee\n    complete ( project_county, approve_year, \n              fill= list (amount=0, loans =0) )|&gt;\n    mutate ( change_loans = loans - lag(loans), \n             pct_change_loans = change_loans / lag(loans) * 100) \n\nPick out the last item in a group, with all of its columns. New verb introduced : slice_tail() . This is particularly useful for chronological events, such as the last thing that happened in a court case, or the most recent complaint against a police officer. This example isn’t a great one, but it gets you the name and other details of the most recent loan for each lender\narrange ( lender, date_approved) |&gt;\ngroup_by ( lender) |&gt;\nslice_tail(n=1)\n\nNOTE: Don’t use slice_max(), because it will give you back multiple rows in the event of ties."
+    "title": "21  Recipes",
+    "section": "21.6 Working with grouped data for subtotals, changes, percents",
+    "text": "21.6 Working with grouped data for subtotals, changes, percents\n\nPercent of total by group\n\n  group_by ( project_county, sector_desc ) |&gt;\n  summarize ( loans = n() , .groups=\"drop_last\") |&gt;\n  mutate ( pct_in_county = loans / sum(loans))\n\nDisplay results as in spreadsheet form\n\nTo see the items across the top, use pivot_wider.\n  group_by ( project_county, new_type) |&gt;\n  summarize ( loan_count = n() ) |&gt;\n  pivot_wider ( names_from = new_type, \n                id_cols = c( project_county), \n                values_from = loan_count)\nYou can add an argument after values_from if you know that any missing values are zero, by using values_fill=0\nYou usually only choose one column to show down the side, one column to spread across the top, and one column to display the value.\n\nComplex example: Get the change by year within each county.\nannual_ppp &lt;- mutate ( approve_year = year (approval_date ) ) |&gt; group_by ( project_county, approve_year) |&gt; summarize ( loans = n() , amount = sum( amount, na.rm=T), .groups=“drop_last”)\n\nNext, compare them within groups. New functions introduced: complete(), which fills in missing information in a sequence. For example, if there were missing years by county, it would create a row to fill it in. lag() refers to the previous item in a group. First, make a new, summarized data frame with just the columns you nee\n    complete ( project_county, approve_year, \n              fill= list (amount=0, loans =0) )|&gt;\n    mutate ( change_loans = loans - lag(loans), \n             pct_change_loans = change_loans / lag(loans) * 100) \n\nPick out the last item in a group, with all of its columns. New verb introduced : slice_tail() . This is particularly useful for chronological events, such as the last thing that happened in a court case, or the most recent complaint against a police officer. This example isn’t a great one, but it gets you the name and other details of the most recent loan for each lender\narrange ( lender, date_approved) |&gt;\ngroup_by ( lender) |&gt;\nslice_tail(n=1)\n\nNOTE: Don’t use slice_max(), because it will give you back multiple rows in the event of ties."
   },
   {
     "objectID": "r-recipes.html#joining-tables",
     "href": "r-recipes.html#joining-tables",
-    "title": "20  Recipes",
-    "section": "20.7 Joining tables",
-    "text": "20.7 Joining tables\nNew options in the dplyr library that contains the join verb let us use the same operators in joining that you use in filtering: ==, &lt;=, etc. Usually you will use the “==” operator. Requires dplyr 1.1, which was released in February 2023.\nppp_orig |&gt; \n  select ( borrower_name, amount, naics_code) |&gt; \n  inner_join ( naics_codes, \n               join_by (naics_code == naics_code) \n  )"
+    "title": "21  Recipes",
+    "section": "21.7 Joining tables",
+    "text": "21.7 Joining tables\nNew options in the dplyr library that contains the join verb let us use the same operators in joining that you use in filtering: ==, &lt;=, etc. Usually you will use the “==” operator. Requires dplyr 1.1, which was released in February 2023.\nppp_orig |&gt; \n  select ( borrower_name, amount, naics_code) |&gt; \n  inner_join ( naics_codes, \n               join_by (naics_code == naics_code) \n  )"
   },
   {
     "objectID": "r-recipes.html#quarto-document-tips",
     "href": "r-recipes.html#quarto-document-tips",
-    "title": "20  Recipes",
-    "section": "20.8 Quarto document tips",
-    "text": "20.8 Quarto document tips\n\nYAML (front matter)\nThe YAML is very picky about exact casing, spacing and other details. If you try to render and you get a YAML error, it is likely at the top.\nTypical YAML front matter that goes within the three dashes beginning on the very first line:\n\ntitle: \"Name of document\"\nauthor: \"Your name\"\noutput: \n  html: \n    theme: cosmo\n    code-tools: true\n    embed-resources: true\n    toc: true\nexecute: \n   warning: true\n   error: true\n\n\nCode chunk options\nThese must be the very first rows in your code chunk, and must start in the first position, with spacing as shown. Use #| label: setup in the first code chunk to run it automatically every time you start up R.\n\n\nNothing runs; everything is gray\nYou probably erased the last three back-ticks at the end of the code chunk. You may need to restart R to get it re-set."
+    "title": "21  Recipes",
+    "section": "21.8 Quarto document tips",
+    "text": "21.8 Quarto document tips\n\nYAML (front matter)\nThe YAML is very picky about exact casing, spacing and other details. If you try to render and you get a YAML error, it is likely at the top.\nTypical YAML front matter that goes within the three dashes beginning on the very first line:\n\ntitle: \"Name of document\"\nauthor: \"Your name\"\noutput: \n  html: \n    theme: cosmo\n    code-tools: true\n    embed-resources: true\n    toc: true\nexecute: \n   warning: true\n   error: true\n\n\nCode chunk options\nThese must be the very first rows in your code chunk, and must start in the first position, with spacing as shown. Use #| label: setup in the first code chunk to run it automatically every time you start up R.\n\n\nNothing runs; everything is gray\nYou probably erased the last three back-ticks at the end of the code chunk. You may need to restart R to get it re-set."
   },
   {
     "objectID": "vis.html#in-this-section",
@@ -870,149 +926,149 @@
   {
     "objectID": "vis-reporting.html#examples",
     "href": "vis-reporting.html#examples",
-    "title": "21  Visualization as a reporting tool",
-    "section": "21.1 Examples",
-    "text": "21.1 Examples\n\n\n\nSize and growth at once\nThis graphic was used for less than five minutes, but helped us see that there was one company that would be a good candidate for further reporting. The details are unimportant, but it showed that this company began quite small and grew quite quickly — the characteristic we were seeking. The others were not as strong candidates.\n\n\n\n\n\n\n\n\nReviewing two data points in a long list\n\n\n\nThis graphic shows the difference between the percent of businesses that got PPP loans in majority Latino zip codes compared to majority white non-Hispanic, based on four different measures. We used it at Reveal to identify places that might be the right place for a story – those where, across a series of measures, the difference was noticable, and the number of loans was large enough to warrant investigating. The team decided to center the story in Los Angeles, though it also reported on Dallas.\n\n\n\nText with graphics\nLooking at a table of numbers is mind-numbing. It’s hard to pick out what’s important. Some of the most imaginative visualizations come in the form of tables. In fact, there’s even a contest among R users to build the best tables. 1\nOne simple example is this analysis of federal prosecutions of gun charges over three administrations. We were curious: which statutes are the ones that Joe Biden is prioritizing compared with previous administrations? And was there a major downturn in prosecutions during the pandemic? These “sparklines” on the data table help show the trends, even when the numbers themselves are very different:\n\n\n\nNetworks of companies, people or programs\nThis is an example of a network analysis of Medicare charges among nephrologists – kidney doctors. We had heard that one consultant was showing doctors how to do procedures in their offices that were traditionally done in hospitals. The procedures had unusually high billing rates in doctors’ offices, but very low ones in hospital settings. These networks looked at the similarity across doctors’ practices by comparing the proportions of billing codes, then clustering them when they matched about 75 percent of the time. The area of red, large circles on the right turned out to be the same consulting practice. Sadly, the reporter who worked on this tip left the paper before we had a chance to publish."
+    "title": "22  Visualization as a reporting tool",
+    "section": "22.1 Examples",
+    "text": "22.1 Examples\n\n\n\nSize and growth at once\nThis graphic was used for less than five minutes, but helped us see that there was one company that would be a good candidate for further reporting. The details are unimportant, but it showed that this company began quite small and grew quite quickly — the characteristic we were seeking. The others were not as strong candidates.\n\n\n\n\n\n\n\n\nReviewing two data points in a long list\n\n\n\nThis graphic shows the difference between the percent of businesses that got PPP loans in majority Latino zip codes compared to majority white non-Hispanic, based on four different measures. We used it at Reveal to identify places that might be the right place for a story – those where, across a series of measures, the difference was noticable, and the number of loans was large enough to warrant investigating. The team decided to center the story in Los Angeles, though it also reported on Dallas.\n\n\n\nText with graphics\nLooking at a table of numbers is mind-numbing. It’s hard to pick out what’s important. Some of the most imaginative visualizations come in the form of tables. In fact, there’s even a contest among R users to build the best tables. 1\nOne simple example is this analysis of federal prosecutions of gun charges over three administrations. We were curious: which statutes are the ones that Joe Biden is prioritizing compared with previous administrations? And was there a major downturn in prosecutions during the pandemic? These “sparklines” on the data table help show the trends, even when the numbers themselves are very different:\n\n\n\nNetworks of companies, people or programs\nThis is an example of a network analysis of Medicare charges among nephrologists – kidney doctors. We had heard that one consultant was showing doctors how to do procedures in their offices that were traditionally done in hospitals. The procedures had unusually high billing rates in doctors’ offices, but very low ones in hospital settings. These networks looked at the similarity across doctors’ practices by comparing the proportions of billing codes, then clustering them when they matched about 75 percent of the time. The area of red, large circles on the right turned out to be the same consulting practice. Sadly, the reporter who worked on this tip left the paper before we had a chance to publish."
   },
   {
     "objectID": "vis-reporting.html#tips-for-exploratory-visualizations",
     "href": "vis-reporting.html#tips-for-exploratory-visualizations",
-    "title": "21  Visualization as a reporting tool",
-    "section": "21.2 Tips for exploratory visualizations",
-    "text": "21.2 Tips for exploratory visualizations\n\nUse small multiples to quickly orient yourself in a large dataset\n“Small multiples” are repeated images, usually over time or place. This example shows where each round of agricultural disaster payments went, which helped us decide on where to focus our reporting – Kansas and North Dakota.\n\n\n\nLook at your data from all directions\nWhen you’re trying to understand a story or a dataset, there’s no wrong way to look at it; try it every way you can think of, and you’ll get a different perspective. If you’re reporting on crime, you might look at one set of charts with change in violent crimes in a year; another might be the percent change; the other might be a comparison to other cities; and another might be a change over time. Use raw numbers, percentages and indexes.\nLook at them on different scales. Try following the rule that the x-axis must be zero. Then break that rule and see if you learn more. Try out logarithms and square roots for data with odd distributions.\nKeep in mind the research done on visual perception. William Cleveland’s experiments showed that the eye sees change in an image when the average slope is about 45 degrees. This suggests you ignore the admonitions to always start at zero and instead work toward the most insightful graphic. Other research in epidemiology has suggested you find a target level as a boundary for your chart. Each of these ways helps you see the data in different ways. When they’ve stopped telling you anything new, you know you’re done.\n\n\nDon’t assume\nNow that you’ve looked at your data a variety of ways, you’ve probably found records that don’t seem right – you may not understand what they meant in the first place, or there are some outliers that seem like they are typos, or there are trends that seem backwards.\nIf you want to publish anything based on your early exploration or in a published visualization, you have to resolve these questions and you can’t make assumptions. They’re either interesting stories or mistakes; interesting challenges to common wisdom or misunderstanding.\nIt’s not unusual for local governments to provide spreadsheets filled with errors, and it’s also easy to misunderstand government jargon in a dataset.\nFirst, walk back your own work. Have you read the documentation, its caveats and does the problem exist in the original version of the data? If everything on your end seems right, then it’s time to pick up the phone. You’re going to have to get it resolved if you plan to use it, so you might as well get started now.\nThat said, not every mistake is important. In campaign finance records, it’s common to have several hundred postal codes that don’t exist in a database of 100,000 records. As long as they’re not all in the same city or within a candidate, the occasional bad data record just doesn’t matter.\nAsk yourself: if I were to use this, would readers have a fundamentally accurate view of what the data says?\n\n\nAvoid obsessiong over precision and presentation details\nThe flip side of not asking enough questions is obsessing over precision before it matters. Your exploratory graphics should be generally correct, but don’t worry if you have various levels of rounding, if they don’t add up to exactly 100 percent or if you are missing one or two years’ data out of 20. This is part of the exploration process. You’ll still see the big trends and know what you have to collect before it’s time for publication.\nIn fact, you might consider taking away labeling and scale markers, much like the charts above, to even better get an overall sense of the data.\n\n\nCreate chronologies of cases and events\nAt the start of any complex story, begin building chronologies of key events and cases. These are easily done in a spreadsheet or a data collection system like Airtable. You will eventually need one.\n\n\nMeet with designers and graphics editors early and often\nBrainstorm about possible graphics with the artists and designers in your newsroom. They will have good ways to look at your data, suggestions of how it might work interactively, and know how to connect data and stories. It will make your reporting much easier if you know what you have to collect early on, or if you can alert your team that a graphic isn’t possible when you can’t collect it."
+    "title": "22  Visualization as a reporting tool",
+    "section": "22.2 Tips for exploratory visualizations",
+    "text": "22.2 Tips for exploratory visualizations\n\nUse small multiples to quickly orient yourself in a large dataset\n“Small multiples” are repeated images, usually over time or place. This example shows where each round of agricultural disaster payments went, which helped us decide on where to focus our reporting – Kansas and North Dakota.\n\n\n\nLook at your data from all directions\nWhen you’re trying to understand a story or a dataset, there’s no wrong way to look at it; try it every way you can think of, and you’ll get a different perspective. If you’re reporting on crime, you might look at one set of charts with change in violent crimes in a year; another might be the percent change; the other might be a comparison to other cities; and another might be a change over time. Use raw numbers, percentages and indexes.\nLook at them on different scales. Try following the rule that the x-axis must be zero. Then break that rule and see if you learn more. Try out logarithms and square roots for data with odd distributions.\nKeep in mind the research done on visual perception. William Cleveland’s experiments showed that the eye sees change in an image when the average slope is about 45 degrees. This suggests you ignore the admonitions to always start at zero and instead work toward the most insightful graphic. Other research in epidemiology has suggested you find a target level as a boundary for your chart. Each of these ways helps you see the data in different ways. When they’ve stopped telling you anything new, you know you’re done.\n\n\nDon’t assume\nNow that you’ve looked at your data a variety of ways, you’ve probably found records that don’t seem right – you may not understand what they meant in the first place, or there are some outliers that seem like they are typos, or there are trends that seem backwards.\nIf you want to publish anything based on your early exploration or in a published visualization, you have to resolve these questions and you can’t make assumptions. They’re either interesting stories or mistakes; interesting challenges to common wisdom or misunderstanding.\nIt’s not unusual for local governments to provide spreadsheets filled with errors, and it’s also easy to misunderstand government jargon in a dataset.\nFirst, walk back your own work. Have you read the documentation, its caveats and does the problem exist in the original version of the data? If everything on your end seems right, then it’s time to pick up the phone. You’re going to have to get it resolved if you plan to use it, so you might as well get started now.\nThat said, not every mistake is important. In campaign finance records, it’s common to have several hundred postal codes that don’t exist in a database of 100,000 records. As long as they’re not all in the same city or within a candidate, the occasional bad data record just doesn’t matter.\nAsk yourself: if I were to use this, would readers have a fundamentally accurate view of what the data says?\n\n\nAvoid obsessiong over precision and presentation details\nThe flip side of not asking enough questions is obsessing over precision before it matters. Your exploratory graphics should be generally correct, but don’t worry if you have various levels of rounding, if they don’t add up to exactly 100 percent or if you are missing one or two years’ data out of 20. This is part of the exploration process. You’ll still see the big trends and know what you have to collect before it’s time for publication.\nIn fact, you might consider taking away labeling and scale markers, much like the charts above, to even better get an overall sense of the data.\n\n\nCreate chronologies of cases and events\nAt the start of any complex story, begin building chronologies of key events and cases. These are easily done in a spreadsheet or a data collection system like Airtable. You will eventually need one.\n\n\nMeet with designers and graphics editors early and often\nBrainstorm about possible graphics with the artists and designers in your newsroom. They will have good ways to look at your data, suggestions of how it might work interactively, and know how to connect data and stories. It will make your reporting much easier if you know what you have to collect early on, or if you can alert your team that a graphic isn’t possible when you can’t collect it."
   },
   {
     "objectID": "vis-reporting.html#footnotes",
     "href": "vis-reporting.html#footnotes",
-    "title": "21  Visualization as a reporting tool",
+    "title": "22  Visualization as a reporting tool",
     "section": "",
     "text": "These are painstakingly built and are meant for publication, but you can see how much they can convey using color, size, shape and other elements within a table of numbers.↩︎"
   },
   {
     "objectID": "vis-tables.html#interactive-tables-with-reactable",
     "href": "vis-tables.html#interactive-tables-with-reactable",
-    "title": "22  Report-making in R",
-    "section": "22.1 Interactive tables with reactable",
-    "text": "22.1 Interactive tables with reactable\nThe reactable library is highly customizable, but it can involve a lot of typing to get a good table. The documentation is excellent and there are great examples in the wild. It was used to produce the table above that contained the sparklines.\nThe basic idea is that you select the columns you want to display, then define their format and other details one by one. There are several useful defaults for the whole table, such as the number format. One way to simplify the typing is to just create the names of the columns you want to see in a select statement, then put a default number format for the dollar amounts. A full list of options is available on the help page for reactable.\n\nppp_orig |&gt; \n1  select (\n    Borrower = borrower_name, \n    City = borrower_city, \n    `Zip code` = borrower_zip, \n    `Loan amount` = amount, \n    `Amt forgiven` = forgiveness_amount\n    ) |&gt; \n  reactable (   \n2     sortable = TRUE,\n     filterable = TRUE,\n     searchable = TRUE,\n     defaultPageSize = 5,\n3     compact = TRUE ,\n     striped=TRUE,\n4     defaultColDef =\n        colDef ( \n                format = \n                   colFormat ( currency=\"USD\", separators = TRUE, digits = 0)\n                )\n     )\n\n\n1\n\nusing the select verb that we’ve already seen, pick out the columns for the table and rename them. Don’t forget the back-ticks if you need something with more than one word.\n\n2\n\nSet the options for sorting, filtering, and searching.\n\n3\n\nI usually use the “compact” style because I’m trying to fit as much in the table as possible.\n\n4\n\nDefine how numbers will show up by default. You can change this for individual columns.\n\n\n\n\n\n\n\n\n\nIt’s a lot of typing for something so simple, but I save it as a code snippet in RStudio and use a shortcut to paste it into a code chunk.\nThere is much, much more that reactable tables can do (like this), but this is a good start toward something readable and presentable to your colleagues. It is also possible to create collapsable rows for large tables, and total rows at the bottom."
+    "title": "23  Report-making in R",
+    "section": "23.1 Interactive tables with reactable",
+    "text": "23.1 Interactive tables with reactable\nThe reactable library is highly customizable, but it can involve a lot of typing to get a good table. The documentation is excellent and there are great examples in the wild. It was used to produce the table above that contained the sparklines.\nThe basic idea is that you select the columns you want to display, then define their format and other details one by one. There are several useful defaults for the whole table, such as the number format. One way to simplify the typing is to just create the names of the columns you want to see in a select statement, then put a default number format for the dollar amounts. A full list of options is available on the help page for reactable.\n\nppp_orig |&gt; \n1  select (\n    Borrower = borrower_name, \n    City = borrower_city, \n    `Zip code` = borrower_zip, \n    `Loan amount` = amount, \n    `Amt forgiven` = forgiveness_amount\n    ) |&gt; \n  reactable (   \n2     sortable = TRUE,\n     filterable = TRUE,\n     searchable = TRUE,\n     defaultPageSize = 5,\n3     compact = TRUE ,\n     striped=TRUE,\n4     defaultColDef =\n        colDef ( \n                format = \n                   colFormat ( currency=\"USD\", separators = TRUE, digits = 0)\n                )\n     )\n\n\n1\n\nusing the select verb that we’ve already seen, pick out the columns for the table and rename them. Don’t forget the back-ticks if you need something with more than one word.\n\n2\n\nSet the options for sorting, filtering, and searching.\n\n3\n\nI usually use the “compact” style because I’m trying to fit as much in the table as possible.\n\n4\n\nDefine how numbers will show up by default. You can change this for individual columns.\n\n\n\n\n\n\n\n\n\nIt’s a lot of typing for something so simple, but I save it as a code snippet in RStudio and use a shortcut to paste it into a code chunk.\nThere is much, much more that reactable tables can do (like this), but this is a good start toward something readable and presentable to your colleagues. It is also possible to create collapsable rows for large tables, and total rows at the bottom."
   },
   {
     "objectID": "vis-tables.html#crosstabs-with-the-janitor-library",
     "href": "vis-tables.html#crosstabs-with-the-janitor-library",
-    "title": "22  Report-making in R",
-    "section": "22.2 Crosstabs with the janitor library",
-    "text": "22.2 Crosstabs with the janitor library\nWe’ve gone through the exercise of creating a grouping, counting, then computing a percentage of total using our usual verbs. The janitor package provides a utility for doing that.\nHere’s an example with the full PPP dataset:\nThe tabyl function does one and only one thing: Count and compute percentages of totals. It creates an output data frame that you can style any way you want. You can do 1-, 2- or 3-way tables, but no more. This example creates the table, then formats it using a datatable as above.\n\nA simple one-way table\n\nppp |&gt; \n1  tabyl ( lender ) |&gt;\n2  arrange ( desc (percent)) |&gt;\n3  head (5) |&gt;\n  datatable (   \n4    rownames = FALSE\n  ) |&gt; \n5  formatPercentage (\"percent\") |&gt;\n6  formatRound (columns =\"n\",  digits=0)\n\n\n1\n\nCreate a one-way data table with count (n) and percent\n\n2\n\nSort it from highest to lowest\n\n3\n\nKeep only the top 5,\n\n4\n\ncreate our default datatable\n\n5\n\nFormat the percentages\n\n6\n\nFormat the raw numbers\n\n\n\n\n\n\n\n\n\n\nTwo-way table with tabyls\nThat’s good, but we have wanted some 2-way tables, the way we did in the group by chapter. Here’s an example:\n\nppp |&gt; \n  tabyl ( loan_status, draw)  \n\n\n\n  \n\n\n\nThis just created the simple counts. There are some “adorn” functions in the package that let you work with the numbers and percentages.\n\nppp |&gt; \n  tabyl (loan_status, draw) |&gt; \n  adorn_totals (c(\"row\", \"col\")) |&gt; \n1  adorn_percentages(\"col\" ) |&gt;\n  adorn_pct_formatting (digits=0) |&gt; \n  adorn_ns() |&gt; \n  adorn_title () \n\n\n1\n\nChange “col” to “row” if you want it to add to 100% across rather than down; “all” if you want it to be a percent of the grand total.\n\n\n\n\n\n\n  \n\n\n\nThis is not a very well documented or developed package, and it balks at giving the “adorn” commands out of order. So sometimes it’s not all that helpful, but it does create good crosstabs quickly."
+    "title": "23  Report-making in R",
+    "section": "23.2 Crosstabs with the janitor library",
+    "text": "23.2 Crosstabs with the janitor library\nWe’ve gone through the exercise of creating a grouping, counting, then computing a percentage of total using our usual verbs. The janitor package provides a utility for doing that.\nHere’s an example with the full PPP dataset:\nThe tabyl function does one and only one thing: Count and compute percentages of totals. It creates an output data frame that you can style any way you want. You can do 1-, 2- or 3-way tables, but no more. This example creates the table, then formats it using a datatable as above.\n\nA simple one-way table\n\nppp |&gt; \n1  tabyl ( lender ) |&gt;\n2  arrange ( desc (percent)) |&gt;\n3  head (5) |&gt;\n  datatable (   \n4    rownames = FALSE\n  ) |&gt; \n5  formatPercentage (\"percent\") |&gt;\n6  formatRound (columns =\"n\",  digits=0)\n\n\n1\n\nCreate a one-way data table with count (n) and percent\n\n2\n\nSort it from highest to lowest\n\n3\n\nKeep only the top 5,\n\n4\n\ncreate our default datatable\n\n5\n\nFormat the percentages\n\n6\n\nFormat the raw numbers\n\n\n\n\n\n\n\n\n\n\nTwo-way table with tabyls\nThat’s good, but we have wanted some 2-way tables, the way we did in the group by chapter. Here’s an example:\n\nppp |&gt; \n  tabyl ( loan_status, draw)  \n\n\n\n  \n\n\n\nThis just created the simple counts. There are some “adorn” functions in the package that let you work with the numbers and percentages.\n\nppp |&gt; \n  tabyl (loan_status, draw) |&gt; \n  adorn_totals (c(\"row\", \"col\")) |&gt; \n1  adorn_percentages(\"col\" ) |&gt;\n  adorn_pct_formatting (digits=0) |&gt; \n  adorn_ns() |&gt; \n  adorn_title () \n\n\n1\n\nChange “col” to “row” if you want it to add to 100% across rather than down; “all” if you want it to be a percent of the grand total.\n\n\n\n\n\n\n  \n\n\n\nThis is not a very well documented or developed package, and it balks at giving the “adorn” commands out of order. So sometimes it’s not all that helpful, but it does create good crosstabs quickly."
   },
   {
     "objectID": "vis-tables.html#static-tables-with-gt",
     "href": "vis-tables.html#static-tables-with-gt",
-    "title": "22  Report-making in R",
-    "section": "22.3 Static tables with gt",
-    "text": "22.3 Static tables with gt\nSection TK."
+    "title": "23  Report-making in R",
+    "section": "23.3 Static tables with gt",
+    "text": "23.3 Static tables with gt\nSection TK."
   },
   {
     "objectID": "vis-demo.html#intro-to-visualization-in-r",
     "href": "vis-demo.html#intro-to-visualization-in-r",
-    "title": "23  Visualization demo",
-    "section": "23.1 Intro to visualization in R",
-    "text": "23.1 Intro to visualization in R\nYou’ll need some additional packages installed for this to work:\n\ngapminder - example data that is used throughout, even though it’s a little boring\nplotly - to make interactive charts.\n\nYou can use this command in the Console to install them at the same time:\n    install.packages(c(\"plotly\", \"gapminder\"))\nThen, load your packages, including the new ones you just installed:\n\nlibrary(gapminder)\nlibrary(tidyverse)\nlibrary(ggplot2)\nlibrary(plotly)\n\noptions(digits=2, scipen=99)"
+    "title": "24  Visualization demo",
+    "section": "24.1 Intro to visualization in R",
+    "text": "24.1 Intro to visualization in R\nYou’ll need some additional packages installed for this to work:\n\ngapminder - example data that is used throughout, even though it’s a little boring\nplotly - to make interactive charts.\n\nYou can use this command in the Console to install them at the same time:\n    install.packages(c(\"plotly\", \"gapminder\"))\nThen, load your packages, including the new ones you just installed:\n\nlibrary(gapminder)\nlibrary(tidyverse)\nlibrary(ggplot2)\nlibrary(plotly)\n\noptions(digits=2, scipen=99)"
   },
   {
     "objectID": "vis-demo.html#the-grammar-of-graphics",
     "href": "vis-demo.html#the-grammar-of-graphics",
-    "title": "23  Visualization demo",
-    "section": "23.2 The Grammar of Graphics",
-    "text": "23.2 The Grammar of Graphics\nThe package we’re going to use for our graphing is called ggplot2, which is part of the tidyverse. Conceptually, it follows the principles in Leland Wilkinson’s “Grammar of Graphics”, written in 1999 as a framework for thinking about visualizations. For our purposes, you just need to understand that any visualization is made up of several fundamental pieces. In ggplot2, they are:\n\naesthetics (called “aes”) : What data are you plotting? It can have more than two dimensions, such as x-axis, y-axis, a column used for color and another used for size. These are things we can see on a graphic, including the values.\ngeometry refers to the shape each column will take.\nscale - any transformation we might make.\nfacets - creating small multiples by splitting up one graph into many similar graphs, based on a variable.\nlayers - adding multiple geometries on top of one another to reveal new information, or add annoations.\n\nIf you’ve used Photoshop or other Adobe products, this idea of layering data, annotation, style and other elements will be familiar to you.\nOne quirk of ggplot is that instead of the %&gt;% pipe command , it uses + instead. There are technical reasons for this, but it’s not worth getting into…\nThe basic structure of the plot is:\n      ggplot (  data = df_name, \n                aes (\n                  x = column_name,\n                  y = column_name,\n                  fill = color inside color,\n                  color = border color,\n                  size = size if relevant\n                  )\n                ) +\n                a geometry (eg, geom_point, geom_bar, etc.) with options such as color, fill and size +\n                facet_wrap ( column ~ column ) +\n                any scale information\nWe’re going to skip the labeling and annotation parts for now. That would come under layering and legends. In practice, these options can be mixed througout the command."
+    "title": "24  Visualization demo",
+    "section": "24.2 The Grammar of Graphics",
+    "text": "24.2 The Grammar of Graphics\nThe package we’re going to use for our graphing is called ggplot2, which is part of the tidyverse. Conceptually, it follows the principles in Leland Wilkinson’s “Grammar of Graphics”, written in 1999 as a framework for thinking about visualizations. For our purposes, you just need to understand that any visualization is made up of several fundamental pieces. In ggplot2, they are:\n\naesthetics (called “aes”) : What data are you plotting? It can have more than two dimensions, such as x-axis, y-axis, a column used for color and another used for size. These are things we can see on a graphic, including the values.\ngeometry refers to the shape each column will take.\nscale - any transformation we might make.\nfacets - creating small multiples by splitting up one graph into many similar graphs, based on a variable.\nlayers - adding multiple geometries on top of one another to reveal new information, or add annoations.\n\nIf you’ve used Photoshop or other Adobe products, this idea of layering data, annotation, style and other elements will be familiar to you.\nOne quirk of ggplot is that instead of the %&gt;% pipe command , it uses + instead. There are technical reasons for this, but it’s not worth getting into…\nThe basic structure of the plot is:\n      ggplot (  data = df_name, \n                aes (\n                  x = column_name,\n                  y = column_name,\n                  fill = color inside color,\n                  color = border color,\n                  size = size if relevant\n                  )\n                ) +\n                a geometry (eg, geom_point, geom_bar, etc.) with options such as color, fill and size +\n                facet_wrap ( column ~ column ) +\n                any scale information\nWe’re going to skip the labeling and annotation parts for now. That would come under layering and legends. In practice, these options can be mixed througout the command."
   },
   {
     "objectID": "vis-demo.html#building-you-graph-in-pieces-using-gapminder",
     "href": "vis-demo.html#building-you-graph-in-pieces-using-gapminder",
-    "title": "23  Visualization demo",
-    "section": "23.3 Building you graph in pieces using gapminder",
-    "text": "23.3 Building you graph in pieces using gapminder\nGapminder is a datast made famous in a viral 2010 TED talk. It contains the life expectency and income (GDP per capita) by country for 200 years. The data is in 5-year increments, through 2007.\n\nSubset to just the latest years\nJust pull out the latest year (2007) for our practice\n\n#what is gapminder? Life expectancy by country, only goes to 2007. Take the latest year.\n\n\ngapminder_2007 &lt;-\n  gapminder %&gt;%\n  filter ( year == 2007)\n\nglimpse(gapminder_2007)\n\nRows: 142\nColumns: 6\n$ country   &lt;fct&gt; \"Afghanistan\", \"Albania\", \"Algeria\", \"Angola\", \"Argentina\", …\n$ continent &lt;fct&gt; Asia, Europe, Africa, Africa, Americas, Oceania, Europe, Asi…\n$ year      &lt;int&gt; 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, …\n$ lifeExp   &lt;dbl&gt; 44, 76, 72, 43, 75, 81, 80, 76, 64, 79, 57, 66, 75, 51, 72, …\n$ pop       &lt;int&gt; 31889923, 3600523, 33333216, 12420476, 40301927, 20434176, 8…\n$ gdpPercap &lt;dbl&gt; 975, 5937, 6223, 4797, 12779, 34435, 36126, 29796, 1391, 336…\n\n\n\n\nThe histogram\nOne of the first pieces of information you often want about a dataset is its distribution. Do all of the values cluster around the center? Or do they spread out? Let’s do a histogram of the life expectency variable.\nStart your plot with the commmand ggplot, then add in the aesthetics and geometry:\n\nggplot( \n        data=gapminder_2007,\n        aes (x = lifeExp )\n      ) +\n      geom_histogram()\n\n\n\n\nWe’ll add a few options to this plot to make it a little easier to read\n\n#make it an outline, with smaller piles.\nggplot ( data= gapminder_2007,\n         aes (x= lifeExp)\n        ) +\n        geom_histogram (binwidth=5, color=\"black\", fill=\"white\")   \n\n\n\n\nBut I’m interested in how the different continents look. Try “faceting” by continent.\n\nggplot ( data= gapminder_2007,\n         aes (x= lifeExp)\n        ) +\n        geom_histogram (binwidth=5, color=\"black\", fill=\"white\")   +\n        facet_wrap ( ~ continent)\n\n\n\n\n\n\nTry a scatter, or dot plot\n\nSave a plot and add to it\nYou can save your plot to an object rather than print it immediately, making it a little easier to troubleshoot.\n\nmy_plot &lt;-\n  ggplot (\n    data = gapminder_2007,\n    aes (x= gdpPercap , y = lifeExp)\n  )\n#what does this look like?\nmy_plot\n\n\n\n\nIt doesn’t look like anything! The reason is that I didn’t include a geometry, or a shape, for the values. Add a point here:\n\nmy_plot &lt;-\n  my_plot +\n  geom_point()\nmy_plot\n\n\n\n\nThere are some built-in themes that take some best practices for mixes of colors and styles, so I’ll add one in.\n\nmy_plot &lt;-\n  my_plot +\n  theme_minimal()\nmy_plot\n\n\n\n\nLet’s add some color. Remember, we keep adding elements to our existing plot, so we don’t have to start over each time.\n\nmy_plot &lt;- \n  my_plot +\n  aes (color = continent)\nmy_plot\n\n\n\n\nAnd now add population for the size of the country\n\nmy_plot &lt;- \n  my_plot +\n  aes (size=pop)\n    \nmy_plot\n\n\n\n\nLet’s build this from scratch, and then also make sure that the big points don’t overlap the little ones too much. Note that I’ve added scale_x_log10 to make the x-axis go up by factors of 10 rather than 1. (In the age of coronavirus, we’ve gotten used to log scales!)\n\nmy_plot &lt;- \n  ggplot (  data= gapminder_2007 , \n            aes ( x= gdpPercap, \n                  y = lifeExp, \n                  color = continent,\n                  size = pop)\n            ) +\n  geom_point (alpha = 0.7) +\n  theme_minimal() +\n  scale_x_log10()\n  \n  \nmy_plot\n\n\n\n\n\n\nFacet\nNow let’s make a little chart for each continent\n\nmy_plot &lt;- \n  my_plot + \n  facet_wrap (~continent)\n\n\nmy_plot\n\n\n\n\n\n\n\nAdding interactivity\nGGPLOT2 is not interactive, so we have to install a different library to allow us to hover over the points. In this case, we’re going to use a function called paste, which puts words together, in the column in aes called “text”:\n\n#install.packages(\"plotly\")\nlibrary(plotly)\n\nmy_plot &lt;-\n  ggplot (data = gapminder_2007,\n          aes(text = paste(\"country: \", country),\n              x= gdpPercap ,\n              y = lifeExp,\n              color= continent,\n              size=pop)\n        ) +\n    geom_point( alpha= 0.7) +\n    theme_minimal() +\n    facet_wrap (~continent) +\n    scale_x_log10()\n#We have to make it a ggplotly to get it interactive.\nmy_plot &lt;- ggplotly(my_plot)\n\nmy_plot"
+    "title": "24  Visualization demo",
+    "section": "24.3 Building you graph in pieces using gapminder",
+    "text": "24.3 Building you graph in pieces using gapminder\nGapminder is a datast made famous in a viral 2010 TED talk. It contains the life expectency and income (GDP per capita) by country for 200 years. The data is in 5-year increments, through 2007.\n\nSubset to just the latest years\nJust pull out the latest year (2007) for our practice\n\n#what is gapminder? Life expectancy by country, only goes to 2007. Take the latest year.\n\n\ngapminder_2007 &lt;-\n  gapminder %&gt;%\n  filter ( year == 2007)\n\nglimpse(gapminder_2007)\n\nRows: 142\nColumns: 6\n$ country   &lt;fct&gt; \"Afghanistan\", \"Albania\", \"Algeria\", \"Angola\", \"Argentina\", …\n$ continent &lt;fct&gt; Asia, Europe, Africa, Africa, Americas, Oceania, Europe, Asi…\n$ year      &lt;int&gt; 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, …\n$ lifeExp   &lt;dbl&gt; 44, 76, 72, 43, 75, 81, 80, 76, 64, 79, 57, 66, 75, 51, 72, …\n$ pop       &lt;int&gt; 31889923, 3600523, 33333216, 12420476, 40301927, 20434176, 8…\n$ gdpPercap &lt;dbl&gt; 975, 5937, 6223, 4797, 12779, 34435, 36126, 29796, 1391, 336…\n\n\n\n\nThe histogram\nOne of the first pieces of information you often want about a dataset is its distribution. Do all of the values cluster around the center? Or do they spread out? Let’s do a histogram of the life expectency variable.\nStart your plot with the commmand ggplot, then add in the aesthetics and geometry:\n\nggplot( \n        data=gapminder_2007,\n        aes (x = lifeExp )\n      ) +\n      geom_histogram()\n\n\n\n\nWe’ll add a few options to this plot to make it a little easier to read\n\n#make it an outline, with smaller piles.\nggplot ( data= gapminder_2007,\n         aes (x= lifeExp)\n        ) +\n        geom_histogram (binwidth=5, color=\"black\", fill=\"white\")   \n\n\n\n\nBut I’m interested in how the different continents look. Try “faceting” by continent.\n\nggplot ( data= gapminder_2007,\n         aes (x= lifeExp)\n        ) +\n        geom_histogram (binwidth=5, color=\"black\", fill=\"white\")   +\n        facet_wrap ( ~ continent)\n\n\n\n\n\n\nTry a scatter, or dot plot\n\nSave a plot and add to it\nYou can save your plot to an object rather than print it immediately, making it a little easier to troubleshoot.\n\nmy_plot &lt;-\n  ggplot (\n    data = gapminder_2007,\n    aes (x= gdpPercap , y = lifeExp)\n  )\n#what does this look like?\nmy_plot\n\n\n\n\nIt doesn’t look like anything! The reason is that I didn’t include a geometry, or a shape, for the values. Add a point here:\n\nmy_plot &lt;-\n  my_plot +\n  geom_point()\nmy_plot\n\n\n\n\nThere are some built-in themes that take some best practices for mixes of colors and styles, so I’ll add one in.\n\nmy_plot &lt;-\n  my_plot +\n  theme_minimal()\nmy_plot\n\n\n\n\nLet’s add some color. Remember, we keep adding elements to our existing plot, so we don’t have to start over each time.\n\nmy_plot &lt;- \n  my_plot +\n  aes (color = continent)\nmy_plot\n\n\n\n\nAnd now add population for the size of the country\n\nmy_plot &lt;- \n  my_plot +\n  aes (size=pop)\n    \nmy_plot\n\n\n\n\nLet’s build this from scratch, and then also make sure that the big points don’t overlap the little ones too much. Note that I’ve added scale_x_log10 to make the x-axis go up by factors of 10 rather than 1. (In the age of coronavirus, we’ve gotten used to log scales!)\n\nmy_plot &lt;- \n  ggplot (  data= gapminder_2007 , \n            aes ( x= gdpPercap, \n                  y = lifeExp, \n                  color = continent,\n                  size = pop)\n            ) +\n  geom_point (alpha = 0.7) +\n  theme_minimal() +\n  scale_x_log10()\n  \n  \nmy_plot\n\n\n\n\n\n\nFacet\nNow let’s make a little chart for each continent\n\nmy_plot &lt;- \n  my_plot + \n  facet_wrap (~continent)\n\n\nmy_plot\n\n\n\n\n\n\n\nAdding interactivity\nGGPLOT2 is not interactive, so we have to install a different library to allow us to hover over the points. In this case, we’re going to use a function called paste, which puts words together, in the column in aes called “text”:\n\n#install.packages(\"plotly\")\nlibrary(plotly)\n\nmy_plot &lt;-\n  ggplot (data = gapminder_2007,\n          aes(text = paste(\"country: \", country),\n              x= gdpPercap ,\n              y = lifeExp,\n              color= continent,\n              size=pop)\n        ) +\n    geom_point( alpha= 0.7) +\n    theme_minimal() +\n    facet_wrap (~continent) +\n    scale_x_log10()\n#We have to make it a ggplotly to get it interactive.\nmy_plot &lt;- ggplotly(my_plot)\n\nmy_plot"
   },
   {
     "objectID": "vis-demo.html#adding-animation",
     "href": "vis-demo.html#adding-animation",
-    "title": "23  Visualization demo",
-    "section": "23.4 Adding animation",
-    "text": "23.4 Adding animation\nSo far, we’ve only looked at the year 2007 – what if we wanted to look at it over time, say 50 or so years? Let’s take the years 1957 to 2007:\n\ndf_gap &lt;- \n  gapminder %&gt;%\n  filter (between (year, 1957, 2007))\n# and let's just check it\n\ndf_gap %&gt;% \n  group_by (year) %&gt;%\n  summarise (n() ) %&gt;%\n  arrange (year)\n\n\n\n  \n\n\n\nLet’s just repeat what we did to get the original interactive plot, with one change: The plotly package lets you specify a “frame” aesthetic, which will create animation by whatever column you specify. Here, we’ve specified the year as that frame:\n\nlibrary(plotly)\n#this option forces it to show real numbers, not the 1e things.\noptions (scipen=99)\nmy_plot &lt;-\n  ggplot (data = df_gap,\n          aes(text = paste(\"country: \", country),\n              x= gdpPercap ,\n              y = lifeExp,\n              color= continent,\n              size=pop, \n              frame = year)\n        ) +\n    geom_point( alpha= 0.7) +\n    theme_minimal() +\n    facet_wrap (~continent) +\n    scale_x_log10()\n#We have to make it a ggplotly to get it interactive.\nmy_plot &lt;- ggplotly(my_plot)\nmy_plot"
+    "title": "24  Visualization demo",
+    "section": "24.4 Adding animation",
+    "text": "24.4 Adding animation\nSo far, we’ve only looked at the year 2007 – what if we wanted to look at it over time, say 50 or so years? Let’s take the years 1957 to 2007:\n\ndf_gap &lt;- \n  gapminder %&gt;%\n  filter (between (year, 1957, 2007))\n# and let's just check it\n\ndf_gap %&gt;% \n  group_by (year) %&gt;%\n  summarise (n() ) %&gt;%\n  arrange (year)\n\n\n\n  \n\n\n\nLet’s just repeat what we did to get the original interactive plot, with one change: The plotly package lets you specify a “frame” aesthetic, which will create animation by whatever column you specify. Here, we’ve specified the year as that frame:\n\nlibrary(plotly)\n#this option forces it to show real numbers, not the 1e things.\noptions (scipen=99)\nmy_plot &lt;-\n  ggplot (data = df_gap,\n          aes(text = paste(\"country: \", country),\n              x= gdpPercap ,\n              y = lifeExp,\n              color= continent,\n              size=pop, \n              frame = year)\n        ) +\n    geom_point( alpha= 0.7) +\n    theme_minimal() +\n    facet_wrap (~continent) +\n    scale_x_log10()\n#We have to make it a ggplotly to get it interactive.\nmy_plot &lt;- ggplotly(my_plot)\nmy_plot"
   },
   {
     "objectID": "vis-practice.html#the-data",
     "href": "vis-practice.html#the-data",
-    "title": "24  Visualization exercise",
-    "section": "24.1 The data",
-    "text": "24.1 The data\nWe saw the Hot 100 data at the very beginning of your exploration of R. We’re going to use a slightly different version of it to create the visualizations because an older version was matched to the Spotify database to extract certain information from each of the songs.\nThis data was used as a TidyTuesday dataset in 2021 (scroll down to get the data dictionaries), so it’s a little older than the one we worked on earlier.\nThis data has two tables:\n\nThe Hot 100 from mid-1958 to mid-2021: https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-09-14/billboard.csv\nFor any songs that are in Spotify, the characteristicso of the song: https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-09-14/audio-features.csv\n\nThe dada dictionary shows the characteristics from the Spotify songs, but it’s a little easier to understand if you take a look at Juan De Dios Santos’ post, asking: “Is my Spotify music boring? An analysis involving music, data, and machine learning”[https://towardsdatascience.com/is-my-spotify-music-boring-an-analysis-involving-music-data-and-machine-learning-47550ae931de]\n\nExamples\nAll of these examples were made with a the version of R and the tidyverse that was used in 2021, so you’ll see a few things in the code that are different, notably the use of %&gt;% instead of |&gt; as the pipe. They are either R scripts (.R extension) or R Markdown (.Rmd extension) , which is the predecessor to Quarto.\n\nTidyTuesday videos\nDavid Robinson, a data scientist at Heap, often creates live demos of his exploration of TidyTuesday datasets – data he’s never seen before. They’re long (usually about an hour), but you can watch him decide what to do and how to do it live. He made this video for the Hot 100 data. (The second half gets into machine learning / AI to make predictions – skip that part if you watch the video).\nHere are a few screen shots from that video – they’re closer to our exploratory visualization than the polished ones below;\n\n\nVisualizing characteristics\nThese are really polished, publication-quality vis examples made in R, but you can take inspiration from these to do draft versions for the purpose of exploration.\nLaurence Horton did a few quick charts to look at the Spotify characteristics.\nLaura Navarroshows us what [annotation can do for a chart]((https://github.com/lau-cloud/TidyTuesdaycode/blob/master/Billboard/script.R) . I’m super surprised at the relatively low showing of hip hop, but …\n\nDanielle Navarro wrote a full tutorial aimed at beginners in R , walking through the data to find and visualize all of Britney Spears’ hits. It has the added feature of explaining how to find her using regular expressions! (Yes, they’re both people named Navarro. I don’t know if they’re the same person.)\nAnd Javier Tamayo-Leiva went a little crazy with colors but found an interesting way to visualize the top hits using two dimensions: Number of weeks at No. 1 and the “energy” of the song. Here is the R code for that chart.\n\n\n\nhot100"
+    "title": "25  Visualization exercise",
+    "section": "25.1 The data",
+    "text": "25.1 The data\nWe saw the Hot 100 data at the very beginning of your exploration of R. We’re going to use a slightly different version of it to create the visualizations because an older version was matched to the Spotify database to extract certain information from each of the songs.\nThis data was used as a TidyTuesday dataset in 2021 (scroll down to get the data dictionaries), so it’s a little older than the one we worked on earlier.\nThis data has two tables:\n\nThe Hot 100 from mid-1958 to mid-2021: https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-09-14/billboard.csv\nFor any songs that are in Spotify, the characteristicso of the song: https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-09-14/audio-features.csv\n\nThe dada dictionary shows the characteristics from the Spotify songs, but it’s a little easier to understand if you take a look at Juan De Dios Santos’ post, asking: “Is my Spotify music boring? An analysis involving music, data, and machine learning”[https://towardsdatascience.com/is-my-spotify-music-boring-an-analysis-involving-music-data-and-machine-learning-47550ae931de]\n\nExamples\nAll of these examples were made with a the version of R and the tidyverse that was used in 2021, so you’ll see a few things in the code that are different, notably the use of %&gt;% instead of |&gt; as the pipe. They are either R scripts (.R extension) or R Markdown (.Rmd extension) , which is the predecessor to Quarto.\n\nTidyTuesday videos\nDavid Robinson, a data scientist at Heap, often creates live demos of his exploration of TidyTuesday datasets – data he’s never seen before. They’re long (usually about an hour), but you can watch him decide what to do and how to do it live. He made this video for the Hot 100 data. (The second half gets into machine learning / AI to make predictions – skip that part if you watch the video).\nHere are a few screen shots from that video – they’re closer to our exploratory visualization than the polished ones below;\n\n\nVisualizing characteristics\nThese are really polished, publication-quality vis examples made in R, but you can take inspiration from these to do draft versions for the purpose of exploration.\nLaurence Horton did a few quick charts to look at the Spotify characteristics.\nLaura Navarroshows us what [annotation can do for a chart]((https://github.com/lau-cloud/TidyTuesdaycode/blob/master/Billboard/script.R) . I’m super surprised at the relatively low showing of hip hop, but …\n\nDanielle Navarro wrote a full tutorial aimed at beginners in R , walking through the data to find and visualize all of Britney Spears’ hits. It has the added feature of explaining how to find her using regular expressions! (Yes, they’re both people named Navarro. I don’t know if they’re the same person.)\nAnd Javier Tamayo-Leiva went a little crazy with colors but found an interesting way to visualize the top hits using two dimensions: Number of weeks at No. 1 and the “energy” of the song. Here is the R code for that chart.\n\n\n\nhot100"
   },
   {
     "objectID": "vis-practice.html#your-turn",
     "href": "vis-practice.html#your-turn",
-    "title": "24  Visualization exercise",
-    "section": "24.2 Your turn",
-    "text": "24.2 Your turn\nYou’ve been through a lot, so let’s review the steps you have to take to make a visualization:\nLoad the data in this case using a read_csv() function. There are two datasets, if you want to use them both.\n\nWrangle the data. You will probably need to filter, group, count or otherwise summarize the data. For example, the “Evolution of Genres” probably starts with a dataset by year and genre.\nDecide on the “geom” that you want to use. We haven’t gone through a lot of these yet, so I’m listing some of the common ones below.\n\nYou might want to change themes and color palettes, which is is a little harder. Get the basic visualization down before you start polishing!\n\nYou might want to use the https://r-graph-gallery.com/ site for inspiration, which always has the packages required and the exact commands you need make it.\n\n\n\n\n\n\nTip\n\n\n\nYou may end up seeing a lot of code in the examples that start with fct_ – these are turning words into categories that are sorted on a particular element, such as the number of times that they show up. It turns words into categories that have numbers underneath the words so that they are always the same and every possible category is defined. These are called factors and we haven’t really dealt with them. But they’re useful. They depend on loading a library of the tidyverse called forcats\n\n\n      ggplot (  data = df_name, \n                aes (\n                  x = column_name,\n                  y = column_name,\n                  fill = column_to_define_the_fill,\n                  color = column_to_define_the_lines_and_borders,\n                  size = column_to_define_the_size\n                  ) \n                ) +\n                geom_something ( aes (`columns that only refer to this geom`) )\n                facet_wrap ( column ~ column ) +\n                extras ()\n  \n\nCommon “geoms”\n\ngeom_point()\ngeom_line()\ngeom_bar()\ngeom_area()\n\nOther tutorials and cheat sheets go into much more detail:\n\nThe “ggplot” chapter of Rafael A. Irizarry’s Introduction to Data Science\nThe Posit cheat sheet for ggplot\nKiernan Healy’s “Data visualization: A practical introduction is a whole book on ggplot."
+    "title": "25  Visualization exercise",
+    "section": "25.2 Your turn",
+    "text": "25.2 Your turn\nYou’ve been through a lot, so let’s review the steps you have to take to make a visualization:\nLoad the data in this case using a read_csv() function. There are two datasets, if you want to use them both.\n\nWrangle the data. You will probably need to filter, group, count or otherwise summarize the data. For example, the “Evolution of Genres” probably starts with a dataset by year and genre.\nDecide on the “geom” that you want to use. We haven’t gone through a lot of these yet, so I’m listing some of the common ones below.\n\nYou might want to change themes and color palettes, which is is a little harder. Get the basic visualization down before you start polishing!\n\nYou might want to use the https://r-graph-gallery.com/ site for inspiration, which always has the packages required and the exact commands you need make it.\n\n\n\n\n\n\nTip\n\n\n\nYou may end up seeing a lot of code in the examples that start with fct_ – these are turning words into categories that are sorted on a particular element, such as the number of times that they show up. It turns words into categories that have numbers underneath the words so that they are always the same and every possible category is defined. These are called factors and we haven’t really dealt with them. But they’re useful. They depend on loading a library of the tidyverse called forcats\n\n\n      ggplot (  data = df_name, \n                aes (\n                  x = column_name,\n                  y = column_name,\n                  fill = column_to_define_the_fill,\n                  color = column_to_define_the_lines_and_borders,\n                  size = column_to_define_the_size\n                  ) \n                ) +\n                geom_something ( aes (`columns that only refer to this geom`) )\n                facet_wrap ( column ~ column ) +\n                extras ()\n  \n\nCommon “geoms”\n\ngeom_point()\ngeom_line()\ngeom_bar()\ngeom_area()\n\nOther tutorials and cheat sheets go into much more detail:\n\nThe “ggplot” chapter of Rafael A. Irizarry’s Introduction to Data Science\nThe Posit cheat sheet for ggplot\nKiernan Healy’s “Data visualization: A practical introduction is a whole book on ggplot."
   },
   {
     "objectID": "vis-practice.html#one-quick-plot",
     "href": "vis-practice.html#one-quick-plot",
-    "title": "24  Visualization exercise",
-    "section": "24.3 One quick plot",
-    "text": "24.3 One quick plot\nThe very first day we looked at R, I showed you one quick plot based on a Hot 100 dataset that I’d created with just the 21 songs that stayed on the Hot 100 for at least a year list during the 2000s. Now that code may make a little more sense to you\n\nhot100 &lt;- readRDS( \n  url (\"https://cronkitedata.s3.amazonaws.com/rdata/top_songs.RDS\")\n  )\n\n\nhot100 |&gt; \n  ggplot ( \n        aes ( x=chart_date, y=hit, color=this_week ) \n        ) +\n       geom_point(  size= .25) +\n  # label the color variable.\n  labs( color = \"Hot 100 #\") +\n  theme_minimal( ) +\n  theme(axis.title.x = element_blank(),\n          axis.title.y = element_blank(), \n        legend.position= \"bottom\") \n\n\n\n\n\nOther sources to learn about ggplot\n\nggplot2: Elegant Graphics for Data Analysis\nData Visualization: A practical Guide\nCedric Sherer’s list of resources, and his materials for the 2-day workshop at the RStudio Conference 2022"
+    "title": "25  Visualization exercise",
+    "section": "25.3 One quick plot",
+    "text": "25.3 One quick plot\nThe very first day we looked at R, I showed you one quick plot based on a Hot 100 dataset that I’d created with just the 21 songs that stayed on the Hot 100 for at least a year list during the 2000s. Now that code may make a little more sense to you\n\nhot100 &lt;- readRDS( \n  url (\"https://cronkitedata.s3.amazonaws.com/rdata/top_songs.RDS\")\n  )\n\n\nhot100 |&gt; \n  ggplot ( \n        aes ( x=chart_date, y=hit, color=this_week ) \n        ) +\n       geom_point(  size= .25) +\n  # label the color variable.\n  labs( color = \"Hot 100 #\") +\n  theme_minimal( ) +\n  theme(axis.title.x = element_blank(),\n          axis.title.y = element_blank(), \n        legend.position= \"bottom\") \n\n\n\n\n\nOther sources to learn about ggplot\n\nggplot2: Elegant Graphics for Data Analysis\nData Visualization: A practical Guide\nCedric Sherer’s list of resources, and his materials for the 2-day workshop at the RStudio Conference 2022"
   },
   {
     "objectID": "vis-maps-begin.html#the-power-of-maps",
     "href": "vis-maps-begin.html#the-power-of-maps",
-    "title": "25  Mapping: An introduction",
-    "section": "25.1 The power of maps",
-    "text": "25.1 The power of maps\nThe most famous use of mapping may be the John Snow map, made in 1854 during a cholera outbreak in London. Officials didn’t know the cause, but Dr. Snow mapped the deaths and quickly saw the common element: The Broad Street water pump.\n\n\n\ncholera map\n\n\nIn data journalism, we use maps to see patterns that would be impossible any other way. Sometimes, they are used to write paragraphs rather than display data.\nIn 2015, we did a story about the tycoons funding the upcoming presidential election. Mapping their locations let us write this paragraph:\n\nNearly all the neighborhoods where they live would fit within the city limits of New Orleans. But minorities make up less than one-fifth of those neighborhoods’ collective population, and virtually no one is Black. Their residents make four and a half times the salary of the average American, and are twice as likely to be college educated.\n\nMapping seems magical and inconceivable until you do it. But it does require a little bit of vocabulary and a few basic concepts before you can accurately and confidently work with maps."
+    "title": "26  Mapping: An introduction",
+    "section": "26.1 The power of maps",
+    "text": "26.1 The power of maps\nThe most famous use of mapping may be the John Snow map, made in 1854 during a cholera outbreak in London. Officials didn’t know the cause, but Dr. Snow mapped the deaths and quickly saw the common element: The Broad Street water pump.\n\n\n\ncholera map\n\n\nIn data journalism, we use maps to see patterns that would be impossible any other way. Sometimes, they are used to write paragraphs rather than display data.\nIn 2015, we did a story about the tycoons funding the upcoming presidential election. Mapping their locations let us write this paragraph:\n\nNearly all the neighborhoods where they live would fit within the city limits of New Orleans. But minorities make up less than one-fifth of those neighborhoods’ collective population, and virtually no one is Black. Their residents make four and a half times the salary of the average American, and are twice as likely to be college educated.\n\nMapping seems magical and inconceivable until you do it. But it does require a little bit of vocabulary and a few basic concepts before you can accurately and confidently work with maps."
   },
   {
     "objectID": "vis-maps-begin.html#mapping-faq",
     "href": "vis-maps-begin.html#mapping-faq",
-    "title": "25  Mapping: An introduction",
-    "section": "25.2 Mapping FAQ",
-    "text": "25.2 Mapping FAQ\n\nWhat is GIS?\nGraphical Information Systems are programs and apps built to work with data that have a spatial, or geographic, element to them. In them, data is meant to be worked with in maps or on a spatial plane or globe rather than in columns and rows. Examples of GIS systems are ESRI’s ArcMap, which is used primarily by researchers and government agencies who create the underlying maps we’ll use in journalism. They are relatively hard on a computer, and require more memory and disk space than we’ve used.\nA popular free and open source GIS system is QGIS, which will work on either Windows or Mac machines. There are excellent tutorials available for QGIS in journalism, including a Knight Center course (but the handout links are dead), and Alexandra Kanik’s prerequisite tutorials for a master class in mapping done for NICAR in 2022.\n\n\nWhat kinds of maps do we use?\n\n“Reference” maps are simply features on a screen or on paper. Think of Google Maps as reference maps. Cartographers take great care in making reference maps readable and useful, from the shape of each line or point to the color of kind of area like a mountain range. Font selection and line types are other elements that cartographers care about, so that it is possible to use the map for navigation or to picture a location.\n“Thematic maps” are the kind that we usually make in journalism: They are designed to illustrate data such as wealth, topography, or crime.\n\nGenerally, your project will begin by finding the right reference map for your work, then layer on top of it thematic elements from your data. Reference maps can come from many places, such as the Census TIGER program, Google Maps ($$$), OpenStreetMap, MapBox, Leaflet or other providers. You don’t have to make them yourself. Turn them into thematic maps by joining data to them or layering data on top.\n\n\nWhat are longitude and latitude?\nThese are the coordinates that define a spot on the globe. In GIS applications, you will usually get these as “decimal degrees” rather than the degree, minute, second form that you learned in junior high.\n\n\n\nSource: mapschool.io\n\n\nUsually, for local data, you will want the numbers to have at least three numbers after the decimal point. However, agencies often mask the actual locations in data they provide by lopping off decimals. Most coding services provide “rooftop” level positions, which have six numbers after the decimal point. For example, longitude (or X) -73.990593 and latitude (Y) 40.740121 define the position of the front of White House.\nIn the United States, longitude (X) is always negative (except Guam), because it’s in the Western Hemisphere. Latitude (Y) is always positive, because there is no place in the United States in the Southern Hemisphere.\n\n\nWhat are features?\nEach layer on your map displays a “feature”, such as roads, schools, or Census tracts. At its simplest, a feature layer is one of the following three types:\n\n\n\nSource: mapschool.io\n\n\n\nPoints: The simplest of the types, a point is a place, defined by its longitude and latitude, or its XY position. You can get that information from the data you’re requesting or obtaining from the government, or you can try to figure it out yourself using a process called “geocoding”. We’ll come back to that.\nLines are strings of points smushed together — they don’t have to be straight, but they do not have the concept of an interior or exterior. You can find points along a line, or distances to a line. While there are many line features in reference maps, I have hardly ever used them in my work. People who work with traffic or water flows, however, use them a lot. You can use them to determine patterns on either side of a border, or a distance to a road.\nPolygons: These lines smushed together to form an enclosed area, such as a state, a Census tract or a Zip Code. They often touch one another, and you can always find out which points fall within each polygon using a “spatial join” – one of the most common tasks that we do. You can turn polygons into points by finding their center point, or centroid.\n\n\n\nRaster data and imagery : For simplification, each of the above types are considered “vector” data – they are defined by positions by connecting a whole bunch of points in space. But raster data, like satellite imagery, are just pixels of different colors. You can technically look at them in any program, but when they are imported to GIS systems, they become “georeferenced”, in that their edges are referenced to position on the globe. For now, we are not going to look at raster data.\n\n\n\n\nSource: mapschool.io\n\n\n“Attributes” are the data elements attached to features – columns in a database! Thematic maps use attributes such as income, population, or temperature to determine the size, color or shape of features on a maps.\n\n\nWhere do I get map data?\nUnlike other parts of the government, sharing map data is very common in agencies. They are so hard to make, and so useful, that most cities and states have GIS departments whose job it is to organize and document the map data.\nMap data comes in various forms – usually a bundle of files called “shapefiles”, which were originally created by ESRI. More modern systems often have “geojson” files, which are easier to use. Google Earth uses “KML” or “KMZ” data, which is also easy to use. JSON files are less efficient than shapefiles, so they are often much larger.\n\nLocal and state governments: In Arizona, the AZGEO hub (accessible with a free user account) coordinates the map data for many agencies of the state. At ASU, the library has a GIS lab with access to much of the base layer information needed, such as county voting districts. They also have contacts in local government who can help you find the underlying map data you need.\nFederal government: At the federal level, the US Geological Service can sometimes serve as a clearinghouse. The Weather Service, USDA and other agencies often have GIS data for free. The National Map also has many layers that can be used.\nThe Census Bureau has mapped every address and block in the US, and creates TIGER files for counties, voting districts, school districts and other political boundaries.\n\nFor very detailed maps, such as the footprints of property parcels in a county , you may have to buy the information. If you get it directly from the county, it shouldn’t be a lot – maybe $100 – but if you buy it privately, it can cost thousands of dollars.\n\nGetting map data from ESRI / ArcMap\nMany government agencies have turned their GIS applications over to the private company, ESRI, and make them public through that system. It’s often quite confusing to download data from ESRI, but it is often possible. If you don’t see a way to use the data yourself, then call the agency responsible for it and ask – you will usually be able to get them to send it to you or give you instructions on where to find it. We’ll look at an example of getting data from an ESRI map in the next chapter.\n\n\n\nWhat if my data doesn’t have longitude and latitude?\nThe process of turning addresses into points on a map is called “geocoding”, and it is one of the most expensive and time-consuming processes in data journalism. The reason is that there are so many ways to represent an address, and that addresses change all the time. It’s not unusual to have 95 percent of your addresses map perfectly, but then you have to look at the remaining 5 percent to see whether there are consistent errors. When I used to do a lot of mapping in Washington, DC, there were three types of addresses that routinely failed – anything on North or South Capitol Street; anything on Martin Luther King, Jr. Avenue, and anything on Nannie Helen Burroughs Avenue. That meant that any map I made, if I didn’t deal with these problems, would be misleading because they would be missing those locations. Geocoding New York City addresses in Queens can also be difficult.\nWe’re not the only ones who struggle with geocoding. In Los Angeles, reporters once found that the crime data supplied to the public was wrong – every time its geocoding failed, it put the crime at 1st and Spring Streets, which was the center of the city map. (Data used inside the police department was correct, because it came from a different source – the GPS data sent by police when they made a stop.)\nThis means that whenever you ask the government for data, you should ALWAYS ask for geographic coordinates and ask if they have GIS files (often using shapefiles). They may not give them to you, but it will save days of work and be much more accurate than trying to put them on a map yourself."
+    "title": "26  Mapping: An introduction",
+    "section": "26.2 Mapping FAQ",
+    "text": "26.2 Mapping FAQ\n\nWhat is GIS?\nGraphical Information Systems are programs and apps built to work with data that have a spatial, or geographic, element to them. In them, data is meant to be worked with in maps or on a spatial plane or globe rather than in columns and rows. Examples of GIS systems are ESRI’s ArcMap, which is used primarily by researchers and government agencies who create the underlying maps we’ll use in journalism. They are relatively hard on a computer, and require more memory and disk space than we’ve used.\nA popular free and open source GIS system is QGIS, which will work on either Windows or Mac machines. There are excellent tutorials available for QGIS in journalism, including a Knight Center course (but the handout links are dead), and Alexandra Kanik’s prerequisite tutorials for a master class in mapping done for NICAR in 2022.\n\n\nWhat kinds of maps do we use?\n\n“Reference” maps are simply features on a screen or on paper. Think of Google Maps as reference maps. Cartographers take great care in making reference maps readable and useful, from the shape of each line or point to the color of kind of area like a mountain range. Font selection and line types are other elements that cartographers care about, so that it is possible to use the map for navigation or to picture a location.\n“Thematic maps” are the kind that we usually make in journalism: They are designed to illustrate data such as wealth, topography, or crime.\n\nGenerally, your project will begin by finding the right reference map for your work, then layer on top of it thematic elements from your data. Reference maps can come from many places, such as the Census TIGER program, Google Maps ($$$), OpenStreetMap, MapBox, Leaflet or other providers. You don’t have to make them yourself. Turn them into thematic maps by joining data to them or layering data on top.\n\n\nWhat are longitude and latitude?\nThese are the coordinates that define a spot on the globe. In GIS applications, you will usually get these as “decimal degrees” rather than the degree, minute, second form that you learned in junior high.\n\n\n\nSource: mapschool.io\n\n\nUsually, for local data, you will want the numbers to have at least three numbers after the decimal point. However, agencies often mask the actual locations in data they provide by lopping off decimals. Most coding services provide “rooftop” level positions, which have six numbers after the decimal point. For example, longitude (or X) -73.990593 and latitude (Y) 40.740121 define the position of the front of White House.\nIn the United States, longitude (X) is always negative (except Guam), because it’s in the Western Hemisphere. Latitude (Y) is always positive, because there is no place in the United States in the Southern Hemisphere.\n\n\nWhat are features?\nEach layer on your map displays a “feature”, such as roads, schools, or Census tracts. At its simplest, a feature layer is one of the following three types:\n\n\n\nSource: mapschool.io\n\n\n\nPoints: The simplest of the types, a point is a place, defined by its longitude and latitude, or its XY position. You can get that information from the data you’re requesting or obtaining from the government, or you can try to figure it out yourself using a process called “geocoding”. We’ll come back to that.\nLines are strings of points smushed together — they don’t have to be straight, but they do not have the concept of an interior or exterior. You can find points along a line, or distances to a line. While there are many line features in reference maps, I have hardly ever used them in my work. People who work with traffic or water flows, however, use them a lot. You can use them to determine patterns on either side of a border, or a distance to a road.\nPolygons: These lines smushed together to form an enclosed area, such as a state, a Census tract or a Zip Code. They often touch one another, and you can always find out which points fall within each polygon using a “spatial join” – one of the most common tasks that we do. You can turn polygons into points by finding their center point, or centroid.\n\n\n\nRaster data and imagery : For simplification, each of the above types are considered “vector” data – they are defined by positions by connecting a whole bunch of points in space. But raster data, like satellite imagery, are just pixels of different colors. You can technically look at them in any program, but when they are imported to GIS systems, they become “georeferenced”, in that their edges are referenced to position on the globe. For now, we are not going to look at raster data.\n\n\n\n\nSource: mapschool.io\n\n\n“Attributes” are the data elements attached to features – columns in a database! Thematic maps use attributes such as income, population, or temperature to determine the size, color or shape of features on a maps.\n\n\nWhere do I get map data?\nUnlike other parts of the government, sharing map data is very common in agencies. They are so hard to make, and so useful, that most cities and states have GIS departments whose job it is to organize and document the map data.\nMap data comes in various forms – usually a bundle of files called “shapefiles”, which were originally created by ESRI. More modern systems often have “geojson” files, which are easier to use. Google Earth uses “KML” or “KMZ” data, which is also easy to use. JSON files are less efficient than shapefiles, so they are often much larger.\n\nLocal and state governments: In Arizona, the AZGEO hub (accessible with a free user account) coordinates the map data for many agencies of the state. At ASU, the library has a GIS lab with access to much of the base layer information needed, such as county voting districts. They also have contacts in local government who can help you find the underlying map data you need.\nFederal government: At the federal level, the US Geological Service can sometimes serve as a clearinghouse. The Weather Service, USDA and other agencies often have GIS data for free. The National Map also has many layers that can be used.\nThe Census Bureau has mapped every address and block in the US, and creates TIGER files for counties, voting districts, school districts and other political boundaries.\n\nFor very detailed maps, such as the footprints of property parcels in a county , you may have to buy the information. If you get it directly from the county, it shouldn’t be a lot – maybe $100 – but if you buy it privately, it can cost thousands of dollars.\n\nGetting map data from ESRI / ArcMap\nMany government agencies have turned their GIS applications over to the private company, ESRI, and make them public through that system. It’s often quite confusing to download data from ESRI, but it is often possible. If you don’t see a way to use the data yourself, then call the agency responsible for it and ask – you will usually be able to get them to send it to you or give you instructions on where to find it. We’ll look at an example of getting data from an ESRI map in the next chapter.\n\n\n\nWhat if my data doesn’t have longitude and latitude?\nThe process of turning addresses into points on a map is called “geocoding”, and it is one of the most expensive and time-consuming processes in data journalism. The reason is that there are so many ways to represent an address, and that addresses change all the time. It’s not unusual to have 95 percent of your addresses map perfectly, but then you have to look at the remaining 5 percent to see whether there are consistent errors. When I used to do a lot of mapping in Washington, DC, there were three types of addresses that routinely failed – anything on North or South Capitol Street; anything on Martin Luther King, Jr. Avenue, and anything on Nannie Helen Burroughs Avenue. That meant that any map I made, if I didn’t deal with these problems, would be misleading because they would be missing those locations. Geocoding New York City addresses in Queens can also be difficult.\nWe’re not the only ones who struggle with geocoding. In Los Angeles, reporters once found that the crime data supplied to the public was wrong – every time its geocoding failed, it put the crime at 1st and Spring Streets, which was the center of the city map. (Data used inside the police department was correct, because it came from a different source – the GPS data sent by police when they made a stop.)\nThis means that whenever you ask the government for data, you should ALWAYS ask for geographic coordinates and ask if they have GIS files (often using shapefiles). They may not give them to you, but it will save days of work and be much more accurate than trying to put them on a map yourself."
   },
   {
     "objectID": "vis-maps-begin.html#a-note-on-projections",
     "href": "vis-maps-begin.html#a-note-on-projections",
-    "title": "25  Mapping: An introduction",
-    "section": "25.3 A note on projections",
-    "text": "25.3 A note on projections\n\nProjections are what we call the mathematical equations that do the trick of turning the world into some flat shape that fits on a printout or a computer screen. It’s a messy task to do, this transformation - there’s no way to smoosh the world onto a screen without distorting it in some way. You either lose direction, or relative size, or come out with something very weird looking.\n-- mapschool.io\n\nThis clip from The West Wing shows how projections can change the way you see the world.\n\nYou can play around with how the standard map distorts size at https://thetruesize.com\nBecause every map has a defined projection, you can’t do anything with one without the maker of the map telling you how it’s been flattened.\n\nGeographic coordinate systems\nGeorgraphic coordinate systems are often used for point features, because they describe posiitons on a globe. They are also often used for reference, or base, maps because they’re easy to use and transform.\nWGS84 with the code ESPG:4326, for World Geodetic System (1984 version) is the standard coordinate system for international data and GPS systems.\nNAD83 or EPSG:4269, is the standard used by the Census Bureau for North American data.\n\n\nProjected coordinate systems\nWhere geographic coordinate systems are measured in degrees, projected coordinate systems use units of measure that we understand – usually feet or meters. They are used more frequently to define direction, distance or shape. Where you would need a lot of geometry to compute the distance between two points on a globe, the distance between two points in a projected system are just the Pythagorean theorem you used in 6th grade: x2 + y2 = distance2\nIf a map is projected, it will usually have the information built into the data file. This is one of the files that makes up the bundle for shapefiles, and the other types have the projection built in. R and other mapping systems will usually recognize it.\nHowever, state and local agencies often leave this part out. In those cases, look for a standard projection used by the state. For Maricopa County, the standard is North American Datum 1983, Arizona State Plane, Central Zone, HARN, in feet (code 2868). For US data, many people use the U.S. national atlas equal area projection, or crs 9311 (especially if you’re only showing the lower 48) or ESRI:102009.\nThere’s a lot of complicated math that goes into flattening the earth into two dimensions for viewing on a screen or a piece of paper. Each of these projections has a number of characteristics, so instead of listing all of them whenever you want to use them, they have codes. You can look up codes using https://spatialreference.org/.\nNow that you have some working knowledge of GIS, we can move onto making maps."
+    "title": "26  Mapping: An introduction",
+    "section": "26.3 A note on projections",
+    "text": "26.3 A note on projections\n\nProjections are what we call the mathematical equations that do the trick of turning the world into some flat shape that fits on a printout or a computer screen. It’s a messy task to do, this transformation - there’s no way to smoosh the world onto a screen without distorting it in some way. You either lose direction, or relative size, or come out with something very weird looking.\n-- mapschool.io\n\nThis clip from The West Wing shows how projections can change the way you see the world.\n\nYou can play around with how the standard map distorts size at https://thetruesize.com\nBecause every map has a defined projection, you can’t do anything with one without the maker of the map telling you how it’s been flattened.\n\nGeographic coordinate systems\nGeorgraphic coordinate systems are often used for point features, because they describe posiitons on a globe. They are also often used for reference, or base, maps because they’re easy to use and transform.\nWGS84 with the code ESPG:4326, for World Geodetic System (1984 version) is the standard coordinate system for international data and GPS systems.\nNAD83 or EPSG:4269, is the standard used by the Census Bureau for North American data.\n\n\nProjected coordinate systems\nWhere geographic coordinate systems are measured in degrees, projected coordinate systems use units of measure that we understand – usually feet or meters. They are used more frequently to define direction, distance or shape. Where you would need a lot of geometry to compute the distance between two points on a globe, the distance between two points in a projected system are just the Pythagorean theorem you used in 6th grade: x2 + y2 = distance2\nIf a map is projected, it will usually have the information built into the data file. This is one of the files that makes up the bundle for shapefiles, and the other types have the projection built in. R and other mapping systems will usually recognize it.\nHowever, state and local agencies often leave this part out. In those cases, look for a standard projection used by the state. For Maricopa County, the standard is North American Datum 1983, Arizona State Plane, Central Zone, HARN, in feet (code 2868). For US data, many people use the U.S. national atlas equal area projection, or crs 9311 (especially if you’re only showing the lower 48) or ESRI:102009.\nThere’s a lot of complicated math that goes into flattening the earth into two dimensions for viewing on a screen or a piece of paper. Each of these projections has a number of characteristics, so instead of listing all of them whenever you want to use them, they have codes. You can look up codes using https://spatialreference.org/.\nNow that you have some working knowledge of GIS, we can move onto making maps."
   },
   {
     "objectID": "vis-maps-firstmap.html#using-tmap-vs.-ggplot-and-leaflet",
     "href": "vis-maps-firstmap.html#using-tmap-vs.-ggplot-and-leaflet",
-    "title": "26  The first map",
-    "section": "26.1 Using tmap vs. ggplot and leaflet",
-    "text": "26.1 Using tmap vs. ggplot and leaflet\nThere are several approaches to getting maps in R. In general, the heavy lifting is done the same way – through the sf (simple features) library. But displaying the maps can be done many different ways.\nThe tmap library does a little more automatically than using the visualization library ggplot to make maps, so that’s the approach this tutorial will take. Neither is better - tmap is a little easier and requires a little less typing, but is not quite as easy to marry with other Javascript interactive maps or to customize using ggplot."
+    "title": "27  The first map",
+    "section": "27.1 Using tmap vs. ggplot and leaflet",
+    "text": "27.1 Using tmap vs. ggplot and leaflet\nThere are several approaches to getting maps in R. In general, the heavy lifting is done the same way – through the sf (simple features) library. But displaying the maps can be done many different ways.\nThe tmap library does a little more automatically than using the visualization library ggplot to make maps, so that’s the approach this tutorial will take. Neither is better - tmap is a little easier and requires a little less typing, but is not quite as easy to marry with other Javascript interactive maps or to customize using ggplot."
   },
   {
     "objectID": "vis-maps-firstmap.html#setup",
     "href": "vis-maps-firstmap.html#setup",
-    "title": "26  The first map",
-    "section": "26.2 Setup",
-    "text": "26.2 Setup\nYou’ll want to change your typical setup chunk to this when you start using maps:\n\n```{r}\n#| label: setup\n\nlibrary(tidyverse)\nlibrary(tigris)\nlibrary(sf)\nlibrary(tmap)\nlibrary(gt)\n\noptions(tigris_class=\"sf\",\n        tigris_use_cache = TRUE)\n```\n\nThe options make it easier to work with the Census bureau’s maps."
+    "title": "27  The first map",
+    "section": "27.2 Setup",
+    "text": "27.2 Setup\nYou’ll want to change your typical setup chunk to this when you start using maps:\n\n```{r}\n#| label: setup\n\nlibrary(tidyverse)\nlibrary(tigris)\nlibrary(sf)\nlibrary(tmap)\nlibrary(gt)\n\noptions(tigris_class=\"sf\",\n        tigris_use_cache = TRUE)\n```\n\nThe options make it easier to work with the Census bureau’s maps."
   },
   {
     "objectID": "vis-maps-firstmap.html#the-first-project",
     "href": "vis-maps-firstmap.html#the-first-project",
-    "title": "26  The first map",
-    "section": "26.3 The first project",
-    "text": "26.3 The first project\nOur first project involves getting points on a map of Maricopa County, then displaying it based on Census demographics. Here, we’ll use voting locations as the points and Census tract level data for income.\n\nPoints\nWe have a list of voting locations that were geocoded by the Arizona Republic last year, courtesy of Caitlin McGlade. The locations aren’t exact – some of the digits were chopped off, and some weren’t found precisely. But they are generally in the right area.\nImport that file using the usual read_csv() method.\n\nlocations &lt;- read_csv(\"https://cronkitedata.s3.amazonaws.com/rdata/maricopa_voting_places.csv\")\n\nglimpse(locations)\n\nRows: 223\nColumns: 7\n$ site_name &lt;chr&gt; \"ASU DOWNTOWN CAMPUS\", \"ASU WEST CAMPUS\", \"AVONDALE CITY HAL…\n$ address   &lt;chr&gt; \"522 N CENTRAL AVE\", \"4701 W THUNDERBIRD RD\", \"11465 W CIVIC…\n$ city      &lt;chr&gt; \"PHOENIX\", \"GLENDALE\", \"AVONDALE\", \"BUCKEYE\", \"QUEEN CREEK\",…\n$ state     &lt;chr&gt; \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", …\n$ zip       &lt;dbl&gt; 85004, 85306, 85323, 85326, 85142, 85335, 85255, 85234, 8525…\n$ longitude &lt;dbl&gt; -112.0740, -112.1599, -112.3031, -112.5839, -111.6361, -112.…\n$ latitude  &lt;dbl&gt; 33.45377, 33.60733, 33.44517, 33.37083, 33.25029, 33.57322, …\n\n\nThe Republic used a geocoding service that provides the data in the geographic reference WGS84, which is code 4326. There are names, addresses, and coordinates in this dataset.\nHere is how to turn the simple data into a geographic dataset:\n\nlocations_map &lt;- \n    locations |&gt; \n    st_as_sf ( \n               coords = c(\"longitude\", \"latitude\"), \n               crs = 4326)\n\nNow take a look at its structure, using the “str” rather than “glimpse” command, which provides a little more detail:\n\nstr(locations_map)\n\nsf [223 × 6] (S3: sf/tbl_df/tbl/data.frame)\n $ site_name: chr [1:223] \"ASU DOWNTOWN CAMPUS\" \"ASU WEST CAMPUS\" \"AVONDALE CITY HALL\" \"BUCKEYE CITY HALL\" ...\n $ address  : chr [1:223] \"522 N CENTRAL AVE\" \"4701 W THUNDERBIRD RD\" \"11465 W CIVIC CENTER DR 200\" \"530 E MONROE AVE\" ...\n $ city     : chr [1:223] \"PHOENIX\" \"GLENDALE\" \"AVONDALE\" \"BUCKEYE\" ...\n $ state    : chr [1:223] \"AZ\" \"AZ\" \"AZ\" \"AZ\" ...\n $ zip      : num [1:223] 85004 85306 85323 85326 85142 ...\n $ geometry :sfc_POINT of length 223; first list element:  'XY' num [1:2] -112.1 33.5\n - attr(*, \"sf_column\")= chr \"geometry\"\n - attr(*, \"agr\")= Factor w/ 3 levels \"constant\",\"aggregate\",..: NA NA NA NA NA\n  ..- attr(*, \"names\")= chr [1:5] \"site_name\" \"address\" \"city\" \"state\" ...\n\n\nThis is hard to read, but take a careful look: Its type of data includes “sf”, along with data frame and table. The longitude and latitude columns are gone, replaced by a column called geometry.\nWe can check to see what projection R thinks it is in using the st_crs() function:\n\nst_crs( locations_map)\n\nCoordinate Reference System:\n  User input: EPSG:4326 \n  wkt:\nGEOGCRS[\"WGS 84\",\n    ENSEMBLE[\"World Geodetic System 1984 ensemble\",\n        MEMBER[\"World Geodetic System 1984 (Transit)\"],\n        MEMBER[\"World Geodetic System 1984 (G730)\"],\n        MEMBER[\"World Geodetic System 1984 (G873)\"],\n        MEMBER[\"World Geodetic System 1984 (G1150)\"],\n        MEMBER[\"World Geodetic System 1984 (G1674)\"],\n        MEMBER[\"World Geodetic System 1984 (G1762)\"],\n        MEMBER[\"World Geodetic System 1984 (G2139)\"],\n        ELLIPSOID[\"WGS 84\",6378137,298.257223563,\n            LENGTHUNIT[\"metre\",1]],\n        ENSEMBLEACCURACY[2.0]],\n    PRIMEM[\"Greenwich\",0,\n        ANGLEUNIT[\"degree\",0.0174532925199433]],\n    CS[ellipsoidal,2],\n        AXIS[\"geodetic latitude (Lat)\",north,\n            ORDER[1],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n        AXIS[\"geodetic longitude (Lon)\",east,\n            ORDER[2],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n    USAGE[\n        SCOPE[\"Horizontal component of 3D system.\"],\n        AREA[\"World.\"],\n        BBOX[-90,-180,90,180]],\n    ID[\"EPSG\",4326]]\n\n\n\n\nUsing the tmap library to display the map\n\ntm_shape ( locations_map) +\n  tm_dots ( )\n\n\n\n\nFor a static map, there are a lot of extra packages that have to be installed and there are some dependencies that we don’t want to deal with. Instead, we’ll just view it as a dynamic map:\n\ntmap_mode(mode=\"view\")\ntm_shape ( locations_map) +\n  tm_dots ( )\n\n\n\n\n\n# turn it back to plot\ntmap_mode(mode=\"plot\")\n\n\n\nAdding census tracts\nThe tigris package gets us access to all of the geographic data from the Census Bureau. Among them are Census tracts.\n\nmaricopa_geo &lt;- \n  tigris::tracts(state = \"04\", county=\"013\", year=\"2020\", \n                 cb=TRUE)\n\nTake a look at the projection that this file is in, since it ought to match when putting two maps together:\n\nst_crs(maricopa_geo)\n\nCoordinate Reference System:\n  User input: NAD83 \n  wkt:\nGEOGCRS[\"NAD83\",\n    DATUM[\"North American Datum 1983\",\n        ELLIPSOID[\"GRS 1980\",6378137,298.257222101,\n            LENGTHUNIT[\"metre\",1]]],\n    PRIMEM[\"Greenwich\",0,\n        ANGLEUNIT[\"degree\",0.0174532925199433]],\n    CS[ellipsoidal,2],\n        AXIS[\"latitude\",north,\n            ORDER[1],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n        AXIS[\"longitude\",east,\n            ORDER[2],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n    ID[\"EPSG\",4269]]\n\n\nThis is a different coordinate system.\nIt’s ok, since we really want them both to be in a system designed for this part of the world.\nWe will present our map in the Arizona State Plane projection, 2868:\n\ntm_shape ( maricopa_geo, projection=2868) +\n  tm_polygons () +\ntm_shape ( locations_map, projection = 2868) +\n  tm_dots ( )\n\n\n\n\n\n\nAdding Census data\nI downloaded some data from the 2018-2021 American Community Survey. You can load it into your environment using this code chunk:\n\nmaricopa_demo &lt;- readRDS(\n                       url (\"https://cronkitedata.s3.amazonaws.com/rdata/maricopa_tract.Rds\")\n)\n\nNow, just like any other data, we can join these two datasets.\n\nmaricopa_demo_map &lt;- \n  maricopa_geo |&gt; \n  left_join ( maricopa_demo, join_by ( GEOID==geoid)) |&gt; \n  select (GEOID, tot_pop:pct_minority )\n  \n\nglimpse (maricopa_demo_map)\n\nRows: 1,009\nColumns: 6\n$ GEOID         &lt;chr&gt; \"04013422102\", \"04013422308\", \"04013103304\", \"0401342070…\n$ tot_pop       &lt;dbl&gt; 4558, 5491, 4164, 5022, 3441, 6215, 4353, 3800, 3990, 58…\n$ nonhisp_white &lt;dbl&gt; 1523, 3237, 1152, 4123, 2540, 3730, 3983, 2371, 3039, 55…\n$ median_inc    &lt;dbl&gt; 34486, 93581, 41853, 110208, 96406, 60598, 141107, 65033…\n$ pct_minority  &lt;dbl&gt; 67, 41, 72, 18, 26, 40, 8, 38, 24, 91, 17, 48, 17, 16, 5…\n$ geometry      &lt;POLYGON [°]&gt; POLYGON ((-111.8746 33.4049..., POLYGON ((-111.8…\n\n\n(You don’t need to select the “geometry” column – it has to be there if it’s a map.)"
+    "title": "27  The first map",
+    "section": "27.3 The first project",
+    "text": "27.3 The first project\nOur first project involves getting points on a map of Maricopa County, then displaying it based on Census demographics. Here, we’ll use voting locations as the points and Census tract level data for income.\n\nPoints\nWe have a list of voting locations that were geocoded by the Arizona Republic last year, courtesy of Caitlin McGlade. The locations aren’t exact – some of the digits were chopped off, and some weren’t found precisely. But they are generally in the right area.\nImport that file using the usual read_csv() method.\n\nlocations &lt;- read_csv(\"https://cronkitedata.s3.amazonaws.com/rdata/maricopa_voting_places.csv\")\n\nglimpse(locations)\n\nRows: 223\nColumns: 7\n$ site_name &lt;chr&gt; \"ASU DOWNTOWN CAMPUS\", \"ASU WEST CAMPUS\", \"AVONDALE CITY HAL…\n$ address   &lt;chr&gt; \"522 N CENTRAL AVE\", \"4701 W THUNDERBIRD RD\", \"11465 W CIVIC…\n$ city      &lt;chr&gt; \"PHOENIX\", \"GLENDALE\", \"AVONDALE\", \"BUCKEYE\", \"QUEEN CREEK\",…\n$ state     &lt;chr&gt; \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", \"AZ\", …\n$ zip       &lt;dbl&gt; 85004, 85306, 85323, 85326, 85142, 85335, 85255, 85234, 8525…\n$ longitude &lt;dbl&gt; -112.0740, -112.1599, -112.3031, -112.5839, -111.6361, -112.…\n$ latitude  &lt;dbl&gt; 33.45377, 33.60733, 33.44517, 33.37083, 33.25029, 33.57322, …\n\n\nThe Republic used a geocoding service that provides the data in the geographic reference WGS84, which is code 4326. There are names, addresses, and coordinates in this dataset.\nHere is how to turn the simple data into a geographic dataset:\n\nlocations_map &lt;- \n    locations |&gt; \n    st_as_sf ( \n               coords = c(\"longitude\", \"latitude\"), \n               crs = 4326)\n\nNow take a look at its structure, using the “str” rather than “glimpse” command, which provides a little more detail:\n\nstr(locations_map)\n\nsf [223 × 6] (S3: sf/tbl_df/tbl/data.frame)\n $ site_name: chr [1:223] \"ASU DOWNTOWN CAMPUS\" \"ASU WEST CAMPUS\" \"AVONDALE CITY HALL\" \"BUCKEYE CITY HALL\" ...\n $ address  : chr [1:223] \"522 N CENTRAL AVE\" \"4701 W THUNDERBIRD RD\" \"11465 W CIVIC CENTER DR 200\" \"530 E MONROE AVE\" ...\n $ city     : chr [1:223] \"PHOENIX\" \"GLENDALE\" \"AVONDALE\" \"BUCKEYE\" ...\n $ state    : chr [1:223] \"AZ\" \"AZ\" \"AZ\" \"AZ\" ...\n $ zip      : num [1:223] 85004 85306 85323 85326 85142 ...\n $ geometry :sfc_POINT of length 223; first list element:  'XY' num [1:2] -112.1 33.5\n - attr(*, \"sf_column\")= chr \"geometry\"\n - attr(*, \"agr\")= Factor w/ 3 levels \"constant\",\"aggregate\",..: NA NA NA NA NA\n  ..- attr(*, \"names\")= chr [1:5] \"site_name\" \"address\" \"city\" \"state\" ...\n\n\nThis is hard to read, but take a careful look: Its type of data includes “sf”, along with data frame and table. The longitude and latitude columns are gone, replaced by a column called geometry.\nWe can check to see what projection R thinks it is in using the st_crs() function:\n\nst_crs( locations_map)\n\nCoordinate Reference System:\n  User input: EPSG:4326 \n  wkt:\nGEOGCRS[\"WGS 84\",\n    ENSEMBLE[\"World Geodetic System 1984 ensemble\",\n        MEMBER[\"World Geodetic System 1984 (Transit)\"],\n        MEMBER[\"World Geodetic System 1984 (G730)\"],\n        MEMBER[\"World Geodetic System 1984 (G873)\"],\n        MEMBER[\"World Geodetic System 1984 (G1150)\"],\n        MEMBER[\"World Geodetic System 1984 (G1674)\"],\n        MEMBER[\"World Geodetic System 1984 (G1762)\"],\n        MEMBER[\"World Geodetic System 1984 (G2139)\"],\n        ELLIPSOID[\"WGS 84\",6378137,298.257223563,\n            LENGTHUNIT[\"metre\",1]],\n        ENSEMBLEACCURACY[2.0]],\n    PRIMEM[\"Greenwich\",0,\n        ANGLEUNIT[\"degree\",0.0174532925199433]],\n    CS[ellipsoidal,2],\n        AXIS[\"geodetic latitude (Lat)\",north,\n            ORDER[1],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n        AXIS[\"geodetic longitude (Lon)\",east,\n            ORDER[2],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n    USAGE[\n        SCOPE[\"Horizontal component of 3D system.\"],\n        AREA[\"World.\"],\n        BBOX[-90,-180,90,180]],\n    ID[\"EPSG\",4326]]\n\n\n\n\nUsing the tmap library to display the map\n\ntm_shape ( locations_map) +\n  tm_dots ( )\n\n\n\n\nFor a static map, there are a lot of extra packages that have to be installed and there are some dependencies that we don’t want to deal with. Instead, we’ll just view it as a dynamic map:\n\ntmap_mode(mode=\"view\")\ntm_shape ( locations_map) +\n  tm_dots ( )\n\n\n\n\n\n# turn it back to plot\ntmap_mode(mode=\"plot\")\n\n\n\nAdding census tracts\nThe tigris package gets us access to all of the geographic data from the Census Bureau. Among them are Census tracts.\n\nmaricopa_geo &lt;- \n  tigris::tracts(state = \"04\", county=\"013\", year=\"2020\", \n                 cb=TRUE)\n\nTake a look at the projection that this file is in, since it ought to match when putting two maps together:\n\nst_crs(maricopa_geo)\n\nCoordinate Reference System:\n  User input: NAD83 \n  wkt:\nGEOGCRS[\"NAD83\",\n    DATUM[\"North American Datum 1983\",\n        ELLIPSOID[\"GRS 1980\",6378137,298.257222101,\n            LENGTHUNIT[\"metre\",1]]],\n    PRIMEM[\"Greenwich\",0,\n        ANGLEUNIT[\"degree\",0.0174532925199433]],\n    CS[ellipsoidal,2],\n        AXIS[\"latitude\",north,\n            ORDER[1],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n        AXIS[\"longitude\",east,\n            ORDER[2],\n            ANGLEUNIT[\"degree\",0.0174532925199433]],\n    ID[\"EPSG\",4269]]\n\n\nThis is a different coordinate system.\nIt’s ok, since we really want them both to be in a system designed for this part of the world.\nWe will present our map in the Arizona State Plane projection, 2868:\n\ntm_shape ( maricopa_geo, projection=2868) +\n  tm_polygons () +\ntm_shape ( locations_map, projection = 2868) +\n  tm_dots ( )\n\n\n\n\n\n\nAdding Census data\nI downloaded some data from the 2018-2021 American Community Survey. You can load it into your environment using this code chunk:\n\nmaricopa_demo &lt;- readRDS(\n                       url (\"https://cronkitedata.s3.amazonaws.com/rdata/maricopa_tract.Rds\")\n)\n\nNow, just like any other data, we can join these two datasets.\n\nmaricopa_demo_map &lt;- \n  maricopa_geo |&gt; \n  left_join ( maricopa_demo, join_by ( GEOID==geoid)) |&gt; \n  select (GEOID, tot_pop:pct_minority )\n  \n\nglimpse (maricopa_demo_map)\n\nRows: 1,009\nColumns: 6\n$ GEOID         &lt;chr&gt; \"04013422102\", \"04013422308\", \"04013103304\", \"0401342070…\n$ tot_pop       &lt;dbl&gt; 4558, 5491, 4164, 5022, 3441, 6215, 4353, 3800, 3990, 58…\n$ nonhisp_white &lt;dbl&gt; 1523, 3237, 1152, 4123, 2540, 3730, 3983, 2371, 3039, 55…\n$ median_inc    &lt;dbl&gt; 34486, 93581, 41853, 110208, 96406, 60598, 141107, 65033…\n$ pct_minority  &lt;dbl&gt; 67, 41, 72, 18, 26, 40, 8, 38, 24, 91, 17, 48, 17, 16, 5…\n$ geometry      &lt;POLYGON [°]&gt; POLYGON ((-111.8746 33.4049..., POLYGON ((-111.8…\n\n\n(You don’t need to select the “geometry” column – it has to be there if it’s a map.)"
   },
   {
     "objectID": "vis-maps-firstmap.html#a-chloropleth-map",
     "href": "vis-maps-firstmap.html#a-chloropleth-map",
-    "title": "26  The first map",
-    "section": "26.4 A chloropleth map",
-    "text": "26.4 A chloropleth map\nNow we can use colors to show the percent of each tract that identified as a racial or ethnic minority in the American Community Surveys.\n\ntm_shape ( maricopa_demo_map, \n           projection = 2868) +\n  tm_polygons ( col = \"pct_minority\", \n                style = \"jenks\", \n                palette = \"YlGnBu\", \n                border.col = NULL)\n\n\n\n\nThere are a lot of options for drawing the maps, but these are the most common. Using “jenks” natural breaks to find the groupings for colors. It tries to find a place where big changes in color won’t happen for minor differences, but it sometimes needs to be tweaked.\n\nColor palettes\nYou may notice I changed the color palette so that the states with a lot of water are in blue and those with little are in yellow! Here are the common palettes used in R, which come from the Color Brewer: color advice for cartography:\n\nRColorBrewer::display.brewer.all()"
+    "title": "27  The first map",
+    "section": "27.4 A chloropleth map",
+    "text": "27.4 A chloropleth map\nNow we can use colors to show the percent of each tract that identified as a racial or ethnic minority in the American Community Surveys.\n\ntm_shape ( maricopa_demo_map, \n           projection = 2868) +\n  tm_polygons ( col = \"pct_minority\", \n                style = \"jenks\", \n                palette = \"YlGnBu\", \n                border.col = NULL)\n\n\n\n\nThere are a lot of options for drawing the maps, but these are the most common. Using “jenks” natural breaks to find the groupings for colors. It tries to find a place where big changes in color won’t happen for minor differences, but it sometimes needs to be tweaked.\n\nColor palettes\nYou may notice I changed the color palette so that the states with a lot of water are in blue and those with little are in yellow! Here are the common palettes used in R, which come from the Color Brewer: color advice for cartography:\n\nRColorBrewer::display.brewer.all()"
   },
   {
     "objectID": "vis-maps-firstmap.html#adding-the-voting-places-and-making-it-interactive",
     "href": "vis-maps-firstmap.html#adding-the-voting-places-and-making-it-interactive",
-    "title": "26  The first map",
-    "section": "26.5 Adding the voting places and making it interactive",
-    "text": "26.5 Adding the voting places and making it interactive\n\ntmap_mode(\"view\")\n\ntm_shape ( maricopa_demo_map, \n           projection = 2868) +\n  tm_polygons ( col = \"pct_minority\", \n                style = \"jenks\", \n                palette = \"YlGnBu\", \n                border.col = NULL, \n                alpha = .3) +\ntm_shape ( locations_map) +\n  tm_dots (col = \"red\")\n\n\n\n\n\n\nYou have now made an interactive map of the Census Tracts and voter locations in Maricopa County, from start to finish! It’s not good enough to publish, but it’s good enough for you to explore on your own.\nThis method of making maps doesn’t have very many options, but it’s the simplest way to get started.\nYou can also use the ggplot library, with additional options added with other packages, to create a publishable static or interactive map."
+    "title": "27  The first map",
+    "section": "27.5 Adding the voting places and making it interactive",
+    "text": "27.5 Adding the voting places and making it interactive\n\ntmap_mode(\"view\")\n\ntm_shape ( maricopa_demo_map, \n           projection = 2868) +\n  tm_polygons ( col = \"pct_minority\", \n                style = \"jenks\", \n                palette = \"YlGnBu\", \n                border.col = NULL, \n                alpha = .3) +\ntm_shape ( locations_map) +\n  tm_dots (col = \"red\")\n\n\n\n\n\n\nYou have now made an interactive map of the Census Tracts and voter locations in Maricopa County, from start to finish! It’s not good enough to publish, but it’s good enough for you to explore on your own.\nThis method of making maps doesn’t have very many options, but it’s the simplest way to get started.\nYou can also use the ggplot library, with additional options added with other packages, to create a publishable static or interactive map."
   },
   {
     "objectID": "advanced.html",
@@ -1024,161 +1080,161 @@
   {
     "objectID": "advanced-pdf.html#the-pdf-problem",
     "href": "advanced-pdf.html#the-pdf-problem",
-    "title": "27  Working with PDF files",
-    "section": "27.1 The PDF problem",
-    "text": "27.1 The PDF problem\nMany government agencies distribute data in PDF, or Adobe Portable Document Format. As the full name implies, these are documents, not data. It is geared at getting text and images properly placed on a page, not creating tidy data as we have been dealing with in previous chapters.\nSome software – Excel, Google Sheets, Microsoft Access and others – create PDF’s as one of their output formats. In fact, there is an entire ecosystem of pdf report makers, such as Crystal Reports. But converting them back into a usable dataset is often a challenge because we want columns and rows, not positions of words on a page.\nSome people call the process of returning a PDF to a data format as “scraping”. Others call it “report mining”. People with a lot of money can use automation tools, notable Altria’s Monarch product, to do this at scale with few errors, but it costs at least $2,000 / year / person for software like this, so few reporters have access.\n\nUnderstanding your pdf\nPDF’s come in two very different types. The first is the more common version: reports generated from other software, such as Quarto, that contain text similar to a formatted Google or Word doc. The other is an image file, which is what is produced by scanning old documents, or sometimes from public records requests in which the agency has printed out a document, redacted it, and scanned it back into an electronic form for distribution.\nIf there are a lot of hand-written annotations on the page, then it is most likely an image. Another test is to try selecting the text on the screen. If you can’t do it (your cursor turns into an image selection tool, or it just selects the whole page), you are really only dealing with pixels and, to the computer, it is nothing different than a picture of your dog. If the selection covers the text and skips around, then it has underlying text and it can be used with most pdf parsing programs.1\n\n\n\n\n\n\n\n(a) selecting text in a pdf\n\n\n\n\n\n\n\n(b) selecting image in a pdf\n\n\n\n\nFigure 27.1: PDF types\n\n\nWhen you have an image, you will need a first step, or more expensive software, to deal with it. Most reporters use Google Pinpoint or DocumentCloud to do the “optical character recognition”, or OCR, necessary to work with them. A full version of Adobe Acrobat (not the reader) will also work. Some of the more expensive products listed below have versions that include OCR. But just know that if you have an image, your ability to parse these could be very difficult.\n\n\nDeveloping a strategy\nEvery PDF project is different. There are a few strategies you can employ to make it a little more replicable, but it’s always a difficult problem that doesn’t translate well from project to project.\nIf you are going to spend money on a tool for anything in data journalism, this is probably the place to spend it.\nHere are some things to think about:\n\nTry every simple tool and see if anything works. Each of the tools that turn PDFs into data work in a slightly different way, and each one makes different kinds of mistakes. It’s very easy to run a test set of documents through each product and see if any works almost perfectly.\nConsider re-typing! If you have a pdf created from a spreadsheet, you will likely spend less time retyping 100 or so rows if the simple tools don’t work.\nYou may have to use a combination of the tool and a program to get the data you want. If that is the case, try to find a conversion that preserves the position on the page, which is the one thing that’s accurate in the PDF. In that case, you write programs or spreadsheet formulas to look for a pattern like a date in certain columns that can then be used to identify different kinds of lines.\nConsider whether you really need the PDF in data form at all. If, for example, your goal is to see if the same names show up in different documents, Pinpoint or DocumentCloud might do just as well using a different process called entity extraction.\n\n\n\nExample document\nThe document we’ll use for testing is the most recent lobbying report from the city of Phoenix. This is a bare-bones report of 14 pages, which has repeated sections of the names of individual lobbyists underneath the name of the lobbying company and the date filed:\n\n\n\nWork from the goal backwards\nThe goal would be to have one dataset with the following columns:\n\nDate of filing\nName of lobbying organization\nLobbyist name 2\n\nThe filing and organization data would be repeated for each of the lobbyists.\nThis seems so simple! In your head, you can see exactly how that would work. But now you have to spend some time thinking about how you’ll get there.\nOne trick in reversing PDFs is to look for anything with a pattern that distinguishes one KIND of row from another – a heading, an indentation, etc. In this case, it’s the line with the date at the end that will make the biggest difference.\nIn this case, we have to find a way to apply the name of the organization and its most recently updated date to all of the rows of the lobbyist names. The pattern is that there is no date on the lobbyist list, but there is always a date on the organization row. (Sometimes indentations and formatting can help, but not always.)"
+    "title": "28  Working with PDF files",
+    "section": "28.1 The PDF problem",
+    "text": "28.1 The PDF problem\nMany government agencies distribute data in PDF, or Adobe Portable Document Format. As the full name implies, these are documents, not data. It is geared at getting text and images properly placed on a page, not creating tidy data as we have been dealing with in previous chapters.\nSome software – Excel, Google Sheets, Microsoft Access and others – create PDF’s as one of their output formats. In fact, there is an entire ecosystem of pdf report makers, such as Crystal Reports. But converting them back into a usable dataset is often a challenge because we want columns and rows, not positions of words on a page.\nSome people call the process of returning a PDF to a data format as “scraping”. Others call it “report mining”. People with a lot of money can use automation tools, notable Altria’s Monarch product, to do this at scale with few errors, but it costs at least $2,000 / year / person for software like this, so few reporters have access.\n\nUnderstanding your pdf\nPDF’s come in two very different types. The first is the more common version: reports generated from other software, such as Quarto, that contain text similar to a formatted Google or Word doc. The other is an image file, which is what is produced by scanning old documents, or sometimes from public records requests in which the agency has printed out a document, redacted it, and scanned it back into an electronic form for distribution.\nIf there are a lot of hand-written annotations on the page, then it is most likely an image. Another test is to try selecting the text on the screen. If you can’t do it (your cursor turns into an image selection tool, or it just selects the whole page), you are really only dealing with pixels and, to the computer, it is nothing different than a picture of your dog. If the selection covers the text and skips around, then it has underlying text and it can be used with most pdf parsing programs.1\n\n\n\n\n\n\n\n(a) selecting text in a pdf\n\n\n\n\n\n\n\n(b) selecting image in a pdf\n\n\n\n\nFigure 28.1: PDF types\n\n\nWhen you have an image, you will need a first step, or more expensive software, to deal with it. Most reporters use Google Pinpoint or DocumentCloud to do the “optical character recognition”, or OCR, necessary to work with them. A full version of Adobe Acrobat (not the reader) will also work. Some of the more expensive products listed below have versions that include OCR. But just know that if you have an image, your ability to parse these could be very difficult.\n\n\nDeveloping a strategy\nEvery PDF project is different. There are a few strategies you can employ to make it a little more replicable, but it’s always a difficult problem that doesn’t translate well from project to project.\nIf you are going to spend money on a tool for anything in data journalism, this is probably the place to spend it.\nHere are some things to think about:\n\nTry every simple tool and see if anything works. Each of the tools that turn PDFs into data work in a slightly different way, and each one makes different kinds of mistakes. It’s very easy to run a test set of documents through each product and see if any works almost perfectly.\nConsider re-typing! If you have a pdf created from a spreadsheet, you will likely spend less time retyping 100 or so rows if the simple tools don’t work.\nYou may have to use a combination of the tool and a program to get the data you want. If that is the case, try to find a conversion that preserves the position on the page, which is the one thing that’s accurate in the PDF. In that case, you write programs or spreadsheet formulas to look for a pattern like a date in certain columns that can then be used to identify different kinds of lines.\nConsider whether you really need the PDF in data form at all. If, for example, your goal is to see if the same names show up in different documents, Pinpoint or DocumentCloud might do just as well using a different process called entity extraction.\n\n\n\nExample document\nThe document we’ll use for testing is the most recent lobbying report from the city of Phoenix. This is a bare-bones report of 14 pages, which has repeated sections of the names of individual lobbyists underneath the name of the lobbying company and the date filed:\n\n\n\nWork from the goal backwards\nThe goal would be to have one dataset with the following columns:\n\nDate of filing\nName of lobbying organization\nLobbyist name 2\n\nThe filing and organization data would be repeated for each of the lobbyists.\nThis seems so simple! In your head, you can see exactly how that would work. But now you have to spend some time thinking about how you’ll get there.\nOne trick in reversing PDFs is to look for anything with a pattern that distinguishes one KIND of row from another – a heading, an indentation, etc. In this case, it’s the line with the date at the end that will make the biggest difference.\nIn this case, we have to find a way to apply the name of the organization and its most recently updated date to all of the rows of the lobbyist names. The pattern is that there is no date on the lobbyist list, but there is always a date on the organization row. (Sometimes indentations and formatting can help, but not always.)"
   },
   {
     "objectID": "advanced-pdf.html#pdf-conversion-tools",
     "href": "advanced-pdf.html#pdf-conversion-tools",
-    "title": "27  Working with PDF files",
-    "section": "27.2 PDF conversion tools",
-    "text": "27.2 PDF conversion tools\nYou can’t just open a pdf or import it into an R program or Excel. Instead, you’ll need a tool to make the conversions. Here are some of the common ones used in newsrooms.\n\nTabula\nTabula.technology is a downloadable program that, once unzipped into your applications or document folder, runs in a browser. The most recent version is more than 5 years old, though, so it isn’t being actively updated or refined. Because it’s a Java-based program, it may not be possible to easily use it on all computers. I’m not sure it’s worth fiddling with if it doesn’t work on yours.\nHere’s a quick video of how Tabula works on this document:\n(I cut at least 5 minutes of waiting out of this video!)\n\n\n\nCometdocs\nUsually, Cometdocs is one of the most reliable conversions. But even after setting up an account, this is what I got:\nAt times, Investigative Reporters and Editors has arranged for free professional accounts. Check if they still have the program if you’re a member.\n\n\n\nzamzar\nZamzar did quite well with this, producing a file that can be used because it kept the concept of two columns and didn’t smush rows together.\n\n\n\nAdobe Acrobat ($)\nEvery ASU student has an automatic account for the full version of Adobe Acrobat. The online version has limited capabilities, but it also allows you to download the full version. But in this instance, it failed miserably:\n\n\n\n\n\n\n\ninput\n\n\n\n\n\n\n\noutput\n\n\n\n\nFigure 27.2: Using Acrobat\n\n\nIt tried to replicate the way the data looks in the page, which smushed together the individual lobbyists even though they’re clearly on separate lines.\n\n\nAble2Extract ($)\nAble2Extract from Invetintech.com is an all-purpose PDF extractor and producer, with the professional version costing $200. It’s got a lot of options for tricky tables, and it also does OCR if necessary before trying to extract any information. (Only Acrobat on the above list does that as well.) It’s the one I use most frequently. In this case, it did nothing different than the free versions, but it also permits you to keep the position on the page instead of converting to an Excel or CSV file, making it easier if you want to manually parse the lines.\n\n\nGoogle Pinpoint’s Table Extractor\nGoogle Pinpoint is a service available only to journalists. You have to apply for an account and, for the time being, apply for the beta add-on.\nHaving Google Pinpoint is a great option if all you want to know is whether the same entity shows up across documents. It also provides free audio and video transcription, and free OCR of image documents. It’s similar to DocumentCloud, but is geared less at publishing documents and more at managing the collection of material vacuumed up in an investigation.\nFor now, it doesn’t work very well on our sample document. It’s just not the right form for it. It works best on collections of individual forms with headings , where one PDF is one form. An example of a good document for Pinpoint is a collection of “Suspected Illegal Alien Form” documents from AZDPS, where each box has a label and it’s not a table:"
+    "title": "28  Working with PDF files",
+    "section": "28.2 PDF conversion tools",
+    "text": "28.2 PDF conversion tools\nYou can’t just open a pdf or import it into an R program or Excel. Instead, you’ll need a tool to make the conversions. Here are some of the common ones used in newsrooms.\n\nTabula\nTabula.technology is a downloadable program that, once unzipped into your applications or document folder, runs in a browser. The most recent version is more than 5 years old, though, so it isn’t being actively updated or refined. Because it’s a Java-based program, it may not be possible to easily use it on all computers. I’m not sure it’s worth fiddling with if it doesn’t work on yours.\nHere’s a quick video of how Tabula works on this document:\n(I cut at least 5 minutes of waiting out of this video!)\n\n\n\nCometdocs\nUsually, Cometdocs is one of the most reliable conversions. But even after setting up an account, this is what I got:\nAt times, Investigative Reporters and Editors has arranged for free professional accounts. Check if they still have the program if you’re a member.\n\n\n\nzamzar\nZamzar did quite well with this, producing a file that can be used because it kept the concept of two columns and didn’t smush rows together.\n\n\n\nAdobe Acrobat ($)\nEvery ASU student has an automatic account for the full version of Adobe Acrobat. The online version has limited capabilities, but it also allows you to download the full version. But in this instance, it failed miserably:\n\n\n\n\n\n\n\ninput\n\n\n\n\n\n\n\noutput\n\n\n\n\nFigure 28.2: Using Acrobat\n\n\nIt tried to replicate the way the data looks in the page, which smushed together the individual lobbyists even though they’re clearly on separate lines.\n\n\nAble2Extract ($)\nAble2Extract from Invetintech.com is an all-purpose PDF extractor and producer, with the professional version costing $200. It’s got a lot of options for tricky tables, and it also does OCR if necessary before trying to extract any information. (Only Acrobat on the above list does that as well.) It’s the one I use most frequently. In this case, it did nothing different than the free versions, but it also permits you to keep the position on the page instead of converting to an Excel or CSV file, making it easier if you want to manually parse the lines.\n\n\nGoogle Pinpoint’s Table Extractor\nGoogle Pinpoint is a service available only to journalists. You have to apply for an account and, for the time being, apply for the beta add-on.\nHaving Google Pinpoint is a great option if all you want to know is whether the same entity shows up across documents. It also provides free audio and video transcription, and free OCR of image documents. It’s similar to DocumentCloud, but is geared less at publishing documents and more at managing the collection of material vacuumed up in an investigation.\nFor now, it doesn’t work very well on our sample document. It’s just not the right form for it. It works best on collections of individual forms with headings , where one PDF is one form. An example of a good document for Pinpoint is a collection of “Suspected Illegal Alien Form” documents from AZDPS, where each box has a label and it’s not a table:"
   },
   {
     "objectID": "advanced-pdf.html#finishing-up",
     "href": "advanced-pdf.html#finishing-up",
-    "title": "27  Working with PDF files",
-    "section": "27.3 Finishing up",
-    "text": "27.3 Finishing up\nUsing the csv file produced by Able2Extract, here’s one method for finishing up, using formulas in Google Sheets to figure out how to put together the pages. Here are the steps I took:\n\nRemove all empty rows, so that copying an pasting is easier.\nAd an ID column of sequential numbers before the first column.\nUse a formula to determine if the cell with the “date” is empty: =if (isblank (c2, \"No\", \"Yes\"))\nFor both the company name and the date: if the flag = “Yes”, then use the value on this this row in the appropriate column. If it’s “No”, use the value directly above the current cell. For example, =if( d2 = \"Yes\", b2, e1)\nConvert the formulas to values by copying, then pasting values only through Paste Special after selecting A1.\nRemove some of the extra rows from page headers\nRemove the rows that are company names , since we now have them next to the individuals, and remove the columns I worked off of.\n\n\nThis Google Sheet shows how some of the data was transferred from the converters, and the final spreadsheet that is in “tidy” format."
+    "title": "28  Working with PDF files",
+    "section": "28.3 Finishing up",
+    "text": "28.3 Finishing up\nUsing the csv file produced by Able2Extract, here’s one method for finishing up, using formulas in Google Sheets to figure out how to put together the pages. Here are the steps I took:\n\nRemove all empty rows, so that copying an pasting is easier.\nAd an ID column of sequential numbers before the first column.\nUse a formula to determine if the cell with the “date” is empty: =if (isblank (c2, \"No\", \"Yes\"))\nFor both the company name and the date: if the flag = “Yes”, then use the value on this this row in the appropriate column. If it’s “No”, use the value directly above the current cell. For example, =if( d2 = \"Yes\", b2, e1)\nConvert the formulas to values by copying, then pasting values only through Paste Special after selecting A1.\nRemove some of the extra rows from page headers\nRemove the rows that are company names , since we now have them next to the individuals, and remove the columns I worked off of.\n\n\nThis Google Sheet shows how some of the data was transferred from the converters, and the final spreadsheet that is in “tidy” format."
   },
   {
     "objectID": "advanced-pdf.html#use-a-program-instead",
     "href": "advanced-pdf.html#use-a-program-instead",
-    "title": "27  Working with PDF files",
-    "section": "27.4 Use a program instead",
-    "text": "27.4 Use a program instead\nWe won’t go through all of the methods of using a program to deconstruct PDFs, but you’ll sometimes find that it’s easier to program it than to go through all of the manual steps. This Quarto document shows how you might process our example . There are a number of functions you’ve never seen in here before, since we’re working with vectors and lists to begin with rather than data frames."
+    "title": "28  Working with PDF files",
+    "section": "28.4 Use a program instead",
+    "text": "28.4 Use a program instead\nWe won’t go through all of the methods of using a program to deconstruct PDFs, but you’ll sometimes find that it’s easier to program it than to go through all of the manual steps. This Quarto document shows how you might process our example . There are a number of functions you’ve never seen in here before, since we’re working with vectors and lists to begin with rather than data frames."
   },
   {
     "objectID": "advanced-pdf.html#footnotes",
     "href": "advanced-pdf.html#footnotes",
-    "title": "27  Working with PDF files",
+    "title": "28  Working with PDF files",
     "section": "",
     "text": "Some documents have BOTH actual text and image text. In those cases, you’ll want to force any program to re-recognize it.↩︎\nIn a future chapter, we’ll look at how to split those into first and last.↩︎"
   },
   {
     "objectID": "advanced-openrefine.html#installing-openrefine",
     "href": "advanced-openrefine.html#installing-openrefine",
-    "title": "28  Open refine walkthrough",
-    "section": "28.1 Installing OpenRefine",
-    "text": "28.1 Installing OpenRefine\nDownload the latest version from the OpenRefine website and follow the instructions there to install.\n\nMac users: Unidentified developer\nFor Mac users, when you try to open it, it may tell you that you can’t because it’s from an unidentified developer. Go to the Applications folder on your computer, and use the CTL button before you click on the OpenRefine diamond. Once you successfully open it once, the program is whitelisted on your computer and you shouldn’t have to do it again.\n\n\nWindows users\nWindows users should take the “Windows kit with embedded Java” rather than trying to install the right version of Java yourself. You may have to go through the Windows Defender settings to set up an exclusion that allows Windows to run it, or get your system administrator to allow you to use it. You should only have to set this up once.\n\n\nLimitations\nOpenRefine works entirely in memory, and will start choking pretty quickly once you reach the 100,000 to 200,000 row range. You might need to partition your data or pull out just the columns you want to work on in order to clean a large dataset.\n\n\nFollowing the walkthrough\nSteps that you should take are shown in bullet points. The rest is explanation."
+    "title": "29  Open refine walkthrough",
+    "section": "29.1 Installing OpenRefine",
+    "text": "29.1 Installing OpenRefine\nDownload the latest version from the OpenRefine website and follow the instructions there to install.\n\nMac users: Unidentified developer\nFor Mac users, when you try to open it, it may tell you that you can’t because it’s from an unidentified developer. Go to the Applications folder on your computer, and use the CTL button before you click on the OpenRefine diamond. Once you successfully open it once, the program is whitelisted on your computer and you shouldn’t have to do it again.\n\n\nWindows users\nWindows users should take the “Windows kit with embedded Java” rather than trying to install the right version of Java yourself. You may have to go through the Windows Defender settings to set up an exclusion that allows Windows to run it, or get your system administrator to allow you to use it. You should only have to set this up once.\n\n\nLimitations\nOpenRefine works entirely in memory, and will start choking pretty quickly once you reach the 100,000 to 200,000 row range. You might need to partition your data or pull out just the columns you want to work on in order to clean a large dataset.\n\n\nFollowing the walkthrough\nSteps that you should take are shown in bullet points. The rest is explanation."
   },
   {
     "objectID": "advanced-openrefine.html#starting-openrefine",
     "href": "advanced-openrefine.html#starting-openrefine",
-    "title": "28  Open refine walkthrough",
-    "section": "28.2 Starting OpenRefine",
-    "text": "28.2 Starting OpenRefine\nOpenRefine is a Java-based program that runs inside your browser, but lives on your computer instead of the internet. It runs much better in Chrome than in Safari.\n\nFind the OpenRefine app on your computer, and double-click. When you first start it up, it will open with your default browser.\nIf it’s not Chrome, open a Chrome window and type the local address into its address bar: localhost:3333 or 127.0.0.1:3333\n\nOnce it’s open in your browser, you’ll see something that looks like this:\n\n\n\nfirst screen"
+    "title": "29  Open refine walkthrough",
+    "section": "29.2 Starting OpenRefine",
+    "text": "29.2 Starting OpenRefine\nOpenRefine is a Java-based program that runs inside your browser, but lives on your computer instead of the internet. It runs much better in Chrome than in Safari.\n\nFind the OpenRefine app on your computer, and double-click. When you first start it up, it will open with your default browser.\nIf it’s not Chrome, open a Chrome window and type the local address into its address bar: localhost:3333 or 127.0.0.1:3333\n\nOnce it’s open in your browser, you’ll see something that looks like this:\n\n\n\nfirst screen"
   },
   {
     "objectID": "advanced-openrefine.html#importing-data",
     "href": "advanced-openrefine.html#importing-data",
-    "title": "28  Open refine walkthrough",
-    "section": "28.3 Importing data",
-    "text": "28.3 Importing data\nOpenRefine works best on comma- or tab-delimited data. To create a dataset like that from R, use the function write_csv(\"path-to-csv-file.csv\") , which can be used like any other verb after a continuation pipe.\n\nTo follow along with this tutorial, download the sample data file and save it into your R project folder .\nIn OpenRefine, look on the left of the screen and choose Create Project-&gt;Choose files, and navigate to the folder where you saved it. Choose the file you just saved, and consider changing the name of the project to something other than the file name. I called mine “ppp refined”.1\n\n\n\n\nimport screen\n\n\nUsually, the default selections at the bottom work. They’re pretty self-explanatory, but if your data looks weird when it is imported, take a look for some options here.\nI usually keep everything as “text” because OpenRefine doesn’t work as well with dates as numbers as it does with text. In this case, I’ve removed the amounts and dates from the data that we’re working with.\n\nClick the “Create Project” button to continue.\n\n\nReturning to your project\nYou can also use that startup screen to re-open a project you’ve previously worked on. You don’t need to save it – OpenRefine saves it for you in the background. Use the Open tab to find your projects. If you need to move your project to another computer, you can use the Export options once it’s open.\n\n\nRename or copy columns\nEventually, you’ll want to re-import the results of your cleaning to your R or other programming environment, and it will cause you no end of trouble if you can’t match it up with the original data, or if columns have the same names.\n\nBefore you do any work on a column, be sure to either rename it or copy it into a new column. Someday you’ll thank me for this.\n\nTo rename something that is relatively simple to clean, such as the project_county in the PPP data, use the dropdown menu, and choose Edit column-&gt;Rename this column . I usually call it whatever the column name started as with a suffix _refined.\n\nRename the columns project_county and project_state now.\n\nTo make a copy, use the dropdown menu Edit column-&gt;Add column based on this column and give it a name."
+    "title": "29  Open refine walkthrough",
+    "section": "29.3 Importing data",
+    "text": "29.3 Importing data\nOpenRefine works best on comma- or tab-delimited data. To create a dataset like that from R, use the function write_csv(\"path-to-csv-file.csv\") , which can be used like any other verb after a continuation pipe.\n\nTo follow along with this tutorial, download the sample data file and save it into your R project folder .\nIn OpenRefine, look on the left of the screen and choose Create Project-&gt;Choose files, and navigate to the folder where you saved it. Choose the file you just saved, and consider changing the name of the project to something other than the file name. I called mine “ppp refined”.1\n\n\n\n\nimport screen\n\n\nUsually, the default selections at the bottom work. They’re pretty self-explanatory, but if your data looks weird when it is imported, take a look for some options here.\nI usually keep everything as “text” because OpenRefine doesn’t work as well with dates as numbers as it does with text. In this case, I’ve removed the amounts and dates from the data that we’re working with.\n\nClick the “Create Project” button to continue.\n\n\nReturning to your project\nYou can also use that startup screen to re-open a project you’ve previously worked on. You don’t need to save it – OpenRefine saves it for you in the background. Use the Open tab to find your projects. If you need to move your project to another computer, you can use the Export options once it’s open.\n\n\nRename or copy columns\nEventually, you’ll want to re-import the results of your cleaning to your R or other programming environment, and it will cause you no end of trouble if you can’t match it up with the original data, or if columns have the same names.\n\nBefore you do any work on a column, be sure to either rename it or copy it into a new column. Someday you’ll thank me for this.\n\nTo rename something that is relatively simple to clean, such as the project_county in the PPP data, use the dropdown menu, and choose Edit column-&gt;Rename this column . I usually call it whatever the column name started as with a suffix _refined.\n\nRename the columns project_county and project_state now.\n\nTo make a copy, use the dropdown menu Edit column-&gt;Add column based on this column and give it a name."
   },
   {
     "objectID": "advanced-openrefine.html#facets",
     "href": "advanced-openrefine.html#facets",
-    "title": "28  Open refine walkthrough",
-    "section": "28.4 Facets",
-    "text": "28.4 Facets\nFor many text columns, you can use facets to edit your data. Under your newly named project_state_refined, choose Facet-&gt;Text facet to get a menu on the left. It will then show you the values it found in that column, and the number of times each one appears, much like at group_by query.\nOpenRefine only acts on rows that are selected, which makes using facets and filtering so powerful. You never have to worry that you’re changing something that you don’t see.\n\n\n\nfacet\n\n\n\nTo edit one by one, click on the state name you want to change, and choose the Edit option. You might combine all of the other states into an “OTHER” state.\n\nA faster way to change everything that is selected at once is through the column’s menu.\n\nSelect “AZ” from the facet list, then choose invert from the top of the box to select everything except AZ.\nUnder the column dropdown, choose Facet-&gt;Edit Cells-&gt;Transform, and type “OTHER” (with the quotes) into the value box. (If you forget the quotes, you’ll see it come up with Null as an answer because it’s looking for a variable called other, not the word “other”)\nWith the same selection on, change the county name to “OUT OF STATE” using the same method."
+    "title": "29  Open refine walkthrough",
+    "section": "29.4 Facets",
+    "text": "29.4 Facets\nFor many text columns, you can use facets to edit your data. Under your newly named project_state_refined, choose Facet-&gt;Text facet to get a menu on the left. It will then show you the values it found in that column, and the number of times each one appears, much like at group_by query.\nOpenRefine only acts on rows that are selected, which makes using facets and filtering so powerful. You never have to worry that you’re changing something that you don’t see.\n\n\n\nfacet\n\n\n\nTo edit one by one, click on the state name you want to change, and choose the Edit option. You might combine all of the other states into an “OTHER” state.\n\nA faster way to change everything that is selected at once is through the column’s menu.\n\nSelect “AZ” from the facet list, then choose invert from the top of the box to select everything except AZ.\nUnder the column dropdown, choose Facet-&gt;Edit Cells-&gt;Transform, and type “OTHER” (with the quotes) into the value box. (If you forget the quotes, you’ll see it come up with Null as an answer because it’s looking for a variable called other, not the word “other”)\nWith the same selection on, change the county name to “OUT OF STATE” using the same method."
   },
   {
     "objectID": "advanced-openrefine.html#clustering",
     "href": "advanced-openrefine.html#clustering",
-    "title": "28  Open refine walkthrough",
-    "section": "28.5 Clustering",
-    "text": "28.5 Clustering\nOne reason a lot of reporters use OpenRefine is to take advantage of its fuzzy-matching to clean items like names and addresses. It provides about a dozen different algorithms that provide suggestions on ways to change unstandardized values in a column that often make a lot of sense.\nThis walkthrough will work on the city name.\n\nMake a copy of the city and call it borrower_city_refined by using the dropdown Edit column-&gt;Add column based on this column . Otherwise you risk over-writing things incorrectly, and you’ll have to try to figure out where you went wrong.\n\n\n\n\nadd column\n\n\n\nNow create a facet on this new column.\n\nYou should see all of the bad values that were entered into that column by lenders and applicants. No one cared whether it was a valid city name.\n\nClustering details\nYou’ll often use many of the clustering approaches, then finish up with some other filters and facets when the clustering doesn’t work.\nSome algorithms work better than others – some don’t catch natural clusters, and others find a lot of things to cluster that shouldn’t be combined. It’s unimportant for us what they actually do behind the scenes. Insetad, just know that they are different methods to find near-exact matches among character values.\nOne might look for all words that have the same letters but in any order, another might look for everything that matches the shortest phrase, and another might look for any 2 words in common. The order suggested by OpenRefine generally goes from the most strict to the most forgiving method, and from the easiest on your computer to the most resource-intensive. 2\n\nOn the facet, press the button that says “Cluster”. The first one chosen, called “key collision” by “fingerprint”, doesn’t come up with any suggestions.\nUse the dropdown menu in the Keying Function box to choose the next option, “ngram-fingerprint”, and leave the option at 2.\n\n\n\n\ncluster\n\n\nNow you can see how close some of the choices are.\n\nLook through your choices, and select any that you want to combine. OpenRefine will change the least popular entry to the most popular, but you can change what you want it to be in the box provided. (At first, you’ll probably “Select all” and then just un-select the bad ones.)\nKeep going with other clustering methods, turning off the ones you’re not sure should be combined and editing the value you want it to be. After exploring all of the options, I often end by going back to the ngram-fingerprint method and changing it to Ngram size of 1. That often catches the last remnants of one-word cities.\n\n\n\n\nngram\n\n\n\nSpend about 10-15 minutes cleaning up the cities, and see how close you can get to a good list.\n\n\nPro tip: Don’t try to do everything through clustering. You can’t really fine-tune the choices, so just stop when it’s getting close. You can then do the remaining things you care about manually. Try not to get too obsessed with getting it perfect. You’ll spend days on things that won’t matter, especially in early stages of your reporting when you don’t know what you care about.\n\n\n\nFacet for long entries\nIn this list, though, there are addresses that are listed as cities. We may just want to remove those city names entirely, since we don’t really know what they are. There are several ways you might do this:\n\nUse the Facet-&gt;Customized facets-&gt;Text length facet menu item to create a facet by string length. Use the scrollbar to choose, say, more than 20 or 25 characters.\nGo through the list and select the ones that should not be changed, then\nInvert the selection and use the Transform menu to change the city to “Unknown”.\n\nThat got rid of most of them. You can go back to the original facet and select the ones at the top and change them as well.\nThe list isn’t perfect, but it’s much better than what we started with."
+    "title": "29  Open refine walkthrough",
+    "section": "29.5 Clustering",
+    "text": "29.5 Clustering\nOne reason a lot of reporters use OpenRefine is to take advantage of its fuzzy-matching to clean items like names and addresses. It provides about a dozen different algorithms that provide suggestions on ways to change unstandardized values in a column that often make a lot of sense.\nThis walkthrough will work on the city name.\n\nMake a copy of the city and call it borrower_city_refined by using the dropdown Edit column-&gt;Add column based on this column . Otherwise you risk over-writing things incorrectly, and you’ll have to try to figure out where you went wrong.\n\n\n\n\nadd column\n\n\n\nNow create a facet on this new column.\n\nYou should see all of the bad values that were entered into that column by lenders and applicants. No one cared whether it was a valid city name.\n\nClustering details\nYou’ll often use many of the clustering approaches, then finish up with some other filters and facets when the clustering doesn’t work.\nSome algorithms work better than others – some don’t catch natural clusters, and others find a lot of things to cluster that shouldn’t be combined. It’s unimportant for us what they actually do behind the scenes. Insetad, just know that they are different methods to find near-exact matches among character values.\nOne might look for all words that have the same letters but in any order, another might look for everything that matches the shortest phrase, and another might look for any 2 words in common. The order suggested by OpenRefine generally goes from the most strict to the most forgiving method, and from the easiest on your computer to the most resource-intensive. 2\n\nOn the facet, press the button that says “Cluster”. The first one chosen, called “key collision” by “fingerprint”, doesn’t come up with any suggestions.\nUse the dropdown menu in the Keying Function box to choose the next option, “ngram-fingerprint”, and leave the option at 2.\n\n\n\n\ncluster\n\n\nNow you can see how close some of the choices are.\n\nLook through your choices, and select any that you want to combine. OpenRefine will change the least popular entry to the most popular, but you can change what you want it to be in the box provided. (At first, you’ll probably “Select all” and then just un-select the bad ones.)\nKeep going with other clustering methods, turning off the ones you’re not sure should be combined and editing the value you want it to be. After exploring all of the options, I often end by going back to the ngram-fingerprint method and changing it to Ngram size of 1. That often catches the last remnants of one-word cities.\n\n\n\n\nngram\n\n\n\nSpend about 10-15 minutes cleaning up the cities, and see how close you can get to a good list.\n\n\nPro tip: Don’t try to do everything through clustering. You can’t really fine-tune the choices, so just stop when it’s getting close. You can then do the remaining things you care about manually. Try not to get too obsessed with getting it perfect. You’ll spend days on things that won’t matter, especially in early stages of your reporting when you don’t know what you care about.\n\n\n\nFacet for long entries\nIn this list, though, there are addresses that are listed as cities. We may just want to remove those city names entirely, since we don’t really know what they are. There are several ways you might do this:\n\nUse the Facet-&gt;Customized facets-&gt;Text length facet menu item to create a facet by string length. Use the scrollbar to choose, say, more than 20 or 25 characters.\nGo through the list and select the ones that should not be changed, then\nInvert the selection and use the Transform menu to change the city to “Unknown”.\n\nThat got rid of most of them. You can go back to the original facet and select the ones at the top and change them as well.\nThe list isn’t perfect, but it’s much better than what we started with."
   },
   {
     "objectID": "advanced-openrefine.html#filtering",
     "href": "advanced-openrefine.html#filtering",
-    "title": "28  Open refine walkthrough",
-    "section": "28.6 Filtering",
-    "text": "28.6 Filtering\nEven after fixing all of the cities we can through clustering and faceting, there are usually some bad entries left over.\n\nSimple filter\n\nInstead of a Text Facet, choose Text filter from the dropdown menu. Generally, I use regular expressions in this box so I can control where in the column my target word or letters falls.\nTry filtering for ^Phoe after checking the “regular expression” box, and you’ll see a few stray entries for Phoenix that you can change in the facet window.\n\nI tend to do this for anything I really care about getting right, such as larger cities or towns that I am going to report on. It’s often not worth the time to get a large list 100% accurate.\n\n\nComplex filter\nA complex filter will help you tag items for later use. One example is to tag items that contain apartment numbers, suite numbers or floor numbers after the address. Eventually, we may want to split the address column in two – one piece for the street address, and one piece for the detailed location.\nFiltering for them first ensures that you won’t inadvertently split something that hasn’t met your condition in the first place.\nRegular expressions work here the same way they did in str_detect in R, except that there is one backslash rather than two for special characters.\n\nHere is an example of how to find anything that has the word Suite or Ste after at least 10 other characters: .{10,}\\bS(ui)?te\\b\n\n\n\n\nexample regex\n\n\nUsing these filters, we can start splitting the values into their pieces.\n\n\nSplit by regex\n\nChoose Edit column-&gt;Split into several columns… from the dropdown menu, then use the regular expression as the “split” character. Make sure to un-check the “Guess cell type” and “Remove this column” boxes.\n\nYou can go back and look for entries that have “floor” or “apt” the same way, and split them in subsequent steps.\n\n\nSplit using OpenRefine programming language\nThat took care of the ones that had Suite numbers in them. Now we have to get the ones that have Apartment numbers in them. Use a regular expression to filter for the apartment numbers, and a Customized Facet-&gt;Facet by null to pick out the ones that haven’t already been split by our previous step.\nOpenRefine has its own special language called GREL, which is really hard to use. But we’ll have to use it for this set because we’ve already split something else. This time, we have to put a value in the two columns based on pieces of a regular expression.\n\nWith your filter on, choose borrower_address_refined_2’s Transform menu. I’ll just show you how to do this, since it’s hard to explain:\n\n-In the “Expression” box, copy and paste this code:\ncells[\"borrower_address_refined\"].value.match(/(.+)?\\b(Apt .*)\\b$/)[1]\nThis tells OpenRefine to extract two pieces from the original address column: The beginning (.+)?, the word “Apt” followed by anything at the end. The [1] tells OpenRefine to take the 2nd of those (its counting starts at zero, not one) and put it in the same column that we used for the Suite numbers. (NOTE: The GREL regular expressions are indicated by the two slashes at the beginning and end of the phrase (/). This is typical for regular expressions, and it distinguishes them from plain text.)\n\nRepeat the process, in the borrower_address_refined_1 column’s transformation, and use the index [0] at the end to get the cleaned street address."
+    "title": "29  Open refine walkthrough",
+    "section": "29.6 Filtering",
+    "text": "29.6 Filtering\nEven after fixing all of the cities we can through clustering and faceting, there are usually some bad entries left over.\n\nSimple filter\n\nInstead of a Text Facet, choose Text filter from the dropdown menu. Generally, I use regular expressions in this box so I can control where in the column my target word or letters falls.\nTry filtering for ^Phoe after checking the “regular expression” box, and you’ll see a few stray entries for Phoenix that you can change in the facet window.\n\nI tend to do this for anything I really care about getting right, such as larger cities or towns that I am going to report on. It’s often not worth the time to get a large list 100% accurate.\n\n\nComplex filter\nA complex filter will help you tag items for later use. One example is to tag items that contain apartment numbers, suite numbers or floor numbers after the address. Eventually, we may want to split the address column in two – one piece for the street address, and one piece for the detailed location.\nFiltering for them first ensures that you won’t inadvertently split something that hasn’t met your condition in the first place.\nRegular expressions work here the same way they did in str_detect in R, except that there is one backslash rather than two for special characters.\n\nHere is an example of how to find anything that has the word Suite or Ste after at least 10 other characters: .{10,}\\bS(ui)?te\\b\n\n\n\n\nexample regex\n\n\nUsing these filters, we can start splitting the values into their pieces.\n\n\nSplit by regex\n\nChoose Edit column-&gt;Split into several columns… from the dropdown menu, then use the regular expression as the “split” character. Make sure to un-check the “Guess cell type” and “Remove this column” boxes.\n\nYou can go back and look for entries that have “floor” or “apt” the same way, and split them in subsequent steps.\n\n\nSplit using OpenRefine programming language\nThat took care of the ones that had Suite numbers in them. Now we have to get the ones that have Apartment numbers in them. Use a regular expression to filter for the apartment numbers, and a Customized Facet-&gt;Facet by null to pick out the ones that haven’t already been split by our previous step.\nOpenRefine has its own special language called GREL, which is really hard to use. But we’ll have to use it for this set because we’ve already split something else. This time, we have to put a value in the two columns based on pieces of a regular expression.\n\nWith your filter on, choose borrower_address_refined_2’s Transform menu. I’ll just show you how to do this, since it’s hard to explain:\n\n-In the “Expression” box, copy and paste this code:\ncells[\"borrower_address_refined\"].value.match(/(.+)?\\b(Apt .*)\\b$/)[1]\nThis tells OpenRefine to extract two pieces from the original address column: The beginning (.+)?, the word “Apt” followed by anything at the end. The [1] tells OpenRefine to take the 2nd of those (its counting starts at zero, not one) and put it in the same column that we used for the Suite numbers. (NOTE: The GREL regular expressions are indicated by the two slashes at the beginning and end of the phrase (/). This is typical for regular expressions, and it distinguishes them from plain text.)\n\nRepeat the process, in the borrower_address_refined_1 column’s transformation, and use the index [0] at the end to get the cleaned street address."
   },
   {
     "objectID": "advanced-openrefine.html#finishing-up",
     "href": "advanced-openrefine.html#finishing-up",
-    "title": "28  Open refine walkthrough",
-    "section": "28.7 Finishing up",
-    "text": "28.7 Finishing up\n\nTry using the clustering and other transformation options to clean up the borrower names. Only change the ones you are sure should be combined.\n\nOnce you’re done, export the project to a new .csv file, and then import it into R. You can then select the cleaned up columns and the loan_id , then join it back to the original PPP data frame. Now you have a good set of cities, better ways to look at popular addresses, etc.\nIf you want to try another exercise, here is an old workshop I did for IRE in 2016."
+    "title": "29  Open refine walkthrough",
+    "section": "29.7 Finishing up",
+    "text": "29.7 Finishing up\n\nTry using the clustering and other transformation options to clean up the borrower names. Only change the ones you are sure should be combined.\n\nOnce you’re done, export the project to a new .csv file, and then import it into R. You can then select the cleaned up columns and the loan_id , then join it back to the original PPP data frame. Now you have a good set of cities, better ways to look at popular addresses, etc.\nIf you want to try another exercise, here is an old workshop I did for IRE in 2016."
   },
   {
     "objectID": "advanced-openrefine.html#footnotes",
     "href": "advanced-openrefine.html#footnotes",
-    "title": "28  Open refine walkthrough",
+    "title": "29  Open refine walkthrough",
     "section": "",
     "text": "Note: It lives in a different place than your folder on your computer. It’s actually a little hard to find, but you shouldn’t have to unless you’re moving between computers. You can always export all of the steps so that you can do them over on a new computer.↩︎\nThere is something called “reconciliation” in OpenRefine that tries to find matches in OpenCorporates lists of companies, or in a list that you provide. We’re not getting into that.↩︎"
   },
   {
     "objectID": "advanced-scrape1.html#where-reporters-get-data",
     "href": "advanced-scrape1.html#where-reporters-get-data",
-    "title": "29  Scraping without programming",
-    "section": "29.1 Where reporters get data",
-    "text": "29.1 Where reporters get data\nReporters can get data from people, using FOIA, asking nicely or by finding a whistleblower to leak it. But we often also get data from publicly published sources, usually on the web.\nThere are three ways to get data from the web:\n\nDownload it, or use an API1. In these cases, the makers of the data have specifically offered it up for your use in a useful format. We’ll cover API’s later, but don’t forget to study the site for a download link or option. If there isn’t one on a government site, you might call the agency and ask that they add one. They might just do it. We’ve been using downloadable data throughout this book.\nFind it on your browser. Often the person making the website delivers structured data to your browser as a convenience. It’s easier for them to make interactive items on their page by using data they’ve already delivered in visualizations and tables. It also reduces the loads on their servers. These are usually in JSON format. You might be able to find it right on your computer. It’s a miracle!\nScrape it. This set of chapters goes over how to scrape content that is delivered in HTML form – a web page. There would be other methods to scrape PDFs, which can be easy or hard."
+    "title": "30  Scraping without programming",
+    "section": "30.1 Where reporters get data",
+    "text": "30.1 Where reporters get data\nReporters can get data from people, using FOIA, asking nicely or by finding a whistleblower to leak it. But we often also get data from publicly published sources, usually on the web.\nThere are three ways to get data from the web:\n\nDownload it, or use an API1. In these cases, the makers of the data have specifically offered it up for your use in a useful format. We’ll cover API’s later, but don’t forget to study the site for a download link or option. If there isn’t one on a government site, you might call the agency and ask that they add one. They might just do it. We’ve been using downloadable data throughout this book.\nFind it on your browser. Often the person making the website delivers structured data to your browser as a convenience. It’s easier for them to make interactive items on their page by using data they’ve already delivered in visualizations and tables. It also reduces the loads on their servers. These are usually in JSON format. You might be able to find it right on your computer. It’s a miracle!\nScrape it. This set of chapters goes over how to scrape content that is delivered in HTML form – a web page. There would be other methods to scrape PDFs, which can be easy or hard."
   },
   {
     "objectID": "advanced-scrape1.html#a-json-miracle-walkthrough",
     "href": "advanced-scrape1.html#a-json-miracle-walkthrough",
-    "title": "29  Scraping without programming",
-    "section": "29.2 A json miracle walkthrough",
-    "text": "29.2 A json miracle walkthrough\nThis walkthrough shows you how to find some json in your browser. Use the Chrome browser for this - Firefox and Safari also have similar features, but they look different.\nHere is a simple page that will show you what the json looks like and how to extract it. This is what a human sees:\n\nThe table of presidents is actually produced using a small javascript program inside the HTML that walks through each item and lists it as a row.\n\nOpen the page in Chrome, then right-click anywhere on the page and choose “Inspect”. It may appear as just two items - the “head” and the “body”. But notice the little arrows - they show you that there is more content underneath. For now, we’ll ignore this, but it will be important later.\nChoose the Network tab, then re-load the page.\n\n\n\n\ninspect network\n\n\nThis shows you everything that the browser is attempting to load into your browser. (You may not see the “favicon” item. I have no idea why it’s showing up on mine - it’s not been requested!)\nYou can ignore most of this. Importantly, the “simple-page.html” is the actual page, and the “simple.json” is the data! Click on the simple.json row, then choose the “Response” tab:\n\n\n\nsimple json\n\n\nThat’s what json looks like - a list of rows within an item called “presidents”, each identified by the name of the column they’ll become.2\n\nRight-click on the simple.json file name, and you’ll see a lot of options. Choose the one that says Copy-&gt;Copy link address.\nGo to a new browser window and search for “json to csv”. This one is one that I often see first.\nPaste your copied link in the tab that says, “Enter URL” and press “Load URL”. You’ll see an option to copy the result as a csv file!\n\n\n\n\njson to csv\n\n\n\nA harder example\nThat was easy! But it’s also trivial. However, this method can often save you from having to page through results of a page. One example is the Maricopa County nightly list of mugshots, which may have several hundred new entries each day. Here’s what today’s looked like on a desktop browser (it looks different on a smaller screen).3:\n\n\n\nmcso list\n\n\nIt looks like you’d have to go through each of the five pages to get all of the names of people who were booked into jail that night, but often a json file contains all of them – they’re just showing you one page at a time.\n\nThis page won’t let you right-click to get the inspector. Instead, on a Mac, press Opt-Cmd-C to open the inspector window. (I think it’s Shft-Ctl-C on Windows, but I’m not sure.)\n\nThis looks like a mess! Don’t worry. Switch to the Network tab, and re-load the page. This time, there are dozens of different things that get loaded on your page, and none of them are obviously json. You have a few strategies to find it.\n\nPress the “Fetch/XHR” tab to see if it shows up there. Use the “Preview” tab to look at what each of them is, and, miracle of miracles, it’s the third one on the list! Even better, it has all 425 entries! (It looks like they’re split into groups of 100, but they really aren’t.)\n\n\n\n\nmiracle 1\n\n\n\nRight-click on the name of the file, and choose Copy-&gt;Copy link, and repeat the process above to convert it to a CSV file.\n\n\n\n\n\n\n\nNote\n\n\n\nIn 2022, the county began limiting the number of results to some random list of 300. It appears that searching for an inmate only checks those first 300 results. (It could have been a coincidence that there were exactly 300 inmate on Jan 3, 2023, but I doubt it.)\n\n\n\n\nAn even harder example\nThe New York Times maintains a map with the vaccination rates for various demographic groups by county on its website. At first, the Times didn’t provide a Github repo for the data. How can we extract the data from this map?\nThe easiest way would be to see if it contains a json miracle!\n\nOpen your inspector panel\nCopy and paste the link to the map page, and open it in your Chrome browser with the inpsectors showing.\nSwitch to the network tab. (If you opened the map before opening the inspector, reload it now. )\n\nYikes! The “Fetch /XHR” button doesn’t help us much here. There are too many different json files to check. We could look one by one and see if they’re right, but sometimes that’s just too hard. Instead,\n\nOpen the “Search” button on your inspector (it’s different from the Filter), and type in a county name (this one is “Maricopa”). You should see only a few of them. The most promising is the “doses_county.json”, so try that one first:\n\n\n\n\nnyt example\n\n\nThis time, it’s hard to find the item in the list of files in the browser. Instead, right-click in the “Preview” area, and copy the object. You can paste that into the box in the JSON to csv converter instead of entering a URL."
+    "title": "30  Scraping without programming",
+    "section": "30.2 A json miracle walkthrough",
+    "text": "30.2 A json miracle walkthrough\nThis walkthrough shows you how to find some json in your browser. Use the Chrome browser for this - Firefox and Safari also have similar features, but they look different.\nHere is a simple page that will show you what the json looks like and how to extract it. This is what a human sees:\n\nThe table of presidents is actually produced using a small javascript program inside the HTML that walks through each item and lists it as a row.\n\nOpen the page in Chrome, then right-click anywhere on the page and choose “Inspect”. It may appear as just two items - the “head” and the “body”. But notice the little arrows - they show you that there is more content underneath. For now, we’ll ignore this, but it will be important later.\nChoose the Network tab, then re-load the page.\n\n\n\n\ninspect network\n\n\nThis shows you everything that the browser is attempting to load into your browser. (You may not see the “favicon” item. I have no idea why it’s showing up on mine - it’s not been requested!)\nYou can ignore most of this. Importantly, the “simple-page.html” is the actual page, and the “simple.json” is the data! Click on the simple.json row, then choose the “Response” tab:\n\n\n\nsimple json\n\n\nThat’s what json looks like - a list of rows within an item called “presidents”, each identified by the name of the column they’ll become.2\n\nRight-click on the simple.json file name, and you’ll see a lot of options. Choose the one that says Copy-&gt;Copy link address.\nGo to a new browser window and search for “json to csv”. This one is one that I often see first.\nPaste your copied link in the tab that says, “Enter URL” and press “Load URL”. You’ll see an option to copy the result as a csv file!\n\n\n\n\njson to csv\n\n\n\nA harder example\nThat was easy! But it’s also trivial. However, this method can often save you from having to page through results of a page. One example is the Maricopa County nightly list of mugshots, which may have several hundred new entries each day. Here’s what today’s looked like on a desktop browser (it looks different on a smaller screen).3:\n\n\n\nmcso list\n\n\nIt looks like you’d have to go through each of the five pages to get all of the names of people who were booked into jail that night, but often a json file contains all of them – they’re just showing you one page at a time.\n\nThis page won’t let you right-click to get the inspector. Instead, on a Mac, press Opt-Cmd-C to open the inspector window. (I think it’s Shft-Ctl-C on Windows, but I’m not sure.)\n\nThis looks like a mess! Don’t worry. Switch to the Network tab, and re-load the page. This time, there are dozens of different things that get loaded on your page, and none of them are obviously json. You have a few strategies to find it.\n\nPress the “Fetch/XHR” tab to see if it shows up there. Use the “Preview” tab to look at what each of them is, and, miracle of miracles, it’s the third one on the list! Even better, it has all 425 entries! (It looks like they’re split into groups of 100, but they really aren’t.)\n\n\n\n\nmiracle 1\n\n\n\nRight-click on the name of the file, and choose Copy-&gt;Copy link, and repeat the process above to convert it to a CSV file.\n\n\n\n\n\n\n\nNote\n\n\n\nIn 2022, the county began limiting the number of results to some random list of 300. It appears that searching for an inmate only checks those first 300 results. (It could have been a coincidence that there were exactly 300 inmate on Jan 3, 2023, but I doubt it.)\n\n\n\n\nAn even harder example\nThe New York Times maintains a map with the vaccination rates for various demographic groups by county on its website. At first, the Times didn’t provide a Github repo for the data. How can we extract the data from this map?\nThe easiest way would be to see if it contains a json miracle!\n\nOpen your inspector panel\nCopy and paste the link to the map page, and open it in your Chrome browser with the inpsectors showing.\nSwitch to the network tab. (If you opened the map before opening the inspector, reload it now. )\n\nYikes! The “Fetch /XHR” button doesn’t help us much here. There are too many different json files to check. We could look one by one and see if they’re right, but sometimes that’s just too hard. Instead,\n\nOpen the “Search” button on your inspector (it’s different from the Filter), and type in a county name (this one is “Maricopa”). You should see only a few of them. The most promising is the “doses_county.json”, so try that one first:\n\n\n\n\nnyt example\n\n\nThis time, it’s hard to find the item in the list of files in the browser. Instead, right-click in the “Preview” area, and copy the object. You can paste that into the box in the JSON to csv converter instead of entering a URL."
   },
   {
     "objectID": "advanced-scrape1.html#no-json-no-problem-maybe",
     "href": "advanced-scrape1.html#no-json-no-problem-maybe",
-    "title": "29  Scraping without programming",
-    "section": "29.3 No json? No problem (maybe)",
-    "text": "29.3 No json? No problem (maybe)\nYou may not be able to find a json file – either it’s too hard to find, or it’s not useful, or it doesn’t exist. For a simple page, there’s no problem getting the data in Google Sheets. (This is one area where Excel lags behind Google Sheets.) We’ll go into HTML tags in more depth in the next chapter, but if your data is held in a table or structured list, you can import it directly into Google sheets.\nNote that this trick is really only useful if your page doesn’t change a lot, or if you just want a one-time snapshot. It doesn’t automatically update, and I don’t know how to capture changes – it would involve a Google scripting program, which I don’t know how to do. I’ve never learned because I usually only gather data for my own use, and it’s easier to program it than to finagle Google Sheets.\nThe trick to using Google Sheets is to find a table tag (&lt;table&gt;) or a list tag (&lt;ul&gt; or &lt;ol&gt;) that contains the data you want.\nHere’s an example, taken from a previous year’s MAIJ cohort: Reporters wanted to know whether Scottsdale was relatively unique in its city council structure, which has no districts. All members are at-large. Some research suggests that this disempowers non-white or less wealthy areas, because more privileged residents are often more active in local politics.\nThe reporters knew it was rare, but one question nagged at them: Was Scottsdale the largest city in the nation with a purely at-large council? That would make a nice tidbit for the story, but it wasn’t worth a major data collection endeavor.\nBallotpedia, a crowdsourced website with information on local governments, had collected a page of city council officials in the 100 largest cities in the US. Extracting this information into a structured table, then using regular expressions, could help make that a relatively simple job. This could even be done in Google Sheets, which also has a regular expression implementation. Because it’s so rare, just getting a list of cities that had no district or ward membership would give them a place to start looking up populations.\nThis information is stored in an HTML table, identified by the “\n\n” element.\nRight-click on the page, and open your inspector. It looks like a mess, but you can search for tables using a simple “Find” using Cmd (or Ctl) -F.\nYou may notice it says you can find by string, selector or XPath.\nIn the box, enter “&lt;table” (with the opening “&lt;”, but no closing one.). You should see “1 of 12” in the result box. As you go through the list, the currently selected table will be highlighted. When you hit “3 of 12”, you’ll notice that the browser has selected the table you want. That’s the information we need.\nOpen a Google Sheet, and copy the page URL to cell A1. This just makes it easier to construct the formula to extract the table.\nIn cell A3, enter the following formula:\n      \n      =ImportHTML(A1, \"table\", 3)That means, “Go to the web address listed in cell A1, look for”table” tags, then return whatever is in the third one.”\n\n\ngooglesheet\n\nWhen you hit “enter” the whole table will populate on your Google Sheet. Unfortunately, you can’t get the link to the city from this method, which means you don’t have a good way to extract a city name. We’ll come back to this when we go to scraping in R. (There is a way to get this in Google Sheets, but it’s not very reliable – it will choke as soon as it encounters a missing URL.)But if you just need the text of a table or list in a spreadsheet, this is an easy way to get it. (To get a list from an “ol” or “ul” (ordered and unordered lists) tag, use “list” instead of “table”.)\n29.4 Recap\nSometimes – especially on modern websites that create interactive elements on the fly – the data you need is already sitting on your computer. In fact, it’s quite hard to scrape those in other ways because the HTML is created when it’s loaded into your browser.\nBut when it’s not, there may be another simple way to get the content.\nThe problem is that getting the content without programming can leave you unsatisfied because you can only get the text, not any of the underlying information. The next chapter shows you one method of getting more information from a web page using CSS selectors.\nI sometimes use a Chrome extension called “Chrome Scraper” to get slightly more complex information out of a website, which uses a language called XPath to parse a web page. It’s harder than the CSS selector method, though, so I’m skipping it for now.\n\n\n\n\n\n“application programming interface”↩︎\nIn R, our style was to name columns in lower case with words separated by underscores. In Javascript, the custom is usually called “camel case”, with words smushed together and the first letter of each upper cased. It’s just a custom, not a rule.↩︎\nI’m hiding names of people to the extent possible, and won’t list them in text here - they’ll only be in the images. Instead, I’ll show pictures of how to find the json when a name is necessary. Although this book is probably not indexed by Google, it’s possible that it could be some day, and I don’t want their names to show up in a Google search.↩︎\n\n\n  \n      \n         28  Open refine walkthrough\n                \n  \n  \n      \n        30  Introduction to scraping in R \n      \n  \n\n  \n    Data reporting for investigative journalism, Spring 2024, written by Sarah Cohen   \n    \n       \n    \n    Produced with Quarto"
+    "title": "30  Scraping without programming",
+    "section": "30.3 No json? No problem (maybe)",
+    "text": "30.3 No json? No problem (maybe)\nYou may not be able to find a json file – either it’s too hard to find, or it’s not useful, or it doesn’t exist. For a simple page, there’s no problem getting the data in Google Sheets. (This is one area where Excel lags behind Google Sheets.) We’ll go into HTML tags in more depth in the next chapter, but if your data is held in a table or structured list, you can import it directly into Google sheets.\nNote that this trick is really only useful if your page doesn’t change a lot, or if you just want a one-time snapshot. It doesn’t automatically update, and I don’t know how to capture changes – it would involve a Google scripting program, which I don’t know how to do. I’ve never learned because I usually only gather data for my own use, and it’s easier to program it than to finagle Google Sheets.\nThe trick to using Google Sheets is to find a table tag (&lt;table&gt;) or a list tag (&lt;ul&gt; or &lt;ol&gt;) that contains the data you want.\nHere’s an example, taken from a previous year’s MAIJ cohort: Reporters wanted to know whether Scottsdale was relatively unique in its city council structure, which has no districts. All members are at-large. Some research suggests that this disempowers non-white or less wealthy areas, because more privileged residents are often more active in local politics.\nThe reporters knew it was rare, but one question nagged at them: Was Scottsdale the largest city in the nation with a purely at-large council? That would make a nice tidbit for the story, but it wasn’t worth a major data collection endeavor.\nBallotpedia, a crowdsourced website with information on local governments, had collected a page of city council officials in the 100 largest cities in the US. Extracting this information into a structured table, then using regular expressions, could help make that a relatively simple job. This could even be done in Google Sheets, which also has a regular expression implementation. Because it’s so rare, just getting a list of cities that had no district or ward membership would give them a place to start looking up populations.\nThis information is stored in an HTML table, identified by the “\n\n” element.\nRight-click on the page, and open your inspector. It looks like a mess, but you can search for tables using a simple “Find” using Cmd (or Ctl) -F.\nYou may notice it says you can find by string, selector or XPath.\nIn the box, enter “&lt;table” (with the opening “&lt;”, but no closing one.). You should see “1 of 12” in the result box. As you go through the list, the currently selected table will be highlighted. When you hit “3 of 12”, you’ll notice that the browser has selected the table you want. That’s the information we need.\nOpen a Google Sheet, and copy the page URL to cell A1. This just makes it easier to construct the formula to extract the table.\nIn cell A3, enter the following formula:\n      \n      =ImportHTML(A1, \"table\", 3)That means, “Go to the web address listed in cell A1, look for”table” tags, then return whatever is in the third one.”\n\n\ngooglesheet\n\nWhen you hit “enter” the whole table will populate on your Google Sheet. Unfortunately, you can’t get the link to the city from this method, which means you don’t have a good way to extract a city name. We’ll come back to this when we go to scraping in R. (There is a way to get this in Google Sheets, but it’s not very reliable – it will choke as soon as it encounters a missing URL.)But if you just need the text of a table or list in a spreadsheet, this is an easy way to get it. (To get a list from an “ol” or “ul” (ordered and unordered lists) tag, use “list” instead of “table”.)\n30.4 Recap\nSometimes – especially on modern websites that create interactive elements on the fly – the data you need is already sitting on your computer. In fact, it’s quite hard to scrape those in other ways because the HTML is created when it’s loaded into your browser.\nBut when it’s not, there may be another simple way to get the content.\nThe problem is that getting the content without programming can leave you unsatisfied because you can only get the text, not any of the underlying information. The next chapter shows you one method of getting more information from a web page using CSS selectors.\nI sometimes use a Chrome extension called “Chrome Scraper” to get slightly more complex information out of a website, which uses a language called XPath to parse a web page. It’s harder than the CSS selector method, though, so I’m skipping it for now.\n\n\n\n\n\n“application programming interface”↩︎\nIn R, our style was to name columns in lower case with words separated by underscores. In Javascript, the custom is usually called “camel case”, with words smushed together and the first letter of each upper cased. It’s just a custom, not a rule.↩︎\nI’m hiding names of people to the extent possible, and won’t list them in text here - they’ll only be in the images. Instead, I’ll show pictures of how to find the json when a name is necessary. Although this book is probably not indexed by Google, it’s possible that it could be some day, and I don’t want their names to show up in a Google search.↩︎\n\n\n  \n      \n         29  Open refine walkthrough\n                \n  \n  \n      \n        31  Introduction to scraping in R \n      \n  \n\n  \n    Data reporting for investigative journalism, Spring 2024, written by Sarah Cohen   \n    \n       \n    \n    Produced with Quarto"
   },
   {
     "objectID": "advanced-scrape1.html#recap",
     "href": "advanced-scrape1.html#recap",
-    "title": "29  Scraping without programming",
-    "section": "29.4 Recap",
-    "text": "29.4 Recap\nSometimes – especially on modern websites that create interactive elements on the fly – the data you need is already sitting on your computer. In fact, it’s quite hard to scrape those in other ways because the HTML is created when it’s loaded into your browser.\nBut when it’s not, there may be another simple way to get the content.\nThe problem is that getting the content without programming can leave you unsatisfied because you can only get the text, not any of the underlying information. The next chapter shows you one method of getting more information from a web page using CSS selectors.\nI sometimes use a Chrome extension called “Chrome Scraper” to get slightly more complex information out of a website, which uses a language called XPath to parse a web page. It’s harder than the CSS selector method, though, so I’m skipping it for now."
+    "title": "30  Scraping without programming",
+    "section": "30.4 Recap",
+    "text": "30.4 Recap\nSometimes – especially on modern websites that create interactive elements on the fly – the data you need is already sitting on your computer. In fact, it’s quite hard to scrape those in other ways because the HTML is created when it’s loaded into your browser.\nBut when it’s not, there may be another simple way to get the content.\nThe problem is that getting the content without programming can leave you unsatisfied because you can only get the text, not any of the underlying information. The next chapter shows you one method of getting more information from a web page using CSS selectors.\nI sometimes use a Chrome extension called “Chrome Scraper” to get slightly more complex information out of a website, which uses a language called XPath to parse a web page. It’s harder than the CSS selector method, though, so I’m skipping it for now."
   },
   {
     "objectID": "advanced-scrape1.html#footnotes",
     "href": "advanced-scrape1.html#footnotes",
-    "title": "29  Scraping without programming",
+    "title": "30  Scraping without programming",
     "section": "",
     "text": "“application programming interface”↩︎\nIn R, our style was to name columns in lower case with words separated by underscores. In Javascript, the custom is usually called “camel case”, with words smushed together and the first letter of each upper cased. It’s just a custom, not a rule.↩︎\nI’m hiding names of people to the extent possible, and won’t list them in text here - they’ll only be in the images. Instead, I’ll show pictures of how to find the json when a name is necessary. Although this book is probably not indexed by Google, it’s possible that it could be some day, and I don’t want their names to show up in a Google search.↩︎"
   },
   {
     "objectID": "advanced-scrape2.html#understanding-a-web-page-and-its-structure",
     "href": "advanced-scrape2.html#understanding-a-web-page-and-its-structure",
-    "title": "30  Introduction to scraping in R",
-    "section": "30.1 Understanding a web page and its structure",
-    "text": "30.1 Understanding a web page and its structure\nWeb pages are written in HTML, even if they don’t have “html” at the end of the file name.\nHTML is like an upside-down tree. It has a trunk, which is an &lt;html&gt; tag, then two main branches: &lt;head&gt; and &lt;body&gt;. The content of the page branches out from the body tag:\n\n\n\nDatacamp html tree\n\n\nAll of HTML is just text. The tags tell your browser how to render each element, while attributes give them extra information, like the URL of a link, or a formatting class.\nWe can navigate the tree using RVest.\n\n\nOpen or create a project in RStudio, and create a new document. You can do this as a new markdown document, or as an R Script.\nThe code to load two libraries:\n\n\n#| label: setup-scrape2\n#| message: false \n#| warning: false\n#| echo: true\n\nlibrary(tidyverse)\nlibrary(rvest) \n\n\nAnd run your setup chunk.\n\n\nThe library rvest splits up the tree into its distinct elements, retaining the structure of the tree. The read_html() function takes a chunk of text, a page saved on your computer, or a page on the internet and parses it into its pieces.\nWe’re going to parse the page using “css selectors”, which tells the program how to navigate the page. The css selector can use the tag , an attribute, or both to find elements on the page. In this case, there is only one table, so we can just find one element using the “table” tag.\nThis method of scraping doesn’t work if the page was created on the fly by executing a Javascript program on your browser, the way that the simple page in the last chapter did. Those pages usually have a json dataset that you can grab more easily.\nHere’s what the page looks like when rendered, with the full tree shown on the right.\n\n\n\nfull page\n\n\nThis code parses the simple page at the address shown into its pieces, and save the result as my_htmlYou may notice that I’ve broken up the code to do one thing at a time. First, it saves the address in a variable called “url”. Then it uses the same piping we used in data work.1\n\n# label: readpres\n# echo: true\n# eval: false\n\nurl &lt;- \"https://cronkitedata.s3.amazonaws.com/docs/presidents.html\"\n\nmy_html &lt;-\n  url |&gt;  \n  read_html()\n\n\nprint(my_html)\n\n{html_document}\n&lt;html&gt;\n[1] &lt;head&gt;\\n&lt;meta http-equiv=\"Content-Type\" content=\"text/html; charset=UTF-8 ...\n[2] &lt;body&gt;\\n   &lt;h1&gt;A heading&lt;/h1&gt;\\n   &lt;p class=\"intro\"&gt; This is  a paragraph  ...\n\n\nThis is a complex object, and you’re only seeing the beginning of it – the two elements that are at the top of the tree. You’ll notice that there is a new object that is a data type called a list rather than a data frame in you environment. Lists are used to store complicated structures that don’t fit neatly into rectangle.\nTo find any element, like the table, use its tag in an html_node() . To find all of the elements of a type, make it plural, like html_nodes().2\nUse the tag name or CSS selector to get just a piece of the page.\n\nmy_html |&gt;\n html_element(\"body\") \n\n{html_node}\n&lt;body&gt;\n[1] &lt;h1&gt;A heading&lt;/h1&gt;\n[2] &lt;p class=\"intro\"&gt; This is  a paragraph with italic&lt;/p&gt;\n[3] &lt;div&gt;\\n     &lt;p&gt; This is a simple table&lt;/p&gt;\\n     &lt;table&gt;\\n&lt;thead&gt;&lt;tr&gt;\\n&lt;t ...\n\n\nTo get all of the paragraphs, make the command plural. Note how you now get the HTML of the selected elements in their entirety.\n\nmy_html |&gt;\n  html_elements (\"p\")\n\n{xml_nodeset (2)}\n[1] &lt;p class=\"intro\"&gt; This is  a paragraph with italic&lt;/p&gt;\n[2] &lt;p&gt; This is a simple table&lt;/p&gt;\n\n\nAnd to get everything with a class of “intro”, use a period to indicate a class, and convert it to text using the html_text function, asking R to remove extra whitespace with the “trim” argument.\n\nmy_html |&gt;\n  html_elements (\".intro\") |&gt;\n  html_text (trim=T)\n\n[1] \"This is  a paragraph with italic\"\n\n\n\nA special type: table\nTables are so commonly scraped that rvest has special way to extract the values, just as we did in Google Sheets, which puts it right into a data frame:\n\nmy_html |&gt;\n  html_element (\"table\")  |&gt;\n  html_table ()\n\n\n\n  \n\n\n\nThe singular version of html_element() picks out the first piece that matches the selector. The plural version would result in a list of all of them, from which you can select the number you want using the odd syntax .[[n]], where “n” is the table number.\n\n\nA harder example: Ballotpedia\nHere’s an example using the ballotpedia page we used in the last section:\n\nurl &lt;-\"https://ballotpedia.org/List_of_current_city_council_officials_of_the_top_100_cities_in_the_United_States\"\n\nballotpedia &lt;- \n  url |&gt;\n  read_html() |&gt;\n  html_elements(\"table\")\n\nprint (ballotpedia)\n\n{xml_nodeset (12)}\n [1] &lt;table class=\"infobox\" style=\"text-align: center; width:200px\"&gt;&lt;tbody&gt;\\n ...\n [2] &lt;table class=\"bptable sortable collapsible collapsed\" style=\"background: ...\n [3] &lt;table class=\"bptable gray sortable\" id=\"officeholder-table\" style=\"widt ...\n [4] &lt;table class=\"wikitable;\" style=\"width=100%\"&gt;&lt;tbody&gt;\\n&lt;tr&gt;\\n&lt;th colspan= ...\n [5] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n [6] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n [7] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n [8] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n [9] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n[10] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n[11] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n[12] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n\n\nLooking at this, we have several ways to get at the proper table. We can pick the third element that we just got:\n\nballotpedia[[3]]\n\n{html_node}\n&lt;table class=\"bptable gray sortable\" id=\"officeholder-table\" style=\"width:auto; border-bottom:1px solid #bcbcbc;\"&gt;\n[1] &lt;thead&gt;&lt;tr colspan=\"5\" style=\"background:#4c4c4c!important;color:#fff!imp ...\n[2] &lt;tbody&gt;\\n&lt;tr&gt;\\n&lt;td style=\"padding-left:10px;\"&gt;&lt;a href=\"https://ballotpedi ...\n\n\nor, you might notice that it has an “id” attribute called ‘officeholder-table’.\n\noffice_holders &lt;-\n  url |&gt;\n  read_html() |&gt;\n  html_node(\"#officeholder-table &gt; tbody\")\n\nIf you print it, you’ll see something like this:\n\n\n\noffice image\n\n\n\n\nGetting the link\nBefore, in Google Sheets, we had no way to pick up the list of links that would tell us what city each member was in. That’s also true when we use the html_table() function to turn it into a data frame:\n\noffice_table &lt;- \n  url |&gt;\n  read_html() |&gt;\n  html_node(\"#officeholder-table\") |&gt;  # keep the whole table to get headings \n  html_table()\n\nBut now we can get a list of the names of cities by extracting an “attribute” from the tag. (This is a little harder than I’d intended because not every row has a link, meaning we have to rejigger the formula to create empty rows when the link doesn’t exist.)\n\ncity_links &lt;-\n  ballotpedia[[3]] |&gt;  # the third table in our list\n  html_nodes (\"tbody &gt; tr\") |&gt;  # all rows\n  html_node (\"td &gt; a\") |&gt;  # justlink tag in the first column \n  html_attr(\"href\")  # the URL\n  \n\ntail(city_links)\n\n[1] \"https://ballotpedia.org/Barbara_Hanes_Burke\"         \n[2] \"https://ballotpedia.org/Jeff_MacIntosh\"              \n[3] \"https://ballotpedia.org/John_Larson_(North_Carolina)\"\n[4] \"https://ballotpedia.org/James_Taylor,_Jr.\"           \n[5] \"https://ballotpedia.org/Kevin_Mundy\"                 \n[6] \"https://ballotpedia.org/Robert_C._Clark\"             \n\n\nWe’ve never done this before, but we can add this list as a column to the data frame using the tidyverse add_column() function of the tidyverse . At the same time, you can put your regular expression muscles to work by “extracting” rather than “detecting” a pattern3:\n\noffice_table |&gt;\n  add_column (city=city_links, .before=\"Office\") |&gt;\n  mutate (city_extracted = str_extract(city, \"ballotpedia.org\\\\/(.*)$\", group=1) , .after=\"city\") |&gt;\n  sample_n(20)\n\n\n\n  \n\n\n\n(The str_extract () function matches everything after the last slash , because we put parentheses around the last part, which is a a capture group. This capability used to be very difficut in R – now it’s a bit easier. )\nIt doesn’t work perfectly, but it generally gets us a bit closer to a city name."
+    "title": "31  Introduction to scraping in R",
+    "section": "31.1 Understanding a web page and its structure",
+    "text": "31.1 Understanding a web page and its structure\nWeb pages are written in HTML, even if they don’t have “html” at the end of the file name.\nHTML is like an upside-down tree. It has a trunk, which is an &lt;html&gt; tag, then two main branches: &lt;head&gt; and &lt;body&gt;. The content of the page branches out from the body tag:\n\n\n\nDatacamp html tree\n\n\nAll of HTML is just text. The tags tell your browser how to render each element, while attributes give them extra information, like the URL of a link, or a formatting class.\nWe can navigate the tree using RVest.\n\n\nOpen or create a project in RStudio, and create a new document. You can do this as a new markdown document, or as an R Script.\nThe code to load two libraries:\n\n\n#| label: setup-scrape2\n#| message: false \n#| warning: false\n#| echo: true\n\nlibrary(tidyverse)\nlibrary(rvest) \n\n\nAnd run your setup chunk.\n\n\nThe library rvest splits up the tree into its distinct elements, retaining the structure of the tree. The read_html() function takes a chunk of text, a page saved on your computer, or a page on the internet and parses it into its pieces.\nWe’re going to parse the page using “css selectors”, which tells the program how to navigate the page. The css selector can use the tag , an attribute, or both to find elements on the page. In this case, there is only one table, so we can just find one element using the “table” tag.\nThis method of scraping doesn’t work if the page was created on the fly by executing a Javascript program on your browser, the way that the simple page in the last chapter did. Those pages usually have a json dataset that you can grab more easily.\nHere’s what the page looks like when rendered, with the full tree shown on the right.\n\n\n\nfull page\n\n\nThis code parses the simple page at the address shown into its pieces, and save the result as my_htmlYou may notice that I’ve broken up the code to do one thing at a time. First, it saves the address in a variable called “url”. Then it uses the same piping we used in data work.1\n\n# label: readpres\n# echo: true\n# eval: false\n\nurl &lt;- \"https://cronkitedata.s3.amazonaws.com/docs/presidents.html\"\n\nmy_html &lt;-\n  url |&gt;  \n  read_html()\n\n\nprint(my_html)\n\n{html_document}\n&lt;html&gt;\n[1] &lt;head&gt;\\n&lt;meta http-equiv=\"Content-Type\" content=\"text/html; charset=UTF-8 ...\n[2] &lt;body&gt;\\n   &lt;h1&gt;A heading&lt;/h1&gt;\\n   &lt;p class=\"intro\"&gt; This is  a paragraph  ...\n\n\nThis is a complex object, and you’re only seeing the beginning of it – the two elements that are at the top of the tree. You’ll notice that there is a new object that is a data type called a list rather than a data frame in you environment. Lists are used to store complicated structures that don’t fit neatly into rectangle.\nTo find any element, like the table, use its tag in an html_node() . To find all of the elements of a type, make it plural, like html_nodes().2\nUse the tag name or CSS selector to get just a piece of the page.\n\nmy_html |&gt;\n html_element(\"body\") \n\n{html_node}\n&lt;body&gt;\n[1] &lt;h1&gt;A heading&lt;/h1&gt;\n[2] &lt;p class=\"intro\"&gt; This is  a paragraph with italic&lt;/p&gt;\n[3] &lt;div&gt;\\n     &lt;p&gt; This is a simple table&lt;/p&gt;\\n     &lt;table&gt;\\n&lt;thead&gt;&lt;tr&gt;\\n&lt;t ...\n\n\nTo get all of the paragraphs, make the command plural. Note how you now get the HTML of the selected elements in their entirety.\n\nmy_html |&gt;\n  html_elements (\"p\")\n\n{xml_nodeset (2)}\n[1] &lt;p class=\"intro\"&gt; This is  a paragraph with italic&lt;/p&gt;\n[2] &lt;p&gt; This is a simple table&lt;/p&gt;\n\n\nAnd to get everything with a class of “intro”, use a period to indicate a class, and convert it to text using the html_text function, asking R to remove extra whitespace with the “trim” argument.\n\nmy_html |&gt;\n  html_elements (\".intro\") |&gt;\n  html_text (trim=T)\n\n[1] \"This is  a paragraph with italic\"\n\n\n\nA special type: table\nTables are so commonly scraped that rvest has special way to extract the values, just as we did in Google Sheets, which puts it right into a data frame:\n\nmy_html |&gt;\n  html_element (\"table\")  |&gt;\n  html_table ()\n\n\n\n  \n\n\n\nThe singular version of html_element() picks out the first piece that matches the selector. The plural version would result in a list of all of them, from which you can select the number you want using the odd syntax .[[n]], where “n” is the table number.\n\n\nA harder example: Ballotpedia\nHere’s an example using the ballotpedia page we used in the last section:\n\nurl &lt;-\"https://ballotpedia.org/List_of_current_city_council_officials_of_the_top_100_cities_in_the_United_States\"\n\nballotpedia &lt;- \n  url |&gt;\n  read_html() |&gt;\n  html_elements(\"table\")\n\nprint (ballotpedia)\n\n{xml_nodeset (12)}\n [1] &lt;table class=\"infobox\" style=\"text-align: center; width:200px\"&gt;&lt;tbody&gt;\\n ...\n [2] &lt;table class=\"bptable sortable collapsible collapsed\" style=\"background: ...\n [3] &lt;table class=\"bptable gray sortable\" id=\"officeholder-table\" style=\"widt ...\n [4] &lt;table class=\"wikitable;\" style=\"width=100%\"&gt;&lt;tbody&gt;\\n&lt;tr&gt;\\n&lt;th colspan= ...\n [5] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n [6] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n [7] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n [8] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n [9] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n[10] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n[11] &lt;table class=\"navbox\" cellspacing=\"0\" style=\";\"&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td style=\"pa ...\n[12] &lt;table cellspacing=\"0\" class=\"nowraplinks collapsible autocollapse\" styl ...\n\n\nLooking at this, we have several ways to get at the proper table. We can pick the third element that we just got:\n\nballotpedia[[3]]\n\n{html_node}\n&lt;table class=\"bptable gray sortable\" id=\"officeholder-table\" style=\"width:auto; border-bottom:1px solid #bcbcbc;\"&gt;\n[1] &lt;thead&gt;&lt;tr colspan=\"5\" style=\"background:#4c4c4c!important;color:#fff!imp ...\n[2] &lt;tbody&gt;\\n&lt;tr&gt;\\n&lt;td style=\"padding-left:10px;\"&gt;&lt;a href=\"https://ballotpedi ...\n\n\nor, you might notice that it has an “id” attribute called ‘officeholder-table’.\n\noffice_holders &lt;-\n  url |&gt;\n  read_html() |&gt;\n  html_node(\"#officeholder-table &gt; tbody\")\n\nIf you print it, you’ll see something like this:\n\n\n\noffice image\n\n\n\n\nGetting the link\nBefore, in Google Sheets, we had no way to pick up the list of links that would tell us what city each member was in. That’s also true when we use the html_table() function to turn it into a data frame:\n\noffice_table &lt;- \n  url |&gt;\n  read_html() |&gt;\n  html_node(\"#officeholder-table\") |&gt;  # keep the whole table to get headings \n  html_table()\n\nBut now we can get a list of the names of cities by extracting an “attribute” from the tag. (This is a little harder than I’d intended because not every row has a link, meaning we have to rejigger the formula to create empty rows when the link doesn’t exist.)\n\ncity_links &lt;-\n  ballotpedia[[3]] |&gt;  # the third table in our list\n  html_nodes (\"tbody &gt; tr\") |&gt;  # all rows\n  html_node (\"td &gt; a\") |&gt;  # justlink tag in the first column \n  html_attr(\"href\")  # the URL\n  \n\ntail(city_links)\n\n[1] \"https://ballotpedia.org/Barbara_Hanes_Burke\"         \n[2] \"https://ballotpedia.org/Jeff_MacIntosh\"              \n[3] \"https://ballotpedia.org/John_Larson_(North_Carolina)\"\n[4] \"https://ballotpedia.org/James_Taylor,_Jr.\"           \n[5] \"https://ballotpedia.org/Kevin_Mundy\"                 \n[6] \"https://ballotpedia.org/Robert_C._Clark\"             \n\n\nWe’ve never done this before, but we can add this list as a column to the data frame using the tidyverse add_column() function of the tidyverse . At the same time, you can put your regular expression muscles to work by “extracting” rather than “detecting” a pattern3:\n\noffice_table |&gt;\n  add_column (city=city_links, .before=\"Office\") |&gt;\n  mutate (city_extracted = str_extract(city, \"ballotpedia.org\\\\/(.*)$\", group=1) , .after=\"city\") |&gt;\n  sample_n(20)\n\n\n\n  \n\n\n\n(The str_extract () function matches everything after the last slash , because we put parentheses around the last part, which is a a capture group. This capability used to be very difficut in R – now it’s a bit easier. )\nIt doesn’t work perfectly, but it generally gets us a bit closer to a city name."
   },
   {
     "objectID": "advanced-scrape2.html#your-turn",
     "href": "advanced-scrape2.html#your-turn",
-    "title": "30  Introduction to scraping in R",
-    "section": "30.2 Your turn",
-    "text": "30.2 Your turn\nA lot of people use IMDB pages as practice for scraping because its HTML is a little primitive. Try extracting the name, year, rating, and rank of each item in this list https://www.imdb.com/chart/toptv/?ref_=nv_tvv_250\nHints:\n\nIf you right-click on the table, you’ll see the table has an attribute of class=\"chart full-width\" . That means you can use the class selector .chart &gt; table . If you use plural, it will be a list with one item in it. If you use singular, it will be the table itself.\nTo get three columns in a data frame of text, html_table(trim=T)\nThe year is held in a span element with a class of “secondaryInfo” in the first column. See if you can figure out how to get at it. 4\nTo extract title and its link, use the a tag\nTo get the full information from the rating column, including the number of votes that it’s based on, use the strong tag then the title attribute. See if you can figure that one out.\nTo put them all together, use the add_column() verb"
+    "title": "31  Introduction to scraping in R",
+    "section": "31.2 Your turn",
+    "text": "31.2 Your turn\nA lot of people use IMDB pages as practice for scraping because its HTML is a little primitive. Try extracting the name, year, rating, and rank of each item in this list https://www.imdb.com/chart/toptv/?ref_=nv_tvv_250\nHints:\n\nIf you right-click on the table, you’ll see the table has an attribute of class=\"chart full-width\" . That means you can use the class selector .chart &gt; table . If you use plural, it will be a list with one item in it. If you use singular, it will be the table itself.\nTo get three columns in a data frame of text, html_table(trim=T)\nThe year is held in a span element with a class of “secondaryInfo” in the first column. See if you can figure out how to get at it. 4\nTo extract title and its link, use the a tag\nTo get the full information from the rating column, including the number of votes that it’s based on, use the strong tag then the title attribute. See if you can figure that one out.\nTo put them all together, use the add_column() verb"
   },
   {
     "objectID": "advanced-scrape2.html#cheat-sheet",
     "href": "advanced-scrape2.html#cheat-sheet",
-    "title": "30  Introduction to scraping in R",
-    "section": "30.3 Cheat sheet",
-    "text": "30.3 Cheat sheet\n\nHTML tags / elements\nEach tag has an opening and closing it. This is what a “p” open and close tag looks like, with the text inside it shown on the page:\n&lt;p&gt; This is a paragraph &lt;/p&gt;\n\nHere is a list of common tags you’ll use in scraping:\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\np\nor a paragraph\n\n\ndiv\na block of text , or division of the page.\n\n\nspan\nAn area within a div or p element that is treated specially without it breaking into a new line.\n\n\na\na link. It should always have an attribute of href, which is the URL to the link.\n\n\nh1 through h6\nwhich are headline levels. “h1” is the headline, “h2” is a sub-head, and so on.\n\n\n\nModern websites might have these sections, which are used instead of the “div” tag:\n\n\n\ntag\ndescription\n\n\n\n\nnav\na navigational menu\n\n\nmain\nthe main block of the page with the content\n\n\naside\na sidebar\n\n\nfooter\nthe stuff at the bottom.\n\n\n\nTables are structured like this:\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\ntable\nthe main container.\n\n\nthead and tbody\nA heading area and the body area. These are always just below the table element\n\n\nth\nthe row that contains the headings . These are the first row within the body\n\n\ntr\nall of the content rows. These are subsequent rows within tbody\n\n\ntd\nall of the cells (columns). These are always WITHIN a tr or th element.\n\n\n\nAny of these can be nested within any others. Typically, a page starts with an “h1” tag, then has “div” tags for different sections, such as the sidebar or the main content. An “a” tag is typically nested within others.\nStandalone tags:\nA few tags don’t have opening and closing versions - they just stand alone:\nimg - an image to show. It would have a src attribute for the link, and an alt attribute for text to show for accessibility. example: &lt;img src=\"path-to-my-image\" alt=\"This is a picture of...\"&gt;\nbr - A hard line break.\n\n\nCommon attributes for tags\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\nhref\nthe URL of a link within an a tag.\n\n\nsrc\nthe path to an image, within an img tag.\n\n\nclass\na reference to a CSS class. More than one class can be identified, separated with a space.\n\n\nid\na unique name for this element using CSS\n\n\n\nPeople can also make up their own attributes - they’re arbitrary.\n\n\nRvest functions\n\n\n\n\n\n\n\nfunction\ndescription\n\n\n\n\nread_html()\nto parse the page. Start with a file name or URL.\n\n\nhtml_elements\nto get ALL elements that match your query. It always gives back a list of objects, even if it’s empty. To pluck one by number, use [[n]]. You might see it as html_nodes(), which is from an older version of the rvest library\n\n\nhtml_element\nthe FIRST element that matches your query. Always returns a single object. (You might see html_node())\n\n\nhtml_table()\nconvert a table to a data frame with just its text\n\n\nhtml_text(trim=T)\nget the text within an element.\n\n\nhtml_attr (attr_name)\nget the value of an attribute. Commonly, this is used as httr_attr(\"href\") to get the link inside an a link element.\n\n\nadd_column\nappend columns to the end of a dataframe from lists/vectors. They must be in the same order, and there have to be the same number of items as there are rows.\n\n\nadd_row\nto append rows at the bottom. These can be by name or position. There can’t be any columns in the row you want to add that aren’t in the one you’re adding to.\n\n\n\n(You’ll often see these operations as cbind and rbind - they’re similar. Our way is just the tidyverse way.)\n\n\nExamples of common CSS selectors\nThis uses an example assuming the tag “p” , class “myclass” and id “myid” are used. You substitute the tags, classes, and id’s you want. See https://www.scraperapi.com/blog/css-selectors-cheat-sheet/ for a more in-depth cheat sheet.\n\n\n\n\n\n\n\nselector\ndescription\n\n\n\n\np\na “p” element. Replace with the element you want to capture.\n\n\n.myclass\nany element with class=“myclass”\n\n\np .myclass\n“p” element with a class of “myclass”.\n\n\np &gt; .myclass\nevery child element of p with a class of ‘myclass’ regardless of the tag. Must be a direct child.\n\n\n#myid\nAny element with “id=‘myid’”\n\n\nbody &gt; div &gt; table .content-table &gt; tbody &gt; tr\nA row within a table classed “content-table” within a div.\n\n\n\nYou have to go through the whole path to an element if you need it, which is why you have to look in the inspector section of your browser or use the CSS Selector Gadget (a chrome extension that I’ve never been able to work properly!)"
+    "title": "31  Introduction to scraping in R",
+    "section": "31.3 Cheat sheet",
+    "text": "31.3 Cheat sheet\n\nHTML tags / elements\nEach tag has an opening and closing it. This is what a “p” open and close tag looks like, with the text inside it shown on the page:\n&lt;p&gt; This is a paragraph &lt;/p&gt;\n\nHere is a list of common tags you’ll use in scraping:\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\np\nor a paragraph\n\n\ndiv\na block of text , or division of the page.\n\n\nspan\nAn area within a div or p element that is treated specially without it breaking into a new line.\n\n\na\na link. It should always have an attribute of href, which is the URL to the link.\n\n\nh1 through h6\nwhich are headline levels. “h1” is the headline, “h2” is a sub-head, and so on.\n\n\n\nModern websites might have these sections, which are used instead of the “div” tag:\n\n\n\ntag\ndescription\n\n\n\n\nnav\na navigational menu\n\n\nmain\nthe main block of the page with the content\n\n\naside\na sidebar\n\n\nfooter\nthe stuff at the bottom.\n\n\n\nTables are structured like this:\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\ntable\nthe main container.\n\n\nthead and tbody\nA heading area and the body area. These are always just below the table element\n\n\nth\nthe row that contains the headings . These are the first row within the body\n\n\ntr\nall of the content rows. These are subsequent rows within tbody\n\n\ntd\nall of the cells (columns). These are always WITHIN a tr or th element.\n\n\n\nAny of these can be nested within any others. Typically, a page starts with an “h1” tag, then has “div” tags for different sections, such as the sidebar or the main content. An “a” tag is typically nested within others.\nStandalone tags:\nA few tags don’t have opening and closing versions - they just stand alone:\nimg - an image to show. It would have a src attribute for the link, and an alt attribute for text to show for accessibility. example: &lt;img src=\"path-to-my-image\" alt=\"This is a picture of...\"&gt;\nbr - A hard line break.\n\n\nCommon attributes for tags\n\n\n\n\n\n\n\ntag\ndescription\n\n\n\n\nhref\nthe URL of a link within an a tag.\n\n\nsrc\nthe path to an image, within an img tag.\n\n\nclass\na reference to a CSS class. More than one class can be identified, separated with a space.\n\n\nid\na unique name for this element using CSS\n\n\n\nPeople can also make up their own attributes - they’re arbitrary.\n\n\nRvest functions\n\n\n\n\n\n\n\nfunction\ndescription\n\n\n\n\nread_html()\nto parse the page. Start with a file name or URL.\n\n\nhtml_elements\nto get ALL elements that match your query. It always gives back a list of objects, even if it’s empty. To pluck one by number, use [[n]]. You might see it as html_nodes(), which is from an older version of the rvest library\n\n\nhtml_element\nthe FIRST element that matches your query. Always returns a single object. (You might see html_node())\n\n\nhtml_table()\nconvert a table to a data frame with just its text\n\n\nhtml_text(trim=T)\nget the text within an element.\n\n\nhtml_attr (attr_name)\nget the value of an attribute. Commonly, this is used as httr_attr(\"href\") to get the link inside an a link element.\n\n\nadd_column\nappend columns to the end of a dataframe from lists/vectors. They must be in the same order, and there have to be the same number of items as there are rows.\n\n\nadd_row\nto append rows at the bottom. These can be by name or position. There can’t be any columns in the row you want to add that aren’t in the one you’re adding to.\n\n\n\n(You’ll often see these operations as cbind and rbind - they’re similar. Our way is just the tidyverse way.)\n\n\nExamples of common CSS selectors\nThis uses an example assuming the tag “p” , class “myclass” and id “myid” are used. You substitute the tags, classes, and id’s you want. See https://www.scraperapi.com/blog/css-selectors-cheat-sheet/ for a more in-depth cheat sheet.\n\n\n\n\n\n\n\nselector\ndescription\n\n\n\n\np\na “p” element. Replace with the element you want to capture.\n\n\n.myclass\nany element with class=“myclass”\n\n\np .myclass\n“p” element with a class of “myclass”.\n\n\np &gt; .myclass\nevery child element of p with a class of ‘myclass’ regardless of the tag. Must be a direct child.\n\n\n#myid\nAny element with “id=‘myid’”\n\n\nbody &gt; div &gt; table .content-table &gt; tbody &gt; tr\nA row within a table classed “content-table” within a div.\n\n\n\nYou have to go through the whole path to an element if you need it, which is why you have to look in the inspector section of your browser or use the CSS Selector Gadget (a chrome extension that I’ve never been able to work properly!)"
   },
   {
     "objectID": "advanced-scrape2.html#going-further",
     "href": "advanced-scrape2.html#going-further",
-    "title": "30  Introduction to scraping in R",
-    "section": "30.4 Going further",
-    "text": "30.4 Going further\nThe CSS selectors shown in this chapter are a little limiting – you’ll find times when the information you’re seeking isn’t defined using those selectors. That’s when the XPATH selectors we saw in the last chapter are used. It’s beyond the scope of this tutorial, but ask for some help or try to find XPATH examples if you can’t figure out how to get to a part of the page with your css selections – it’s pretty common for this to happen."
+    "title": "31  Introduction to scraping in R",
+    "section": "31.4 Going further",
+    "text": "31.4 Going further\nThe CSS selectors shown in this chapter are a little limiting – you’ll find times when the information you’re seeking isn’t defined using those selectors. That’s when the XPATH selectors we saw in the last chapter are used. It’s beyond the scope of this tutorial, but ask for some help or try to find XPATH examples if you can’t figure out how to get to a part of the page with your css selections – it’s pretty common for this to happen."
   },
   {
     "objectID": "advanced-scrape2.html#footnotes",
     "href": "advanced-scrape2.html#footnotes",
-    "title": "30  Introduction to scraping in R",
+    "title": "31  Introduction to scraping in R",
     "section": "",
     "text": "We glossed over this before, but any time you use a pipe, whatever comes above a command is used as the first argument of the current command. So this code is the same as read_html(url). Sometimes you need it as something other than the first argument, in which case you reference it using a period.↩︎\nA newer version of rvest prefers the use of html_element() instead of node. For us, they mean the same thing and both work. The newer syntax throws a warning in your RStudio environment that I can’t troubleshoot, so I’m waiting for an update to the rvest package to switch.↩︎\nstr_extract changed in December 2022. Make sure your packages are up to date.↩︎\n html_elements (\"tbody &gt; tr &gt; .titleColumn &gt; .secondaryInfo\") |&gt; html_text(trim=T)↩︎"
   },
@@ -1244,96 +1300,5 @@
     "title": "Appendix C — Documentation for PPP data chapters",
     "section": "C.3 Other data for use in with PPP data",
     "text": "C.3 Other data for use in with PPP data\n\nThe NAICS code, a standardized code created by the Office of Management and Budget and used widely by the federal government, will be turned into words in the chapter on joining data.\nThe business Zip Code is the mailing address. I mapped them to actual physical address Zip Codes and downloaded some Census data to attach to each loan. This is rough estimate, and not a great representation of neighborhoods, but it is useful for us to get a sense of the demographics during the join part of the R training.\nReveal worked with https://geocod.io to add latitude and longitude to each loan, allowing them to match it to Census tracts instead of Zip Codes, a much better way to look at neighborhoods. We will look at that data in the mapping / geography portion of the course"
-  },
-  {
-    "objectID": "start-groupby.html#summarizing-with-groups",
-    "href": "start-groupby.html#summarizing-with-groups",
-    "title": "16  Summarizing detail into totals",
-    "section": "16.1 Summarizing with groups",
-    "text": "16.1 Summarizing with groups\nSummarizing a list of items in R is done using the verbs group_by and summarize. Think of grouping as answering the questions, “How many?” and “How much?”. They are particularly powerful when your question also has the words “the most” or the “the least” or “of each”. Some examples:\n\nWhich Zip Code had the most crimes?\nWhat month had the least total rainfall?\nHow much did each candidate raise last quarter?\nIn playing cards, how many of each suit do I have in my hand?\nOn average, are Cronkite students taller or shorter than in other schools?\n\n\nConfusing grouping with sorting or arranging\nMany reporters confuse this summarization with “sorting”. One reason is that this is how we express the concept in plain language: “I want to sort Skittles by color”.\nBut in data analysis, sorting and and grouping are very different things. Sorting, which is done in R via the arrange() verb, involves shuffling a data frame’s rows into some order based on the values in a column. Grouping is a way to aggregate and compute summary statistics such as a count (the number of items), sum (how much they add up to), or average for category. It means “make piles and compute statistics for each one.”\n\n\nWhen to use filter vs. summary\nSomething that trips up beginners is a desire to see details and totals at the same time, which is more difficult than it sounds.\nA filter is used to display your selected items as a list. You’ll get to see all of the detail and every column. As a convenience, R shows you how many items are in that filtered list (usually). That’s great when you want to just look at them, or get more information about them. For instance, if you had a list of crimes by ZIP code, you might just want to see the list in your neighborhood – where, exactly, were they? When did they happen? Was it at night or the morning? What crimes happened on which blocks?\nAggregates are used when you just want to see summaries – does my ZIP code have more crime than others? Are robberies more common than car theft in my Zip code, and how does that compare to others?\nIn practice, you’ll go back and forth between summary and detail. They’re both important, just different."
-  },
-  {
-    "objectID": "start-groupby.html#motivational-exercise",
-    "href": "start-groupby.html#motivational-exercise",
-    "title": "16  Summarizing detail into totals",
-    "section": "16.2 Motivational exercise",
-    "text": "16.2 Motivational exercise\nThis exercise helps you internalize the differences between filtering and aggregating (or grouping). It uses a list of fatal police shootings logged by the Washington Post, downloaded in late 2022. The column names are pretty self-explanatory, and we’ll only be looking at the ethnicity column. (Scroll to the right if you can’t see it.)\n\n\nCreate a Quarto document in your project\nAdd the usual setup chunk to load the tidyverse, using the library command. Copy and paste the chunk from a previous lesson if you need to.\nCopy and paste this into a code chunk to read the data.\n\n\n\nwaposhootings &lt;- \n  readRDS ( \n        url ( \n           \"https://cronkitedata.s3.amazonaws.com/rdata/waposhootings.RDS\"\n           )\n        )\n\nExamine the data:\n\nwaposhootings |&gt; \n  sample_n (50)\n\n\n\n  \n\n\n\n\nFiltering method\nUsing a pen and paper, write out something that looks like the table below. The total number comes from the number of rows in the data frame.\n\n\n\n\nEthnicity\n# of victims\n% of total\n\n\n\n\nBlack, non-Hispanic\n\n\n\n\nWhite, non-Hispanic\n\n\n\n\nHispanic\n\n\n\n\nTOTAL\n5,945\n100%\n\n\n\n\nCopy and paste this code to list all fatal shootings of Black, non-Hispanic victims. Look at the number of rows returned and enter it into your handmade table.\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"Black, non-Hispanic\")\n\n\nRun this code to get the number of White, non-Hispanic victims and write the answer down on your handmade table\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"White, non-Hispanic\")\n\n\nRun this filter to get the number of Hispanic victims, and write the answer down.\n\n\nwaposhootings |&gt; \n  filter ( ethnicity == \"Hispanic\")\n\n\nCompute the percentages\n\n\nCompute the percent of total each group represents. Do this on a calculator (or just type the formula into Google). Remember, the percent of total formula is: group / total * 100\nWrite the answers into your handwritten chart.\nNow, rewrite your chart so that the highest percentage is on the first row, the second highest on the second row, and the third highest on the third row.\nRepeat your calculations, since you always have to do every hand calculation twice. They won’t add up to 100% because we left out people who had an unknown ethnicity or other ethnicity."
-  },
-  {
-    "objectID": "start-groupby.html#aggregating-for-an-easier-way",
-    "href": "start-groupby.html#aggregating-for-an-easier-way",
-    "title": "16  Summarizing detail into totals",
-    "section": "16.3 Aggregating for an easier way",
-    "text": "16.3 Aggregating for an easier way\nThat’s a lot of work. Imagine if you had to do this 50 times for 50 states, or even more for every Zip code? There has to be an easier way, and there is: Grouping and summarizing.\nWe’ll build the code one piece at a time so you can see what’s happening:\n\nCopy this code, one chunk at a time, into your Quarto document and run them to see what the answers are.\n\n\nGet the totals\n\n\nwaposhootings |&gt; \n  summarize ( num_shootings = n() ) \n\n\nAdd a “group” to summarize by ethnicity\n\n\nwaposhootings |&gt; \n  group_by (ethnicity) |&gt; \n  summarize (num_shootings = n() )\n\n\nArrange to get the highest number at the top\n\n\nwaposhootings |&gt; \n  group_by ( ethnicity) |&gt; \n  summarize (num_shootings = n() ) |&gt; \n  arrange ( desc (num_shootings ))\n\n\nCompute the percent\nNotice that the “mutate” line is exactly the same formula you typed into your calculator. The only difference is that we computed the total number of shootings instead of typing it. \nThis is one of the most common patterns you’ll use in data reporting, and we can get much more sophisticated with it. Don’t expect to understand it fully right away, but you should recognize the idea: If you find yourself filtering in order to count something, consider using grouping instead.\n\n\nwaposhootings |&gt; \n  group_by ( ethnicity) |&gt; \n  summarize ( num_shootings = n() ) |&gt; \n  mutate ( total = sum ( num_shootings ) , \n           percent = num_shootings /total * 100 ) |&gt; \n  arrange ( desc ( num_shootings))"
-  },
-  {
-    "objectID": "start-groupby.html#differences",
-    "href": "start-groupby.html#differences",
-    "title": "16  Summarizing detail into totals",
-    "section": "16.4 Differences",
-    "text": "16.4 Differences\nWhen you filtered, you could see the details of every row. That’s great if you want to explore it or find examples. But it’s not great for that sentence you need when you want to say which group is the largest or smallest, or compare the numbers across groups.\nWhen you grouped, you got statistics by ethnicity, but you couldn’t see the details.\nTypically, you’ll go back and forth between aggregating and filtering, using each method to get to different goals. Don’t fall in the trap of trying to force all of your questions into a filter, or all of your questions into aggregation. Instead, think about which method will give you the answers you want most efficiently."
-  },
-  {
-    "objectID": "start-groupby.html#your-questions",
-    "href": "start-groupby.html#your-questions",
-    "title": "16  Summarizing detail into totals",
-    "section": "16.5 Your questions:",
-    "text": "16.5 Your questions:\n\nOn your Quarto document, try to write out, in plain English, what you just did. See if you can articulate what the difference between filtering and aggregating is, in your own words."
-  },
-  {
-    "objectID": "r-verb-filter.html#footnotes",
-    "href": "r-verb-filter.html#footnotes",
-    "title": "14  Verbs in depth: Select, arrange, filter",
-    "section": "",
-    "text": "That’s why I’ve removed all of the punctuation from your the borrower names and addresses in your data file, and turned all of the names into upper case.↩︎"
-  },
-  {
-    "objectID": "r-verb-mutate.html#math-on-columns",
-    "href": "r-verb-mutate.html#math-on-columns",
-    "title": "15  Verbs in depth: New from old data with Mutate",
-    "section": "15.2 Math on columns",
-    "text": "15.2 Math on columns\nHere’s an example of computing the portion of the original amount that was forgiven.\nUse the single “=” sign to provide a name for the new column and create more than one new column using a comma between them:\n\nppp_orig |&gt;\n1    select ( borrower_name, borrower_city, amount, forgiveness_amount) |&gt;\n    mutate ( forgiven_pct = \n2               ( forgiveness_amount  /  amount )  * 100\n            )   |&gt;                   \n3  sample_n (50)\n\n\n1\n\nPick out a few columns to work with\n\n2\n\nCompute a new column using math\n\n3\n\nPrint out a random sample of 50 rows\n\n\n\n\n\n\n\n\n\n\nNote\n\n\n\nA value over 100 means that the government also waived fees and interest; a value under 100 means that less than all of the loan was forgiven.\n\n\n\n\n\n\n\n\n\n(I did some fancy formatting so you could read the results more clearly. We’ll come back to that in a later lesson.)\n\nConverting NA to 0\nIf you look through the pages, you’ll realize that there are a lot of rows with NA instead of a value. That’s because whenever you do anything with a missing value, the result is missing. Missing values infect everything they touch. You often have to deal with them before you can do anything else.\nWe would like to convert the forgiven amount from a missing value to zero, under the idea that if they have not filled it out, nothing has (yet) been forgiven. Of course, we’d have to check that with the SBA before publication.\nThere is a specific function used for that: replace_na(), with two arguments: The column you want to check for missing values, and the value you want to use instead. This isn’t limited to numbers – you can do the same thing with words, often replacing NA with “Unknown”.\n\nppp_forgiven_fixed &lt;- \n  ppp_orig |&gt;\n  mutate (amount_forgiven = replace_na(forgiveness_amount, 0))\n\n(Note that nothing came out in this code chunk because the result was saved into a new data frame)\n\n\nDetail and total with summary statistics\nYou can use mutate to put summary statistics next to your values so you can see whether or not they are similar to the average or median.1\n\nppp_forgiven_fixed |&gt; \n  mutate ( avg_forgiven = mean(amount_forgiven), \n           median_forgiven = median(amount_forgiven)) |&gt; \n  select (borrower_name, amount_forgiven, avg_forgiven, median_forgiven ) |&gt; \n  head()"
-  },
-  {
-    "objectID": "r-verb-mutate.html#creating-two-categories-using-if_else",
-    "href": "r-verb-mutate.html#creating-two-categories-using-if_else",
-    "title": "15  Verbs in depth: New from old data with Mutate",
-    "section": "15.3 Creating two categories using if_else()",
-    "text": "15.3 Creating two categories using if_else()\nVery often, you’ll want to categorize entries in a database in order to make it simpler to count and sum the values in a meaningful way. For example, the business_type column has 24 different values, including NA . Here’s a list with the number of loans in each category:\n\n\n\n\n  \n\n\n\nOne way to work with these is to create new columns with yes-no indicators for certain types of businesses like non-profits or individuals vs. companies.\nThe function to do this is if_else() , which tests a condition exactly the same way filter did, but then assigns a value based on whether it’s met or not. You list the condition first, then a comma , then what should happen if the condition is true, and then what should happen if it’s false.\nThe condition is done the same way you did it in a filter, but instead of picking out the rows that match, mutate acts on them separately from the rows that don’t match.\nThere’s no “maybe”, except for NA’s in the original value. Here’s the general form of what it looks like:\n\nnew_column_name = if_else ( test the old column for something as in a a filter,\n                         give it a value if it's true,\n                         give it another value if it's not true)\nSo here is a way to do this with the business_type using the same %in% operator you used in the filter lesson, saving it to new data frame in your Environment, then displaying the first 10 types of businesses using count()\n\nppp_category_indiv &lt;- \n  ppp_forgiven_fixed |&gt;\n  mutate ( is_individual = \n              if_else ( business_type %in% \n                          c(\"Independent Contractors\", \n                            \"Sole Proprietorship\", \n                            \"Self-Employed Individuals\", \n                            \"Single Member LLC\"), \n                 \"Individual\", \n                 \"Organization\")\n  )  \n\nNow print off a sampling of a some of the rows to take a look:\n\nppp_category_indiv |&gt; \n  select ( is_individual, business_type) |&gt; \n  sample_n (50) \n\n\n\n  \n\n\n\n\nTwo categories using fuzzy matching\nYou can also use the same str_like() function you used in filtering. Here, it sets whether or not the borrower was a non-profit. Remember, a % means “anything or nothing, and a _ means”any single character but it must be present”. In our case, the word is shown as “Non Profit” and “Non-Profit”. Once. you’ve set it to your new words, you can just filter on the new value:\n\nppp_category_nonprofit &lt;-\n  ppp_forgiven_fixed |&gt; \n  mutate ( is_nonprofit = \n             if_else ( str_like(business_type, \"%Non_Profit%\") , \n                       \"Is nonprofit\", \n                       \"Not nonprofit\"))  \n\nYou can check it using a new verb, count(), to check the number of items by category:\n\nppp_category_nonprofit |&gt; \n  filter ( is_nonprofit == \"Is nonprofit\") |&gt; \n  count ( is_nonprofit, business_type)\n\n\n\n  \n\n\n\n(The profit categorization is unclear for some of the original types, such as professional associations , tribal concerns and cooperatives.)"
-  },
-  {
-    "objectID": "r-verb-mutate.html#more-than-two-categories-using-case_when",
-    "href": "r-verb-mutate.html#more-than-two-categories-using-case_when",
-    "title": "15  Verbs in depth: New from old data with Mutate",
-    "section": "15.4 More than two categories using case_when()",
-    "text": "15.4 More than two categories using case_when()\nSometimes you will want more than one outcome, such as setting a value for “High”, “Medium” and “Low”. Instead of if_then, use the function case_when, which lets you string along conditions and their outcomes. The tilde (~) is used to show what happens if it’s true. At the end, you can add an argument .default=\"Other\" to say “everything else”.\n\noriginal data |&gt; \n  case_when ( first condition ~ what if it's true,\n              second condition ~ what if  it's true, \n              third condition  ~ what if it's true, \n              .default =  what to do with everything that's left\n              )"
-  },
-  {
-    "objectID": "r-verb-mutate.html#creating-categories-for-easier-analysis",
-    "href": "r-verb-mutate.html#creating-categories-for-easier-analysis",
-    "title": "15  Verbs in depth: New from old data with Mutate",
-    "section": "15.3 Creating categories for easier analysis",
-    "text": "15.3 Creating categories for easier analysis\nVery often, you’ll want to categorize entries in a database in order to make it simpler to count and sum the values in a meaningful way. For example, the business_type column has 24 different values, including NA . Here’s a list with the number of loans in each category:\n\n\n\n\n  \n\n\n\n\nYes-No categories using if_else()\nOne way to work with these is to create new columns with yes-no indicators for certain types of businesses like non-profits or individuals vs. companies.\nThe function to do this is if_else() , which tests a condition exactly the same way filter did, but then assigns a value based on whether it’s met or not. You list the condition first, then a comma , then what should happen if the condition is true, and then what should happen if it’s false.\nThe condition is done the same way you did it in a filter, but instead of picking out the rows that match, mutate acts on them separately from the rows that don’t match.\nThere’s no “maybe”, except for NA’s in the original value. Here’s the general form of what it looks like:\n\nnew_column_name = if_else ( test the old column for something as in a a filter,\n                         give it a value if it's true,\n                         give it another value if it's not true)\nSo here is a way to do this with the business_type using the same %in% operator you used in the filter lesson, saving it to new data frame in your Environment, then displaying the first 10 types of businesses using count()\n\nExample 1: Two categories using exact conditions\n\nppp_category_indiv &lt;- \n  ppp_forgiven_fixed |&gt;\n  mutate ( is_individual = \n              if_else ( business_type %in% \n                          c(\"Independent Contractors\", \n                            \"Sole Proprietorship\", \n                            \"Self-Employed Individuals\", \n                            \"Single Member LLC\"), \n                 \"Individual\", \n                 \"Organization\")\n  )  \n\nNow print off a sampling of a some of the rows to take a look:\n\nppp_category_indiv |&gt; \n  select ( is_individual, business_type) |&gt; \n  sample_n (50) \n\n\n\n  \n\n\n\n\n\nExample 2: Two categories using wildcards\nYou can also use the same str_like() function you used in filtering. Here, it sets whether or not the borrower was a non-profit. Remember, a % means “anything or nothing, and a _ means”any single character but it must be present”. In our case, the word is shown as “Non Profit” and “Non-Profit”. Once. you’ve set it to your new words, you can just filter on the new value:\n\nppp_category_nonprofit &lt;-\n  ppp_forgiven_fixed |&gt; \n  mutate ( is_nonprofit = \n             if_else ( str_like(business_type, \"%Non_Profit%\") , \n                       \"Is nonprofit\", \n                       \"Not nonprofit\"))  \n\nYou can check it using a new verb, count(), to check the number of items by category:\n\nppp_category_nonprofit |&gt; \n  filter ( is_nonprofit == \"Is nonprofit\") |&gt; \n  count ( is_nonprofit, business_type)\n\n\n\n  \n\n\n\n(The profit categorization is unclear for some of the original types, such as professional associations , tribal concerns and cooperatives.)\n\n\n\nMore than two categories using case_when()\nSometimes you will want more than one outcome, such as setting a value for “High”, “Medium” and “Low”. Instead of if_then, use the function case_when, which lets you string along conditions and their outcomes. The tilde (~) is used to show what happens if it’s true. At the end, you can add an argument .default=\"Other\" to say “everything else”.\n\noriginal data |&gt; \n  case_when ( first condition ~ what if it's true,\n              second condition ~ what if  it's true, \n              third condition  ~ what if it's true, \n              .default =  what to do with everything that's left\n              )"
-  },
-  {
-    "objectID": "r-verb-groupby.html#summarize-for-statistics",
-    "href": "r-verb-groupby.html#summarize-for-statistics",
-    "title": "17  Verbs in depth: Aggregating with groups",
-    "section": "17.1 summarize for statistics",
-    "text": "17.1 summarize for statistics\nsummarize computes summary statistics such as the number of rows in a data frame or the sum of dollar values. It removes the original columns completely, and only produces the summary statistics you compute within that statement. Using summarize alone produces a data frame with one row. It’s the equivalent of putting nothing in your pivot table in Excel other than the “Values” area.\nAnother way to think of summarize is that it collapses your list of items (loans, in our example) into a statistical report.\n\nThe dreaded NA\nYou saw in the mutate section that missing values are always a problem. Because they’re unknown, they can’t match anything else, they can’t be considered 0, and they can warp any answers you get. But there’s usually nothing you can do about missing data, so you have to tell the program exactly what to do about them.\nThere are two choices:\n\nLet them infect everything they touch, turning everything into NA. In this scenario, a total of the dollar values in a column would be NA if any of the values in that column is missing:\n\n\nppp_orig |&gt; \n  summarize ( total = sum(forgiveness_amount))\n\n\n\n  \n\n\n\n\nIgnore them in a computation completely, effectively removing that value from your calculation.\n\nThere’s no right answer, and it depends on what you’re doing. In some cases, you know that they stand for the value 0, and in others you don’t. We will usually ignore them by adding an argument to every summary function that could be infected by them : na.rm = TRUE , which means, “remove NA’s before you do anything.”.\n\nppp_orig |&gt; \n  summarize ( total = sum (forgiveness_amount, na.rm=TRUE))\n\n\n\n  \n\n\n\n\n\nSummary functions\nSome of the common functions you’ll use to summarize are :\n\nmean (column_name, na.rm=T) – for an average : Numbers only\nsum (column_name, na.rm = T): Numbers only\nn() – for “how many”, or “count”. Anything - this counts rows, not values\nn_distinct ( column_name) : The number of unique entries in the column. Use it to see how many categories there are in a column.\nmedian (column_name, na.rm=T): Numbers only\nmin (column_name, na.rm=T): Dates and numbers\nmax (column_name , na.rm=T): Dates and numbers\n\nWhen used on the whole data frame, it’s customary to just glimpse the output, since there’s only one row:\n\nppp_orig |&gt; \n  summarize ( n(), \n              mean (amount, na.rm=T), \n              mean (forgiveness_amount, na.rm=T), \n              min (date_approved, na.rm=T), \n              max (date_approved, na.rm= T), \n              n_distinct ( business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 6\n$ `n()`                                 &lt;int&gt; 169259\n$ `mean(amount, na.rm = T)`             &lt;dbl&gt; 73206.66\n$ `mean(forgiveness_amount, na.rm = T)` &lt;dbl&gt; 77050.41\n$ `min(date_approved, na.rm = T)`       &lt;date&gt; 2020-04-03\n$ `max(date_approved, na.rm = T)`       &lt;date&gt; 2021-06-29\n$ `n_distinct(business_type)`           &lt;int&gt; 24\n\n\nThis produced a data frame with 1 row and 5 columns. The column names are the same as the formulas that created them, which is difficult to work with. Create new column names using the name (in back-ticks if it’s got spaces or special characters) and assign them the values of the summaries using the = sign:\n\nppp_orig |&gt; \n  summarize ( number_of_rows =  n(), \n              mean_amount = mean (amount, na.rm=T),    \n              median_amount = median (amount, na.rm=T),\n              mean_forgiven = mean (forgiveness_amount, na.rm=T),   \n              first_loan = min (date_approved, na.rm=T),   \n              last_loan = max (date_approved, na.rm= T), \n              business_type_count = n_distinct(business_type)\n  ) |&gt;\n  glimpse()\n\nRows: 1\nColumns: 7\n$ number_of_rows      &lt;int&gt; 169259\n$ mean_amount         &lt;dbl&gt; 73206.66\n$ median_amount       &lt;dbl&gt; 20800\n$ mean_forgiven       &lt;dbl&gt; 77050.41\n$ first_loan          &lt;date&gt; 2020-04-03\n$ last_loan           &lt;date&gt; 2021-06-29\n$ business_type_count &lt;int&gt; 24\n\n\nNote that the mean forgiven removes those with missing values for the forgiven amount, which is wrong! We need to turn them into zeroes first."
-  },
-  {
-    "objectID": "r-verb-groupby.html#grouping-for-lists",
-    "href": "r-verb-groupby.html#grouping-for-lists",
-    "title": "17  Verbs in depth: Aggregating with groups",
-    "section": "17.2 Grouping for lists",
-    "text": "17.2 Grouping for lists\nNow that you know how to summarize the whole data frame, you’ll want to start getting totals by category. This is the same thing as a pivot table in spreadsheets: the column names that create the “groups” are the equivalent of the Rows area a spreadsheet pivot table:\n\n\nGrouping by one column\nIn the PPP data, the “draw” refers to which of the two programs was involved - the original one created in April 2020, or the one with stricter criteria passed by Congress that December.\nHere’s how we’d get some key statistics by draw:\n\nppp_orig |&gt;\n  group_by ( draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              `# of loans` = n() \n  )\n\n\n\n  \n\n\n\nHere are a couple of things to note about grouped output:\n\nThe only columns saved are the ones that are shown in either the group_by or summarize rows. All of the other original columns have been eliminated. You no longer have them to work with .\nThe names of the columns for the summary statistics are the ones defined before the “=” sign in the summarize statement.\nTRAP! Don’t ever name your summary columns the same thing as a group_by column. It will override those names, and your output will be unintelligible.\n\n\n\n\n\n\n\nNaming your columns\n\n\n\nNote that the name of the columns doesn’t always follow our standard. In this case, # of loans has a special character and spaces. In order to create or use it, you must enclose them in back-tics (`) or you’ll get an error.\n\n\n\n\nGrouping by more than one column\nIf you wanted to know the numbers outstanding and forgiven by draw, you could add another column to the group by:\n\nppp_orig |&gt;\n  group_by ( loan_status, draw ) |&gt;\n  summarize ( first_loan = min ( date_approved ), \n              total_amount = sum (amount), \n              total_forgiven = sum (forgiveness_amount, na.rm=T), \n              loan_ct = n() \n  )\n\n\n\n  \n\n\n\n\n\nA shortcut : count()\nIf all you want to do is count or add by group, you can use the count() function as a shortcut. It does the exact same thing as a combination of group_by() and summarize( n() ) and arrange()` to get the number of items in each category, sorted by the most frequent to least:\n\nppp_orig |&gt; \n  count ( loan_status, draw, \n          sort=TRUE,\n          name = \"loan_ct\")\n\n\n\n  \n\n\n\n\n\n\n\n\n\nNew ways to summarize\n\n\n\nA new version of the tidyverse, in 2023, has changed some of the options for the summarize() verb that make grouping unnecessary much of the time, but it’s confusing. I’m skipping it for now."
-  },
-  {
-    "objectID": "r-verb-join.html#join-syntax",
-    "href": "r-verb-join.html#join-syntax",
-    "title": "18  Verbs in depth: Matchmaking with joins",
-    "section": "18.2 join syntax",
-    "text": "18.2 join syntax\n\n\n\n\n\n\nTip\n\n\n\nThere are actually pretty good explanations of the concept of joining and the variations of it in R in the documentation.\n\n\nThere are several kinds of joins, but the syntax is similar across them.\n  old_table |&gt;\n     inner_join (new_table , \n     join_by = (name of old_table column == name of new_table column) )\n     \n\n\n\n\n\n\nTip\n\n\n\nTry to piece together what this is saying, similar to the filter conditions: If the two columns are equal (double-equal signs), put them together.\n\n\n\nTypes of joins\n\nAn inner_join means that the value(s) in the common columns must match in BOTH tables – it will eliminate any row without a match\nA left_join or right_join keeps everything from one table, and only the information that matches from the other. Those columns will contain NA wherever the match fails.\nA full_join puts together both tables no matter whether there is any match. It’s pretty rare to use this."
   }
 ]
\ No newline at end of file
diff --git a/docs/sitemap.xml b/docs/sitemap.xml
index 07ba6f9..c1cd5a2 100644
--- a/docs/sitemap.xml
+++ b/docs/sitemap.xml
@@ -2,162 +2,162 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/index.html</loc>
-    <lastmod>2023-12-27T17:15:55.862Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.193Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start.html</loc>
-    <lastmod>2023-12-22T21:49:27.320Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.198Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-story.html</loc>
-    <lastmod>2023-12-28T17:02:59.590Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.203Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-data-def.html</loc>
-    <lastmod>2023-12-22T21:49:27.330Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.208Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-data-document.html</loc>
-    <lastmod>2023-12-22T21:49:27.334Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.213Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-hunt.html</loc>
-    <lastmod>2023-12-22T21:49:27.341Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.220Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-build-own.html</loc>
-    <lastmod>2023-12-22T21:49:27.346Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.225Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/quickstart.html</loc>
-    <lastmod>2023-12-25T17:18:01.609Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.229Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-install.html</loc>
-    <lastmod>2023-12-22T21:49:27.354Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.233Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/appendix-program.html</loc>
-    <lastmod>2023-12-22T21:49:27.361Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.242Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-start.html</loc>
-    <lastmod>2023-12-22T21:49:27.366Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.247Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-quarto.html</loc>
-    <lastmod>2023-12-22T21:49:27.373Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.254Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-data-import.html</loc>
-    <lastmod>2023-12-22T21:49:27.383Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.265Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-verbs.html</loc>
-    <lastmod>2023-12-27T18:54:32.902Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.274Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/rdepth.html</loc>
-    <lastmod>2023-12-27T18:44:44.446Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.278Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-tidy.html</loc>
-    <lastmod>2023-12-22T21:49:27.402Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.301Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/start-filter-sort-story.html</loc>
-    <lastmod>2023-12-27T17:22:42.980Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.305Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-verb-filter.html</loc>
-    <lastmod>2023-12-27T18:03:22.006Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.313Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-verb-mutate.html</loc>
-    <lastmod>2023-12-27T19:38:55.774Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.322Z</lastmod>
+  </url>
+  <url>
+    <loc>https://cronkitedata.github.io/djtextbook/start-groupby.html</loc>
+    <lastmod>2023-12-28T22:05:30.331Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-verb-groupby.html</loc>
-    <lastmod>2023-12-27T22:19:22.684Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.342Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-verb-join.html</loc>
-    <lastmod>2023-12-27T22:34:22.313Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.351Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-regex.html</loc>
-    <lastmod>2023-12-22T21:49:27.463Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.358Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-regex-r.html</loc>
-    <lastmod>2023-12-22T21:49:27.472Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.370Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/r-recipes.html</loc>
-    <lastmod>2023-12-22T21:49:27.481Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.378Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis.html</loc>
-    <lastmod>2023-12-22T21:49:27.485Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.382Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-reporting.html</loc>
-    <lastmod>2023-12-22T21:49:27.490Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.389Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-tables.html</loc>
-    <lastmod>2023-12-22T21:49:27.500Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.397Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-demo.html</loc>
-    <lastmod>2023-12-22T21:49:27.516Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.416Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-practice.html</loc>
-    <lastmod>2023-12-22T21:49:27.524Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.421Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-maps-begin.html</loc>
-    <lastmod>2023-12-22T21:49:27.529Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.427Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/vis-maps-firstmap.html</loc>
-    <lastmod>2023-12-22T21:49:27.566Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.468Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced.html</loc>
-    <lastmod>2023-12-22T21:49:27.569Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.471Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-pdf.html</loc>
-    <lastmod>2023-12-22T21:49:27.578Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.477Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-openrefine.html</loc>
-    <lastmod>2023-12-22T21:49:27.584Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.483Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-scrape1.html</loc>
-    <lastmod>2023-12-22T21:49:27.589Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.491Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/advanced-scrape2.html</loc>
-    <lastmod>2023-12-22T21:49:27.602Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.501Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/appendix-math.html</loc>
-    <lastmod>2023-12-22T21:49:27.609Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.511Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/appendix-filetypes.html</loc>
-    <lastmod>2023-12-22T21:49:27.615Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.515Z</lastmod>
   </url>
   <url>
     <loc>https://cronkitedata.github.io/djtextbook/appendix-ppp.html</loc>
-    <lastmod>2023-12-25T17:32:30.672Z</lastmod>
-  </url>
-  <url>
-    <loc>https://cronkitedata.github.io/djtextbook/start-groupby.html</loc>
-    <lastmod>2023-12-27T19:39:54.135Z</lastmod>
+    <lastmod>2023-12-28T22:05:30.520Z</lastmod>
   </url>
 </urlset>
diff --git a/docs/start-build-own.html b/docs/start-build-own.html
index 0c7db9e..50f25bb 100644
--- a/docs/start-build-own.html
+++ b/docs/start-build-own.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/start-data-def.html b/docs/start-data-def.html
index 626cd0c..6280e0c 100644
--- a/docs/start-data-def.html
+++ b/docs/start-data-def.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -392,7 +398,7 @@ <h2 id="toc-title">In this chapter:</h2>
   <ul class="collapse">
   <li><a href="#example-death-certificates" id="toc-example-death-certificates" class="nav-link" data-scroll-target="#example-death-certificates">Example: Death certificates</a></li>
   </ul></li>
-  <li><a href="#nouns" id="toc-nouns" class="nav-link" data-scroll-target="#nouns"><span class="header-section-number">2.3</span> Nouns</a></li>
+  <li><a href="#unit-of-analysis-know-your-nouns" id="toc-unit-of-analysis-know-your-nouns" class="nav-link" data-scroll-target="#unit-of-analysis-know-your-nouns"><span class="header-section-number">2.3</span> Unit of analysis: Know your nouns</a></li>
   <li><a href="#further-reading" id="toc-further-reading" class="nav-link" data-scroll-target="#further-reading"><span class="header-section-number">2.4</span> Further reading</a></li>
   <li><a href="#exercises" id="toc-exercises" class="nav-link" data-scroll-target="#exercises"><span class="header-section-number">2.5</span> Exercises</a></li>
   </ul>
@@ -426,7 +432,7 @@ <h1 class="title"><span id="start-data-def" class="quarto-section-identifier"><s
 <section id="the-birth-of-a-dataset" class="level2" data-number="2.1">
 <h2 data-number="2.1" class="anchored" data-anchor-id="the-birth-of-a-dataset"><span class="header-section-number">2.1</span> The birth of a dataset</h2>
 <p>In “<a href="https://www.amazon.com/Art-Access-Strategies-Acquiring-Records/dp/1604265507">The Art of Access</a>”, David Cuillier and Charles N. Davis describe a process of tracking down the life and times of a dataset. Their purpose is to make sure they know how to request it from a government agency. The same idea applies to using data that we acquire elsewhere.</p>
-<p>As reporters, we usually deal with data that was created in the process of doing something else – conducting an inspection or paying a parking ticket. These adminsitrative datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. In the sciences, those are often called “administrative records”.</p>
+<p>As reporters, we usually deal with data that was created in the process of doing something else such as conducting an inspection or paying a parking ticket. These datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. These are often called “administrative records”.</p>
 <p>Another type of data might be considered “digital trace” data, which often refers to social media posts, online publications and other items that are born in electronic form and posted publicly for anyone to view and use.</p>
 <p>Finally, there are datasets that are compiled or collected specifically for the purpose of studying something. They might be collected in the form of a survey or a poll, or they might be monitoring systems such as pollution or weather stations. In this instance, the information has intrinsic value AS information.</p>
 </section>
@@ -445,8 +451,8 @@ <h3 class="anchored" data-anchor-id="example-death-certificates">Example: Death
 </div>
 <p><a href="https://www.cdc.gov/nchs/data/dvs/death11-03final-acc.pdf">Here is a blank version</a> of the standard US death certificate form – notice the detail and the detailed instructions on how it is supposed to be filled out.</p>
 <p>A good reporter could imagine many stories coming out of these little boxes. Limiting yourself to just to COVID-19-related stories: You could profile the local doctor who signed the most COVID-19-related death certificates in their city, or examine the number of deaths that had COVID as a contributing, but not underlying or immediate, cause of death. Or maybe you would want to map the deaths to find the block in your town most devastated by the virus.</p>
-<p><img src="assets/images/start-data-def-01.png" class="float-md-end img-fluid m-3 img-fluid" style="width:60.0%"></p>
 <p>Early in the pandemic, Coulter Jones and Jon Kamp of the Wall Street Journal examined the records from one of the few states that makes them public, and concluded that “<a href="https://www.wsj.com/articles/coronavirus-deaths-were-likely-missed-in-michigan-death-certificates-suggest-11590073280%5D%20%5E%5BWall%20Street%20Journal,%20May%2021,%202020">Coronavirus Deaths were Likely Missed in Michigan, Death Certificates Suggest</a>”</p>
+<p><img src="assets/images/start-data-def-01.png" class="float-md-end img-fluid m-3 img-fluid" style="width:60.0%"></p>
 <p>But you probably can’t do that. The reason is that, in most states, death certificates are not public records and are treated as secrets. <a href="#fn2" class="footnote-ref" id="fnref2" role="doc-noteref"><sup>2</sup></a>. Instead, state and local governments provide limited statistics related to the deaths, usually by county, with no detail.</p>
 <p>Here’s an example from Arizona — note that we can only see statistics in the way the data source has decided we want to examine them, without access to the underlying information. There’s no way to look at age and race and gender combined for each county, just the generalized statistics for each category alone.</p>
 <p>Here are some of the typical (not universal) characteristics of granular vs.&nbsp;aggregated data:</p>
@@ -492,10 +498,10 @@ <h3 class="anchored" data-anchor-id="example-death-certificates">Example: Death
 <p>For our purposes, it’s important to remember that we can always create aggregated numbers like the ones shown in the Arizona COVID page out of individual items, but you can never de-aggregate statistics into more granular data like the boxes filled out in death certificates.</p>
 </section>
 </section>
-<section id="nouns" class="level2" data-number="2.3">
-<h2 data-number="2.3" class="anchored" data-anchor-id="nouns"><span class="header-section-number">2.3</span> Nouns</h2>
+<section id="unit-of-analysis-know-your-nouns" class="level2" data-number="2.3">
+<h2 data-number="2.3" class="anchored" data-anchor-id="unit-of-analysis-know-your-nouns"><span class="header-section-number">2.3</span> Unit of analysis: Know your nouns</h2>
 <p>That brings us to one of the most important things you must find out about any data you begin to analyze: What <strong>noun</strong> describes each row? In statistics, these rows might be called <em>observations</em> or <em>cases</em>. In data science, they’re usually called <em>records</em>. Either way, every row must represent the same thing – a person, a place, a year, a water sample or a school. And you can’t really do anything with it until you figure out what that noun is.</p>
-<p>Sometimes the only noun you can obtain is problematic. In 2015, we did a story at The New York Times called <a href="https://www.nytimes.com/2014/04/07/us/more-deportations-follow-minor-crimes-data-shows.html">“More Deportation Follow Minor Crimes, Records Show”</a> . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions.</p>
+<p>Sometimes the only level you can obtain creates problems. In 2015, we did a story at The New York Times called <a href="https://www.nytimes.com/2014/04/07/us/more-deportations-follow-minor-crimes-data-shows.html">“More Deportation Follow Minor Crimes, Records Show”</a> . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions.</p>
 <p>In writing the piece, we had to work around a problem in our data: the agency refused to provide us anything that would help us distinguish one individual from another. All we knew was that each row represented one deportation, not a person. Without a column that uniquely identified people – say, name and date of birth, or some scrambled version of an their DHS number – we had no way to even estimate how often people were deported multiple times. If you read the story, you’ll see the very careful wording, except when we had reported out and spoken to people on the ground.</p>
 </section>
 <section id="further-reading" class="level2" data-number="2.4">
diff --git a/docs/start-hunt.html b/docs/start-hunt.html
index 7bb7797..c1efffe 100644
--- a/docs/start-hunt.html
+++ b/docs/start-hunt.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/start-tidy.html b/docs/start-tidy.html
index 801705c..10e46d9 100644
--- a/docs/start-tidy.html
+++ b/docs/start-tidy.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/start.html b/docs/start.html
index aa1cfb7..de7d45e 100644
--- a/docs/start.html
+++ b/docs/start.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
diff --git a/docs/vis-demo.html b/docs/vis-demo.html
index 2ddcdb2..e3e0713 100644
--- a/docs/vis-demo.html
+++ b/docs/vis-demo.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 23&nbsp; Visualization demo</title>
+<title>Data Reporting - 24&nbsp; Visualization demo</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -121,7 +121,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-demo.html"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-demo.html"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -278,35 +278,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -323,37 +329,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -370,25 +376,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -432,16 +438,16 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#intro-to-visualization-in-r" id="toc-intro-to-visualization-in-r" class="nav-link active" data-scroll-target="#intro-to-visualization-in-r"><span class="header-section-number">23.1</span> Intro to visualization in R</a></li>
-  <li><a href="#the-grammar-of-graphics" id="toc-the-grammar-of-graphics" class="nav-link" data-scroll-target="#the-grammar-of-graphics"><span class="header-section-number">23.2</span> The Grammar of Graphics</a></li>
-  <li><a href="#building-you-graph-in-pieces-using-gapminder" id="toc-building-you-graph-in-pieces-using-gapminder" class="nav-link" data-scroll-target="#building-you-graph-in-pieces-using-gapminder"><span class="header-section-number">23.3</span> Building you graph in pieces using gapminder</a>
+  <li><a href="#intro-to-visualization-in-r" id="toc-intro-to-visualization-in-r" class="nav-link active" data-scroll-target="#intro-to-visualization-in-r"><span class="header-section-number">24.1</span> Intro to visualization in R</a></li>
+  <li><a href="#the-grammar-of-graphics" id="toc-the-grammar-of-graphics" class="nav-link" data-scroll-target="#the-grammar-of-graphics"><span class="header-section-number">24.2</span> The Grammar of Graphics</a></li>
+  <li><a href="#building-you-graph-in-pieces-using-gapminder" id="toc-building-you-graph-in-pieces-using-gapminder" class="nav-link" data-scroll-target="#building-you-graph-in-pieces-using-gapminder"><span class="header-section-number">24.3</span> Building you graph in pieces using gapminder</a>
   <ul class="collapse">
   <li><a href="#subset-to-just-the-latest-years" id="toc-subset-to-just-the-latest-years" class="nav-link" data-scroll-target="#subset-to-just-the-latest-years">Subset to just the latest years</a></li>
   <li><a href="#the-histogram" id="toc-the-histogram" class="nav-link" data-scroll-target="#the-histogram">The histogram</a></li>
   <li><a href="#try-a-scatter-or-dot-plot" id="toc-try-a-scatter-or-dot-plot" class="nav-link" data-scroll-target="#try-a-scatter-or-dot-plot">Try a scatter, or dot plot</a></li>
   <li><a href="#adding-interactivity" id="toc-adding-interactivity" class="nav-link" data-scroll-target="#adding-interactivity">Adding interactivity</a></li>
   </ul></li>
-  <li><a href="#adding-animation" id="toc-adding-animation" class="nav-link" data-scroll-target="#adding-animation"><span class="header-section-number">23.4</span> Adding animation</a></li>
+  <li><a href="#adding-animation" id="toc-adding-animation" class="nav-link" data-scroll-target="#adding-animation"><span class="header-section-number">24.4</span> Adding animation</a></li>
   </ul>
 </nav>
     </div>
@@ -450,7 +456,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="vis-demo" class="quarto-section-identifier"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></h1>
+<h1 class="title"><span id="vis-demo" class="quarto-section-identifier"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></h1>
 </div>
 
 
@@ -480,8 +486,8 @@ <h1 class="title"><span id="vis-demo" class="quarto-section-identifier"><span cl
 <li>“<a href="https://datajournalism.com/read/handbook/one/understanding-data/using-data-visualization-to-find-insights-in-data">Using data visualization to find insights in data</a>”, by Gregor Aisch, former New York Times assistant graphics editor. His examples were created before the Tidyverse became popular, so his R code may not translate well for you.</li>
 </ul>
 </div>
-<section id="intro-to-visualization-in-r" class="level2" data-number="23.1">
-<h2 data-number="23.1" class="anchored" data-anchor-id="intro-to-visualization-in-r"><span class="header-section-number">23.1</span> Intro to visualization in R</h2>
+<section id="intro-to-visualization-in-r" class="level2" data-number="24.1">
+<h2 data-number="24.1" class="anchored" data-anchor-id="intro-to-visualization-in-r"><span class="header-section-number">24.1</span> Intro to visualization in R</h2>
 <p>You’ll need some additional packages installed for this to work:</p>
 <ul>
 <li>gapminder - example data that is used throughout, even though it’s a little boring</li>
@@ -499,8 +505,8 @@ <h2 data-number="23.1" class="anchored" data-anchor-id="intro-to-visualization-i
 <span id="cb2-6"><a href="#cb2-6" aria-hidden="true" tabindex="-1"></a><span class="fu">options</span>(<span class="at">digits=</span><span class="dv">2</span>, <span class="at">scipen=</span><span class="dv">99</span>)</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
 </section>
-<section id="the-grammar-of-graphics" class="level2" data-number="23.2">
-<h2 data-number="23.2" class="anchored" data-anchor-id="the-grammar-of-graphics"><span class="header-section-number">23.2</span> The Grammar of Graphics</h2>
+<section id="the-grammar-of-graphics" class="level2" data-number="24.2">
+<h2 data-number="24.2" class="anchored" data-anchor-id="the-grammar-of-graphics"><span class="header-section-number">24.2</span> The Grammar of Graphics</h2>
 <p>The package we’re going to use for our graphing is called <code>ggplot2</code>, which is part of the tidyverse. Conceptually, it follows the principles in Leland Wilkinson’s “Grammar of Graphics”, written in 1999 as a framework for thinking about visualizations. For our purposes, you just need to understand that any visualization is made up of several fundamental pieces. In <code>ggplot2</code>, they are:</p>
 <ul>
 <li><em>aesthetics</em> (called “aes”) : What <em>data</em> are you plotting? It can have more than two dimensions, such as x-axis, y-axis, a column used for color and another used for size. These are things we can see on a graphic, including the values.</li>
@@ -526,8 +532,8 @@ <h2 data-number="23.2" class="anchored" data-anchor-id="the-grammar-of-graphics"
 <span id="cb3-12"><a href="#cb3-12" aria-hidden="true" tabindex="-1"></a>                any scale information</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>We’re going to skip the labeling and annotation parts for now. That would come under layering and legends. In practice, these options can be mixed througout the command.</p>
 </section>
-<section id="building-you-graph-in-pieces-using-gapminder" class="level2" data-number="23.3">
-<h2 data-number="23.3" class="anchored" data-anchor-id="building-you-graph-in-pieces-using-gapminder"><span class="header-section-number">23.3</span> Building you graph in pieces using gapminder</h2>
+<section id="building-you-graph-in-pieces-using-gapminder" class="level2" data-number="24.3">
+<h2 data-number="24.3" class="anchored" data-anchor-id="building-you-graph-in-pieces-using-gapminder"><span class="header-section-number">24.3</span> Building you graph in pieces using gapminder</h2>
 <p>Gapminder is a datast made famous in a viral 2010 TED talk. It contains the life expectency and income (GDP per capita) by country for 200 years. The data is in 5-year increments, through 2007.</p>
 <section id="subset-to-just-the-latest-years" class="level3">
 <h3 class="anchored" data-anchor-id="subset-to-just-the-latest-years">Subset to just the latest years</h3>
@@ -714,8 +720,8 @@ <h3 class="anchored" data-anchor-id="adding-interactivity">Adding interactivity<
 </div>
 </section>
 </section>
-<section id="adding-animation" class="level2" data-number="23.4">
-<h2 data-number="23.4" class="anchored" data-anchor-id="adding-animation"><span class="header-section-number">23.4</span> Adding animation</h2>
+<section id="adding-animation" class="level2" data-number="24.4">
+<h2 data-number="24.4" class="anchored" data-anchor-id="adding-animation"><span class="header-section-number">24.4</span> Adding animation</h2>
 <p>So far, we’ve only looked at the year 2007 – what if we wanted to look at it over time, say 50 or so years? Let’s take the years 1957 to 2007:</p>
 <div class="cell">
 <div class="sourceCode cell-code" id="cb17"><pre class="sourceCode r code-with-copy"><code class="sourceCode r"><span id="cb17-1"><a href="#cb17-1" aria-hidden="true" tabindex="-1"></a>df_gap <span class="ot">&lt;-</span> </span>
@@ -1004,12 +1010,12 @@ <h2 data-number="23.4" class="anchored" data-anchor-id="adding-animation"><span
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-tables.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-practice.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/vis-maps-begin.html b/docs/vis-maps-begin.html
index 6040158..351f603 100644
--- a/docs/vis-maps-begin.html
+++ b/docs/vis-maps-begin.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 25&nbsp; Mapping: An introduction</title>
+<title>Data Reporting - 26&nbsp; Mapping: An introduction</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -76,7 +76,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-maps-begin.html"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-maps-begin.html"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -387,8 +393,8 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#the-power-of-maps" id="toc-the-power-of-maps" class="nav-link active" data-scroll-target="#the-power-of-maps"><span class="header-section-number">25.1</span> The power of maps</a></li>
-  <li><a href="#mapping-faq" id="toc-mapping-faq" class="nav-link" data-scroll-target="#mapping-faq"><span class="header-section-number">25.2</span> Mapping FAQ</a>
+  <li><a href="#the-power-of-maps" id="toc-the-power-of-maps" class="nav-link active" data-scroll-target="#the-power-of-maps"><span class="header-section-number">26.1</span> The power of maps</a></li>
+  <li><a href="#mapping-faq" id="toc-mapping-faq" class="nav-link" data-scroll-target="#mapping-faq"><span class="header-section-number">26.2</span> Mapping FAQ</a>
   <ul class="collapse">
   <li><a href="#what-is-gis" id="toc-what-is-gis" class="nav-link" data-scroll-target="#what-is-gis">What is GIS?</a></li>
   <li><a href="#what-kinds-of-maps-do-we-use" id="toc-what-kinds-of-maps-do-we-use" class="nav-link" data-scroll-target="#what-kinds-of-maps-do-we-use">What kinds of maps do we use?</a></li>
@@ -397,7 +403,7 @@ <h2 id="toc-title">In this chapter:</h2>
   <li><a href="#where-do-i-get-map-data" id="toc-where-do-i-get-map-data" class="nav-link" data-scroll-target="#where-do-i-get-map-data">Where do I get map data?</a></li>
   <li><a href="#what-if-my-data-doesnt-have-longitude-and-latitude" id="toc-what-if-my-data-doesnt-have-longitude-and-latitude" class="nav-link" data-scroll-target="#what-if-my-data-doesnt-have-longitude-and-latitude">What if my data doesn’t have longitude and latitude?</a></li>
   </ul></li>
-  <li><a href="#a-note-on-projections" id="toc-a-note-on-projections" class="nav-link" data-scroll-target="#a-note-on-projections"><span class="header-section-number">25.3</span> A note on projections</a></li>
+  <li><a href="#a-note-on-projections" id="toc-a-note-on-projections" class="nav-link" data-scroll-target="#a-note-on-projections"><span class="header-section-number">26.3</span> A note on projections</a></li>
   </ul>
 </nav>
     </div>
@@ -406,7 +412,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="vis-maps-begin.qmd" class="quarto-section-identifier"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></h1>
+<h1 class="title"><span id="vis-maps-begin.qmd" class="quarto-section-identifier"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></h1>
 </div>
 
 
@@ -428,8 +434,8 @@ <h1 class="title"><span id="vis-maps-begin.qmd" class="quarto-section-identifier
 <li><p>Compare areas to one another, identifying disparities or differences based on geography rather than other characteristics.</p></li>
 <li><p>Aggregate data based on location, such as crime hot spots.</p></li>
 </ul>
-<section id="the-power-of-maps" class="level2" data-number="25.1">
-<h2 data-number="25.1" class="anchored" data-anchor-id="the-power-of-maps"><span class="header-section-number">25.1</span> The power of maps</h2>
+<section id="the-power-of-maps" class="level2" data-number="26.1">
+<h2 data-number="26.1" class="anchored" data-anchor-id="the-power-of-maps"><span class="header-section-number">26.1</span> The power of maps</h2>
 <p>The most famous use of mapping may be the John Snow map, made in 1854 during a cholera outbreak in London. Officials didn’t know the cause, but Dr.&nbsp;Snow mapped the deaths and quickly saw the common element: The Broad Street water pump.</p>
 <div class="quarto-figure quarto-figure-center">
 <figure class="figure">
@@ -444,8 +450,8 @@ <h2 data-number="25.1" class="anchored" data-anchor-id="the-power-of-maps"><span
 </div>
 <p>Mapping seems magical and inconceivable until you do it. But it does require a little bit of vocabulary and a few basic concepts before you can accurately and confidently work with maps.</p>
 </section>
-<section id="mapping-faq" class="level2" data-number="25.2">
-<h2 data-number="25.2" class="anchored" data-anchor-id="mapping-faq"><span class="header-section-number">25.2</span> Mapping FAQ</h2>
+<section id="mapping-faq" class="level2" data-number="26.2">
+<h2 data-number="26.2" class="anchored" data-anchor-id="mapping-faq"><span class="header-section-number">26.2</span> Mapping FAQ</h2>
 <section id="what-is-gis" class="level3">
 <h3 class="anchored" data-anchor-id="what-is-gis">What is GIS?</h3>
 <p>Graphical Information Systems are programs and apps built to work with data that have a spatial, or geographic, element to them. In them, data is meant to be worked with in maps or on a spatial plane or globe rather than in columns and rows. Examples of GIS systems are ESRI’s ArcMap, which is used primarily by researchers and government agencies who create the underlying maps we’ll use in journalism. They are relatively hard on a computer, and require more memory and disk space than we’ve used.</p>
@@ -519,8 +525,8 @@ <h3 class="anchored" data-anchor-id="what-if-my-data-doesnt-have-longitude-and-l
 <p>This means that whenever you ask the government for data, you should ALWAYS ask for geographic coordinates and ask if they have GIS files (often using shapefiles). They may not give them to you, but it will save days of work and be much more accurate than trying to put them on a map yourself.</p>
 </section>
 </section>
-<section id="a-note-on-projections" class="level2" data-number="25.3">
-<h2 data-number="25.3" class="anchored" data-anchor-id="a-note-on-projections"><span class="header-section-number">25.3</span> A note on projections</h2>
+<section id="a-note-on-projections" class="level2" data-number="26.3">
+<h2 data-number="26.3" class="anchored" data-anchor-id="a-note-on-projections"><span class="header-section-number">26.3</span> A note on projections</h2>
 <div class="blockquote">
 <p>Projections are what we call the mathematical equations that do the trick of turning the world into some flat shape that fits on a printout or a computer screen. It’s a messy task to do, this transformation - there’s no way to smoosh the world onto a screen without distorting it in some way. You either lose direction, or relative size, or come out with something very weird looking.</p>
 <p>-- mapschool.io</p>
@@ -784,12 +790,12 @@ <h4 class="anchored" data-anchor-id="projected-coordinate-systems">Projected coo
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-practice.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-maps-firstmap.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/vis-maps-firstmap.html b/docs/vis-maps-firstmap.html
index f5dfad3..b1ab0c2 100644
--- a/docs/vis-maps-firstmap.html
+++ b/docs/vis-maps-firstmap.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 26&nbsp; The first map</title>
+<title>Data Reporting - 27&nbsp; The first map</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -124,7 +124,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-maps-firstmap.html"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-maps-firstmap.html"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -281,35 +281,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -326,37 +332,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -373,25 +379,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -435,17 +441,17 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#using-tmap-vs.-ggplot-and-leaflet" id="toc-using-tmap-vs.-ggplot-and-leaflet" class="nav-link active" data-scroll-target="#using-tmap-vs.-ggplot-and-leaflet"><span class="header-section-number">26.1</span> Using tmap vs.&nbsp;ggplot and leaflet</a></li>
-  <li><a href="#setup" id="toc-setup" class="nav-link" data-scroll-target="#setup"><span class="header-section-number">26.2</span> Setup</a></li>
-  <li><a href="#the-first-project" id="toc-the-first-project" class="nav-link" data-scroll-target="#the-first-project"><span class="header-section-number">26.3</span> The first project</a>
+  <li><a href="#using-tmap-vs.-ggplot-and-leaflet" id="toc-using-tmap-vs.-ggplot-and-leaflet" class="nav-link active" data-scroll-target="#using-tmap-vs.-ggplot-and-leaflet"><span class="header-section-number">27.1</span> Using tmap vs.&nbsp;ggplot and leaflet</a></li>
+  <li><a href="#setup" id="toc-setup" class="nav-link" data-scroll-target="#setup"><span class="header-section-number">27.2</span> Setup</a></li>
+  <li><a href="#the-first-project" id="toc-the-first-project" class="nav-link" data-scroll-target="#the-first-project"><span class="header-section-number">27.3</span> The first project</a>
   <ul class="collapse">
   <li><a href="#points" id="toc-points" class="nav-link" data-scroll-target="#points">Points</a></li>
   <li><a href="#using-the-tmap-library-to-display-the-map" id="toc-using-the-tmap-library-to-display-the-map" class="nav-link" data-scroll-target="#using-the-tmap-library-to-display-the-map">Using the <code>tmap</code> library to display the map</a></li>
   <li><a href="#adding-census-tracts" id="toc-adding-census-tracts" class="nav-link" data-scroll-target="#adding-census-tracts">Adding census tracts</a></li>
   <li><a href="#adding-census-data" id="toc-adding-census-data" class="nav-link" data-scroll-target="#adding-census-data">Adding Census data</a></li>
   </ul></li>
-  <li><a href="#a-chloropleth-map" id="toc-a-chloropleth-map" class="nav-link" data-scroll-target="#a-chloropleth-map"><span class="header-section-number">26.4</span> A chloropleth map</a></li>
-  <li><a href="#adding-the-voting-places-and-making-it-interactive" id="toc-adding-the-voting-places-and-making-it-interactive" class="nav-link" data-scroll-target="#adding-the-voting-places-and-making-it-interactive"><span class="header-section-number">26.5</span> Adding the voting places and making it interactive</a></li>
+  <li><a href="#a-chloropleth-map" id="toc-a-chloropleth-map" class="nav-link" data-scroll-target="#a-chloropleth-map"><span class="header-section-number">27.4</span> A chloropleth map</a></li>
+  <li><a href="#adding-the-voting-places-and-making-it-interactive" id="toc-adding-the-voting-places-and-making-it-interactive" class="nav-link" data-scroll-target="#adding-the-voting-places-and-making-it-interactive"><span class="header-section-number">27.5</span> Adding the voting places and making it interactive</a></li>
   </ul>
 </nav>
     </div>
@@ -454,7 +460,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="vis-map-firstmap.qmd" class="quarto-section-identifier"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></h1>
+<h1 class="title"><span id="vis-map-firstmap.qmd" class="quarto-section-identifier"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></h1>
 </div>
 
 
@@ -488,13 +494,13 @@ <h1 class="title"><span id="vis-map-firstmap.qmd" class="quarto-section-identifi
 <li><code>tmaptools</code> to allow for an underlying reference map</li>
 </ul>
 <p>This tutorial is based on one created by Aaron Kessler, senior data reporter at the Associated Press, for the 2022 NICAR hands-on workshop on visualization. His materials for that session are on his <a href="https://github.com/amkessler/NICAR2022_Dataviz_in_R/blob/master/Handson_GISMapping.Rmd">github repo</a></p>
-<section id="using-tmap-vs.-ggplot-and-leaflet" class="level2" data-number="26.1">
-<h2 data-number="26.1" class="anchored" data-anchor-id="using-tmap-vs.-ggplot-and-leaflet"><span class="header-section-number">26.1</span> Using tmap vs.&nbsp;ggplot and leaflet</h2>
+<section id="using-tmap-vs.-ggplot-and-leaflet" class="level2" data-number="27.1">
+<h2 data-number="27.1" class="anchored" data-anchor-id="using-tmap-vs.-ggplot-and-leaflet"><span class="header-section-number">27.1</span> Using tmap vs.&nbsp;ggplot and leaflet</h2>
 <p>There are several approaches to getting maps in R. In general, the heavy lifting is done the same way – through the sf (simple features) library. But displaying the maps can be done many different ways.</p>
 <p>The tmap library does a little more automatically than using the visualization library ggplot to make maps, so that’s the approach this tutorial will take. Neither is better - tmap is a little easier and requires a little less typing, but is not quite as easy to marry with other Javascript interactive maps or to customize using ggplot.</p>
 </section>
-<section id="setup" class="level2" data-number="26.2">
-<h2 data-number="26.2" class="anchored" data-anchor-id="setup"><span class="header-section-number">26.2</span> Setup</h2>
+<section id="setup" class="level2" data-number="27.2">
+<h2 data-number="27.2" class="anchored" data-anchor-id="setup"><span class="header-section-number">27.2</span> Setup</h2>
 <p>You’ll want to change your typical setup chunk to this when you start using maps:</p>
 <div class="cell">
 <div class="sourceCode cell-code" id="cb1"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb1-1"><a href="#cb1-1" aria-hidden="true" tabindex="-1"></a><span class="in">```{r}</span></span>
@@ -512,8 +518,8 @@ <h2 data-number="26.2" class="anchored" data-anchor-id="setup"><span class="head
 </div>
 <p>The options make it easier to work with the Census bureau’s maps.</p>
 </section>
-<section id="the-first-project" class="level2" data-number="26.3">
-<h2 data-number="26.3" class="anchored" data-anchor-id="the-first-project"><span class="header-section-number">26.3</span> The first project</h2>
+<section id="the-first-project" class="level2" data-number="27.3">
+<h2 data-number="27.3" class="anchored" data-anchor-id="the-first-project"><span class="header-section-number">27.3</span> The first project</h2>
 <p>Our first project involves getting points on a map of Maricopa County, then displaying it based on Census demographics. Here, we’ll use voting locations as the points and Census tract level data for income.</p>
 <section id="points" class="level3">
 <h3 class="anchored" data-anchor-id="points">Points</h3>
@@ -695,8 +701,8 @@ <h3 class="anchored" data-anchor-id="adding-census-data">Adding Census data</h3>
 <p>(You don’t need to select the “geometry” column – it has to be there if it’s a map.)</p>
 </section>
 </section>
-<section id="a-chloropleth-map" class="level2" data-number="26.4">
-<h2 data-number="26.4" class="anchored" data-anchor-id="a-chloropleth-map"><span class="header-section-number">26.4</span> A chloropleth map</h2>
+<section id="a-chloropleth-map" class="level2" data-number="27.4">
+<h2 data-number="27.4" class="anchored" data-anchor-id="a-chloropleth-map"><span class="header-section-number">27.4</span> A chloropleth map</h2>
 <p>Now we can use colors to show the percent of each tract that identified as a racial or ethnic minority in the American Community Surveys.</p>
 <div class="cell">
 <div class="sourceCode cell-code" id="cb19"><pre class="sourceCode r code-with-copy"><code class="sourceCode r"><span id="cb19-1"><a href="#cb19-1" aria-hidden="true" tabindex="-1"></a><span class="fu">tm_shape</span> ( maricopa_demo_map, </span>
@@ -721,8 +727,8 @@ <h4 class="anchored" data-anchor-id="color-palettes">Color palettes</h4>
 </div>
 </section>
 </section>
-<section id="adding-the-voting-places-and-making-it-interactive" class="level2" data-number="26.5">
-<h2 data-number="26.5" class="anchored" data-anchor-id="adding-the-voting-places-and-making-it-interactive"><span class="header-section-number">26.5</span> Adding the voting places and making it interactive</h2>
+<section id="adding-the-voting-places-and-making-it-interactive" class="level2" data-number="27.5">
+<h2 data-number="27.5" class="anchored" data-anchor-id="adding-the-voting-places-and-making-it-interactive"><span class="header-section-number">27.5</span> Adding the voting places and making it interactive</h2>
 <div class="cell">
 <div class="sourceCode cell-code" id="cb21"><pre class="sourceCode r code-with-copy"><code class="sourceCode r"><span id="cb21-1"><a href="#cb21-1" aria-hidden="true" tabindex="-1"></a><span class="fu">tmap_mode</span>(<span class="st">"view"</span>)</span>
 <span id="cb21-2"><a href="#cb21-2" aria-hidden="true" tabindex="-1"></a></span>
@@ -985,7 +991,7 @@ <h2 data-number="26.5" class="anchored" data-anchor-id="adding-the-voting-places
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-maps-begin.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
diff --git a/docs/vis-practice.html b/docs/vis-practice.html
index 92b19a9..5c8330d 100644
--- a/docs/vis-practice.html
+++ b/docs/vis-practice.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 24&nbsp; Visualization exercise</title>
+<title>Data Reporting - 25&nbsp; Visualization exercise</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -113,7 +113,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-practice.html"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-practice.html"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -270,35 +270,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -315,37 +321,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -362,25 +368,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -424,15 +430,15 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#the-data" id="toc-the-data" class="nav-link active" data-scroll-target="#the-data"><span class="header-section-number">24.1</span> The data</a>
+  <li><a href="#the-data" id="toc-the-data" class="nav-link active" data-scroll-target="#the-data"><span class="header-section-number">25.1</span> The data</a>
   <ul class="collapse">
   <li><a href="#examples" id="toc-examples" class="nav-link" data-scroll-target="#examples">Examples</a></li>
   </ul></li>
-  <li><a href="#your-turn" id="toc-your-turn" class="nav-link" data-scroll-target="#your-turn"><span class="header-section-number">24.2</span> Your turn</a>
+  <li><a href="#your-turn" id="toc-your-turn" class="nav-link" data-scroll-target="#your-turn"><span class="header-section-number">25.2</span> Your turn</a>
   <ul class="collapse">
   <li><a href="#common-geoms" id="toc-common-geoms" class="nav-link" data-scroll-target="#common-geoms">Common “geoms”</a></li>
   </ul></li>
-  <li><a href="#one-quick-plot" id="toc-one-quick-plot" class="nav-link" data-scroll-target="#one-quick-plot"><span class="header-section-number">24.3</span> One quick plot</a>
+  <li><a href="#one-quick-plot" id="toc-one-quick-plot" class="nav-link" data-scroll-target="#one-quick-plot"><span class="header-section-number">25.3</span> One quick plot</a>
   <ul class="collapse">
   <li><a href="#other-sources-to-learn-about-ggplot" id="toc-other-sources-to-learn-about-ggplot" class="nav-link" data-scroll-target="#other-sources-to-learn-about-ggplot">Other sources to learn about ggplot</a></li>
   </ul></li>
@@ -444,7 +450,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="vis-practice" class="quarto-section-identifier"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></h1>
+<h1 class="title"><span id="vis-practice" class="quarto-section-identifier"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></h1>
 </div>
 
 
@@ -472,8 +478,8 @@ <h1 class="title"><span id="vis-practice" class="quarto-section-identifier"><spa
 </ul>
 </div>
 <p>There are a lot of extra add-ons to the standard <code>ggplot</code> library that comes with the tidyverse, so you may end up having to load them to do what you want.</p>
-<section id="the-data" class="level2" data-number="24.1">
-<h2 data-number="24.1" class="anchored" data-anchor-id="the-data"><span class="header-section-number">24.1</span> The data</h2>
+<section id="the-data" class="level2" data-number="25.1">
+<h2 data-number="25.1" class="anchored" data-anchor-id="the-data"><span class="header-section-number">25.1</span> The data</h2>
 <p>We saw the Hot 100 data at the very beginning of your exploration of R. We’re going to use a slightly different version of it to create the visualizations because an older version was matched to the Spotify database to extract certain information from each of the songs.</p>
 <p>This data was used as a <a href="https://github.com/rfordatascience/tidytuesday/blob/master/data/2021/2021-09-14/readme.md">TidyTuesday dataset in 2021</a> (scroll down to get the data dictionaries), so it’s a little older than the one we worked on earlier.</p>
 <p>This data has two tables:</p>
@@ -507,8 +513,8 @@ <h4 class="anchored" data-anchor-id="visualizing-characteristics">Visualizing ch
 </section>
 </section>
 </section>
-<section id="your-turn" class="level2" data-number="24.2">
-<h2 data-number="24.2" class="anchored" data-anchor-id="your-turn"><span class="header-section-number">24.2</span> Your turn</h2>
+<section id="your-turn" class="level2" data-number="25.2">
+<h2 data-number="25.2" class="anchored" data-anchor-id="your-turn"><span class="header-section-number">25.2</span> Your turn</h2>
 <p>You’ve been through a lot, so let’s review the steps you have to take to make a visualization:</p>
 <p>Load the data in this case using a <code>read_csv()</code> function. There are two datasets, if you want to use them both.</p>
 <ol type="1">
@@ -560,8 +566,8 @@ <h3 class="anchored" data-anchor-id="common-geoms">Common “geoms”</h3>
 </ul>
 </section>
 </section>
-<section id="one-quick-plot" class="level2 page-columns page-full" data-number="24.3">
-<h2 data-number="24.3" class="anchored" data-anchor-id="one-quick-plot"><span class="header-section-number">24.3</span> One quick plot</h2>
+<section id="one-quick-plot" class="level2 page-columns page-full" data-number="25.3">
+<h2 data-number="25.3" class="anchored" data-anchor-id="one-quick-plot"><span class="header-section-number">25.3</span> One quick plot</h2>
 <p>The very first day we looked at R, I showed you one quick plot based on a Hot 100 dataset that I’d created with just the 21 songs that stayed on the Hot 100 for at least a year list during the 2000s. Now that code may make a little more sense to you</p>
 <div class="cell" data-lael="hot-100-read">
 <div class="sourceCode cell-code" id="cb2"><pre class="sourceCode r code-with-copy"><code class="sourceCode r"><span id="cb2-1"><a href="#cb2-1" aria-hidden="true" tabindex="-1"></a>hot100 <span class="ot">&lt;-</span> <span class="fu">readRDS</span>( </span>
@@ -834,12 +840,12 @@ <h3 class="anchored" data-anchor-id="other-sources-to-learn-about-ggplot">Other
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-demo.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-maps-begin.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/vis-reporting.html b/docs/vis-reporting.html
index f00eb9a..2c527ad 100644
--- a/docs/vis-reporting.html
+++ b/docs/vis-reporting.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 21&nbsp; Visualization as a reporting tool</title>
+<title>Data Reporting - 22&nbsp; Visualization as a reporting tool</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -76,7 +76,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-reporting.html"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-reporting.html"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -387,8 +393,8 @@
     <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
-  <li><a href="#examples" id="toc-examples" class="nav-link active" data-scroll-target="#examples"><span class="header-section-number">21.1</span> Examples</a></li>
-  <li><a href="#tips-for-exploratory-visualizations" id="toc-tips-for-exploratory-visualizations" class="nav-link" data-scroll-target="#tips-for-exploratory-visualizations"><span class="header-section-number">21.2</span> Tips for exploratory visualizations</a>
+  <li><a href="#examples" id="toc-examples" class="nav-link active" data-scroll-target="#examples"><span class="header-section-number">22.1</span> Examples</a></li>
+  <li><a href="#tips-for-exploratory-visualizations" id="toc-tips-for-exploratory-visualizations" class="nav-link" data-scroll-target="#tips-for-exploratory-visualizations"><span class="header-section-number">22.2</span> Tips for exploratory visualizations</a>
   <ul class="collapse">
   <li><a href="#use-small-multiples-to-quickly-orient-yourself-in-a-large-dataset" id="toc-use-small-multiples-to-quickly-orient-yourself-in-a-large-dataset" class="nav-link" data-scroll-target="#use-small-multiples-to-quickly-orient-yourself-in-a-large-dataset">Use small multiples to quickly orient yourself in a large dataset</a></li>
   <li><a href="#look-at-your-data-from-all-directions" id="toc-look-at-your-data-from-all-directions" class="nav-link" data-scroll-target="#look-at-your-data-from-all-directions">Look at your data from all directions</a></li>
@@ -405,7 +411,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span id="viz-reporting" class="quarto-section-identifier"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></h1>
+<h1 class="title"><span id="viz-reporting" class="quarto-section-identifier"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></h1>
 </div>
 
 
@@ -449,8 +455,8 @@ <h1 class="title"><span id="viz-reporting" class="quarto-section-identifier"><sp
 </figure>
 </div>
 <p>It shows the portion of farm income associated with subsidies and key events over the past 45 years, and was built over a series of months. Finding data that could be used over time with similar definitions and similar meanings was a challenge. Investigating all of the peaks and troughs helped us keep context in mind as we did the rest of our reporting. It also meant that one chore was pretty much finished before the stories were written.</p>
-<section id="examples" class="level2" data-number="21.1">
-<h2 data-number="21.1" class="anchored" data-anchor-id="examples"><span class="header-section-number">21.1</span> Examples</h2>
+<section id="examples" class="level2" data-number="22.1">
+<h2 data-number="22.1" class="anchored" data-anchor-id="examples"><span class="header-section-number">22.1</span> Examples</h2>
 <div class="columns">
 <div class="column" style="width:50%;">
 <section id="size-and-growth-at-once" class="level4">
@@ -483,8 +489,8 @@ <h4 class="anchored" data-anchor-id="networks-of-companies-people-or-programs">N
 <p><img src="assets/images/vis-reporting-gephi.png" class="img-fluid" style="width:75.0%"></p>
 </section>
 </section>
-<section id="tips-for-exploratory-visualizations" class="level2" data-number="21.2">
-<h2 data-number="21.2" class="anchored" data-anchor-id="tips-for-exploratory-visualizations"><span class="header-section-number">21.2</span> Tips for exploratory visualizations</h2>
+<section id="tips-for-exploratory-visualizations" class="level2" data-number="22.2">
+<h2 data-number="22.2" class="anchored" data-anchor-id="tips-for-exploratory-visualizations"><span class="header-section-number">22.2</span> Tips for exploratory visualizations</h2>
 <section id="use-small-multiples-to-quickly-orient-yourself-in-a-large-dataset" class="level3 unnumbered">
 <h3 class="unnumbered anchored" data-anchor-id="use-small-multiples-to-quickly-orient-yourself-in-a-large-dataset">Use small multiples to quickly orient yourself in a large dataset</h3>
 <p>“Small multiples” are repeated images, usually over time or place. This example shows where each round of agricultural disaster payments went, which helped us decide on where to focus our reporting – Kansas and North Dakota.</p>
@@ -771,7 +777,7 @@ <h3 class="unnumbered anchored" data-anchor-id="meet-with-designers-and-graphics
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-tables.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/vis-tables.html b/docs/vis-tables.html
index 1fe10f3..e92dfdd 100644
--- a/docs/vis-tables.html
+++ b/docs/vis-tables.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>Data Reporting - 22&nbsp; Report-making in R</title>
+<title>Data Reporting - 23&nbsp; Report-making in R</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -128,7 +128,7 @@
       <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
         <i class="bi bi-layout-text-sidebar-reverse"></i>
       </button>
-      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-tables.html"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></a></li></ol></nav>
+      <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="./vis.html">Visualization</a></li><li class="breadcrumb-item"><a href="./vis-tables.html"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></a></li></ol></nav>
       <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar,#quarto-sidebar-glass" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
       </a>
       <button type="button" class="btn quarto-search-button" aria-label="" onclick="window.quartoOpenSearch();">
@@ -285,35 +285,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -330,37 +336,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -377,25 +383,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -440,12 +446,12 @@ <h2 id="toc-title">In this chapter:</h2>
    
   <ul>
   <li><a href="#interactive-tables-with-dt" id="toc-interactive-tables-with-dt" class="nav-link active" data-scroll-target="#interactive-tables-with-dt">Interactive tables with <code>DT</code></a></li>
-  <li><a href="#interactive-tables-with-reactable" id="toc-interactive-tables-with-reactable" class="nav-link" data-scroll-target="#interactive-tables-with-reactable"><span class="header-section-number">22.1</span> Interactive tables with <code>reactable</code></a></li>
-  <li><a href="#crosstabs-with-the-janitor-library" id="toc-crosstabs-with-the-janitor-library" class="nav-link" data-scroll-target="#crosstabs-with-the-janitor-library"><span class="header-section-number">22.2</span> Crosstabs with the <code>janitor</code> library</a>
+  <li><a href="#interactive-tables-with-reactable" id="toc-interactive-tables-with-reactable" class="nav-link" data-scroll-target="#interactive-tables-with-reactable"><span class="header-section-number">23.1</span> Interactive tables with <code>reactable</code></a></li>
+  <li><a href="#crosstabs-with-the-janitor-library" id="toc-crosstabs-with-the-janitor-library" class="nav-link" data-scroll-target="#crosstabs-with-the-janitor-library"><span class="header-section-number">23.2</span> Crosstabs with the <code>janitor</code> library</a>
   <ul class="collapse">
   <li><a href="#a-simple-one-way-table" id="toc-a-simple-one-way-table" class="nav-link" data-scroll-target="#a-simple-one-way-table">A simple one-way table</a></li>
   </ul></li>
-  <li><a href="#static-tables-with-gt" id="toc-static-tables-with-gt" class="nav-link" data-scroll-target="#static-tables-with-gt"><span class="header-section-number">22.3</span> Static tables with gt</a></li>
+  <li><a href="#static-tables-with-gt" id="toc-static-tables-with-gt" class="nav-link" data-scroll-target="#static-tables-with-gt"><span class="header-section-number">23.3</span> Static tables with gt</a></li>
   </ul>
 </nav>
     </div>
@@ -454,7 +460,7 @@ <h2 id="toc-title">In this chapter:</h2>
 
 <header id="title-block-header" class="quarto-title-block default">
 <div class="quarto-title">
-<h1 class="title"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></h1>
+<h1 class="title"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></h1>
 </div>
 
 
@@ -529,15 +535,15 @@ <h3 class="anchored" data-anchor-id="interactive-tables-with-dt">Interactive tab
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-1" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="5" data-code-cell="annotated-cell-1">Create a datatable.</span>
+<span data-code-lines="5" data-code-cell="annotated-cell-1" data-code-annotation="1">Create a datatable.</span>
 </dd>
 <dt data-target-cell="annotated-cell-1" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-annotation="2" data-code-lines="7" data-code-cell="annotated-cell-1">Turn on a left-right scrollbar and reduce the default number of rows shown.</span>
+<span data-code-lines="7" data-code-cell="annotated-cell-1" data-code-annotation="2">Turn on a left-right scrollbar and reduce the default number of rows shown.</span>
 </dd>
 <dt data-target-cell="annotated-cell-1" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-annotation="3" data-code-lines="12" data-code-cell="annotated-cell-1">Format the two amounts as currency. <a href="https://rstudio.github.io/DT/functions.html">Other format options</a> exist for regular numbers and percentages.</span>
+<span data-code-lines="12" data-code-cell="annotated-cell-1" data-code-annotation="3">Format the two amounts as currency. <a href="https://rstudio.github.io/DT/functions.html">Other format options</a> exist for regular numbers and percentages.</span>
 </dd>
 </dl>
 </div>
@@ -549,8 +555,8 @@ <h3 class="anchored" data-anchor-id="interactive-tables-with-dt">Interactive tab
 </div>
 <p>One problem with the <code>DT</code> package is that the documentation is inscrutable if you don’t know Javascript. It is bare bones in R, and most Google searches send you to the Javascript documentation. I usually futz around and look for examples.</p>
 </section>
-<section id="interactive-tables-with-reactable" class="level2" data-number="22.1">
-<h2 data-number="22.1" class="anchored" data-anchor-id="interactive-tables-with-reactable"><span class="header-section-number">22.1</span> Interactive tables with <code>reactable</code></h2>
+<section id="interactive-tables-with-reactable" class="level2" data-number="23.1">
+<h2 data-number="23.1" class="anchored" data-anchor-id="interactive-tables-with-reactable"><span class="header-section-number">23.1</span> Interactive tables with <code>reactable</code></h2>
 <p>The <code>reactable</code> library is highly customizable, but it can involve a lot of typing to get a good table. The documentation is excellent and there are great examples in the wild. It was used to produce the table above that contained the sparklines.</p>
 <p>The basic idea is that you select the columns you want to display, then define their format and other details one by one. There are several useful defaults for the whole table, such as the number format. One way to simplify the typing is to just create the names of the columns you want to see in a <code>select</code> statement, then put a default number format for the dollar amounts. A <a href="https://glin.github.io/reactable/reference/colDef.html">full list of options</a> is available on the help page for reactable.</p>
 <div class="cell" data-code-annotations="hover">
@@ -579,19 +585,19 @@ <h2 data-number="22.1" class="anchored" data-anchor-id="interactive-tables-with-
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-2" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="2" data-code-cell="annotated-cell-2">using the <code>select</code> verb that we’ve already seen, pick out the columns for the table and rename them. Don’t forget the back-ticks if you need something with more than one word.</span>
+<span data-code-lines="2" data-code-cell="annotated-cell-2" data-code-annotation="1">using the <code>select</code> verb that we’ve already seen, pick out the columns for the table and rename them. Don’t forget the back-ticks if you need something with more than one word.</span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-annotation="2" data-code-lines="10" data-code-cell="annotated-cell-2">Set the options for sorting, filtering, and searching.</span>
+<span data-code-lines="10" data-code-cell="annotated-cell-2" data-code-annotation="2">Set the options for sorting, filtering, and searching.</span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-annotation="3" data-code-lines="14" data-code-cell="annotated-cell-2">I usually use the “compact” style because I’m trying to fit as much in the table as possible.</span>
+<span data-code-lines="14" data-code-cell="annotated-cell-2" data-code-annotation="3">I usually use the “compact” style because I’m trying to fit as much in the table as possible.</span>
 </dd>
 <dt data-target-cell="annotated-cell-2" data-target-annotation="4">4</dt>
 <dd>
-<span data-code-annotation="4" data-code-lines="16,19" data-code-cell="annotated-cell-2">Define how numbers will show up by default. You can change this for individual columns.</span>
+<span data-code-lines="16,19" data-code-cell="annotated-cell-2" data-code-annotation="4">Define how numbers will show up by default. You can change this for individual columns.</span>
 </dd>
 </dl>
 </div>
@@ -604,8 +610,8 @@ <h2 data-number="22.1" class="anchored" data-anchor-id="interactive-tables-with-
 <p>It’s a lot of typing for something so simple, but I save it as a code snippet in RStudio and use a shortcut to paste it into a code chunk.</p>
 <p>There is much, much more that reactable tables can do (like <a href="https://glin.github.io/reactable/articles/womens-world-cup/womens-world-cup.html">this</a>), but this is a good start toward something readable and presentable to your colleagues. It is also possible to create collapsable rows for large tables, and total rows at the bottom.</p>
 </section>
-<section id="crosstabs-with-the-janitor-library" class="level2" data-number="22.2">
-<h2 data-number="22.2" class="anchored" data-anchor-id="crosstabs-with-the-janitor-library"><span class="header-section-number">22.2</span> Crosstabs with the <code>janitor</code> library</h2>
+<section id="crosstabs-with-the-janitor-library" class="level2" data-number="23.2">
+<h2 data-number="23.2" class="anchored" data-anchor-id="crosstabs-with-the-janitor-library"><span class="header-section-number">23.2</span> Crosstabs with the <code>janitor</code> library</h2>
 <p>We’ve gone through the exercise of creating a grouping, counting, then computing a percentage of total using our usual verbs. The <code>janitor</code> package provides a utility for doing that.</p>
 <p>Here’s an example with the full PPP dataset:</p>
 <p>The <code>tabyl</code> function does one and only one thing: Count and compute percentages of totals. It creates an output data frame that you can style any way you want. You can do 1-, 2- or 3-way tables, but no more. This example creates the table, then formats it using a datatable as above.</p>
@@ -625,27 +631,27 @@ <h3 class="anchored" data-anchor-id="a-simple-one-way-table">A simple one-way ta
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-3" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="2" data-code-cell="annotated-cell-3">Create a one-way data table with count (n) and percent</span>
+<span data-code-lines="2" data-code-cell="annotated-cell-3" data-code-annotation="1">Create a one-way data table with count (n) and percent</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="2">2</dt>
 <dd>
-<span data-code-annotation="2" data-code-lines="3" data-code-cell="annotated-cell-3">Sort it from highest to lowest</span>
+<span data-code-lines="3" data-code-cell="annotated-cell-3" data-code-annotation="2">Sort it from highest to lowest</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="3">3</dt>
 <dd>
-<span data-code-annotation="3" data-code-lines="4" data-code-cell="annotated-cell-3">Keep only the top 5,</span>
+<span data-code-lines="4" data-code-cell="annotated-cell-3" data-code-annotation="3">Keep only the top 5,</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="4">4</dt>
 <dd>
-<span data-code-annotation="4" data-code-lines="6" data-code-cell="annotated-cell-3">create our default datatable</span>
+<span data-code-lines="6" data-code-cell="annotated-cell-3" data-code-annotation="4">create our default datatable</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="5">5</dt>
 <dd>
-<span data-code-annotation="5" data-code-lines="8" data-code-cell="annotated-cell-3">Format the percentages</span>
+<span data-code-lines="8" data-code-cell="annotated-cell-3" data-code-annotation="5">Format the percentages</span>
 </dd>
 <dt data-target-cell="annotated-cell-3" data-target-annotation="6">6</dt>
 <dd>
-<span data-code-annotation="6" data-code-lines="9" data-code-cell="annotated-cell-3">Format the raw numbers</span>
+<span data-code-lines="9" data-code-cell="annotated-cell-3" data-code-annotation="6">Format the raw numbers</span>
 </dd>
 </dl>
 </div>
@@ -683,7 +689,7 @@ <h4 class="anchored" data-anchor-id="two-way-table-with-tabyls">Two-way table wi
 <dl class="code-annotation-container-grid">
 <dt data-target-cell="annotated-cell-5" data-target-annotation="1">1</dt>
 <dd>
-<span data-code-annotation="1" data-code-lines="4" data-code-cell="annotated-cell-5">Change “col” to “row” if you want it to add to 100% across rather than down; “all” if you want it to be a percent of the grand total.</span>
+<span data-code-lines="4" data-code-cell="annotated-cell-5" data-code-annotation="1">Change “col” to “row” if you want it to add to 100% across rather than down; “all” if you want it to be a percent of the grand total.</span>
 </dd>
 </dl>
 </div>
@@ -700,8 +706,8 @@ <h4 class="anchored" data-anchor-id="two-way-table-with-tabyls">Two-way table wi
 </section>
 </section>
 </section>
-<section id="static-tables-with-gt" class="level2" data-number="22.3">
-<h2 data-number="22.3" class="anchored" data-anchor-id="static-tables-with-gt"><span class="header-section-number">22.3</span> Static tables with gt</h2>
+<section id="static-tables-with-gt" class="level2" data-number="23.3">
+<h2 data-number="23.3" class="anchored" data-anchor-id="static-tables-with-gt"><span class="header-section-number">23.3</span> Static tables with gt</h2>
 <p>Section TK.</p>
 
 
@@ -944,12 +950,12 @@ <h2 data-number="22.3" class="anchored" data-anchor-id="static-tables-with-gt"><
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./vis-reporting.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-demo.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/docs/vis.html b/docs/vis.html
index 86d5f8a..ed9b49d 100644
--- a/docs/vis.html
+++ b/docs/vis.html
@@ -233,35 +233,41 @@
   <a href="./r-verb-mutate.html" class="sidebar-item-text sidebar-link">
  <span class="menu-text"><span class="chapter-number">15</span>&nbsp; <span class="chapter-title">Verbs in depth: New from old data with Mutate</span></span></a>
   </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="./start-groupby.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Summarizing detail into totals</span></span></a>
+  </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-groupby.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">16</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
+ <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Aggregating with groups</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-verb-join.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">17</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
+ <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Verbs in depth: Matchmaking with joins</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">18</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
+ <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Regular Expressions for pattern matching</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-regex-r.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">19</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Inexact matching and regular expressions in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./r-recipes.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
+ <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span></a>
   </div>
 </li>
       </ul>
@@ -278,37 +284,37 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-reporting.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
+ <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-tables.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Report-making in R</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-demo.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">23</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
+ <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization demo</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-practice.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">24</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
+ <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Visualization exercise</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-begin.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">25</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
+ <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">Mapping: An introduction</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./vis-maps-firstmap.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">26</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
+ <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">The first map</span></span></a>
   </div>
 </li>
       </ul>
@@ -325,25 +331,25 @@
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-pdf.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">27</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
+ <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Working with PDF files</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-openrefine.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">28</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
+ <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Open refine walkthrough</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape1.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">29</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
+ <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Scraping without programming</span></span></a>
   </div>
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
   <a href="./advanced-scrape2.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text"><span class="chapter-number">30</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
+ <span class="menu-text"><span class="chapter-number">31</span>&nbsp; <span class="chapter-title">Introduction to scraping in R</span></span></a>
   </div>
 </li>
       </ul>
@@ -679,12 +685,12 @@ <h2 class="unnumbered anchored" data-anchor-id="visualization-tools">Visualizati
 <nav class="page-navigation column-body">
   <div class="nav-page nav-page-previous">
       <a href="./r-recipes.html" class="pagination-link">
-        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">20</span>&nbsp; <span class="chapter-title">Recipes</span></span>
+        <i class="bi bi-arrow-left-short"></i> <span class="nav-page-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Recipes</span></span>
       </a>          
   </div>
   <div class="nav-page nav-page-next">
       <a href="./vis-reporting.html" class="pagination-link">
-        <span class="nav-page-text"><span class="chapter-number">21</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span> <i class="bi bi-arrow-right-short"></i>
+        <span class="nav-page-text"><span class="chapter-number">22</span>&nbsp; <span class="chapter-title">Visualization as a reporting tool</span></span> <i class="bi bi-arrow-right-short"></i>
       </a>
   </div>
 </nav>
diff --git a/start-data-def.qmd b/start-data-def.qmd
index b60a575..8723b56 100644
--- a/start-data-def.qmd
+++ b/start-data-def.qmd
@@ -11,7 +11,7 @@
 
 In "[The Art of Access](https://www.amazon.com/Art-Access-Strategies-Acquiring-Records/dp/1604265507)", David Cuillier  and Charles N. Davis describe a process of tracking down the life and times of a dataset. Their purpose is to make sure they know how to request it from a government agency. The same idea applies to using data that we acquire elsewhere. 
 
-As reporters, we usually deal with data that was created in the process of doing something else -- conducting an inspection or paying a parking ticket. These adminsitrative datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. In the sciences, those are often called "administrative records". 
+As reporters, we usually deal with data that was created in the process of doing something else such as conducting an inspection or paying a parking ticket. These datasets are created in government as part of carrying out their work. They form the basis of much investigative reporting and they are often the subject of public records and FOIA requests. They were born as part of the government doing its job, without any thought given to how it might be used in another way. These are often called "administrative records". 
 
 Another type of data might be considered "digital trace" data, which often refers to social media posts, online publications and other items that are born in electronic form and posted publicly for anyone to view and use. 
 
@@ -35,14 +35,11 @@ When someone dies in the US, a standard death certificate is filled out by a ser
 A good reporter could imagine many stories coming out of these little boxes. Limiting yourself to just to COVID-19-related stories: You could profile the local doctor who signed the most COVID-19-related death certificates in their city, or  examine the number of deaths that had COVID as a contributing, but not underlying or immediate, cause of death. Or maybe you would want to map the deaths to find the block in your town most devastated by the virus. 
 
 
+Early in the pandemic, Coulter Jones and Jon Kamp of the Wall Street Journal examined the records from one of the few states that makes them public, and concluded that "[Coronavirus Deaths were Likely Missed in Michigan, Death Certificates Suggest](https://www.wsj.com/articles/coronavirus-deaths-were-likely-missed-in-michigan-death-certificates-suggest-11590073280] ^[Wall Street Journal, May 21, 2020)"
 
 
 ![](assets/images/start-data-def-01.png){.float-md-end .img-fluid  .m-3 width="60%"}
 
-
-Early in the pandemic, Coulter Jones and Jon Kamp of the Wall Street Journal examined the records from one of the few states that makes them public, and concluded that "[Coronavirus Deaths were Likely Missed in Michigan, Death Certificates Suggest](https://www.wsj.com/articles/coronavirus-deaths-were-likely-missed-in-michigan-death-certificates-suggest-11590073280] ^[Wall Street Journal, May 21, 2020)"
-
-
 But you probably can't do that. The reason is that, in most states, death certificates are not public records and are treated as secrets. ^[See "[Secrecy in Death Records: A call to action](https://www.researchgate.net/publication/348322257_Secrecy_in_Death_Records_A_Call_to_Action)", by Megain Craig and Madeleine Davison, Journal of Civic Information, December 2020]. Instead, state and local governments provide limited statistics related to the deaths, usually by county, with no detail.  
 
 
@@ -64,11 +61,11 @@ We often have to consider the trade-offs. Granular data with the detail we need
 
 For our purposes, it's important to remember that we can always create aggregated numbers like the ones shown in the Arizona COVID page out of individual items, but you can never de-aggregate statistics into more granular data like the boxes filled out in death certificates.
 
-## Nouns 
+## Unit of analysis: Know your nouns 
 
 That brings us to one of the most important things you must find out about any data you begin to analyze: What **noun** describes each row?  In statistics, these rows might be called *observations* or *cases*.  In data science, they're usually called *records*. Either way, every row must represent the same thing -- a person, a place, a year, a water sample or a school. And you can't really do anything with it until you figure out what that noun is. 
 
-Sometimes the only noun you can obtain is problematic. In 2015, we did a story at The New York Times called ["More Deportation Follow Minor Crimes, Records Show"](https://www.nytimes.com/2014/04/07/us/more-deportations-follow-minor-crimes-data-shows.html) . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions. 
+Sometimes the only level you can obtain creates problems. In 2015, we did a story at The New York Times called ["More Deportation Follow Minor Crimes, Records Show"](https://www.nytimes.com/2014/04/07/us/more-deportations-follow-minor-crimes-data-shows.html) . The government had claimed it was only removing hardened criminals from the country, but our analysis of the data suggested that many of them were for minor infractions. 
 
 In writing the piece, we had to work around a problem in our data: the agency refused to provide us anything that would help us distinguish one individual from another.  All we knew was that each row represented one deportation, not a person. Without a column that uniquely identified people -- say, name and date of birth, or some scrambled version of an their DHS number --  we had no way to even estimate how often people were deported multiple times. If you read the story, you'll see the very careful wording, except when we had reported out and spoken to people on the ground.