Skip to content


Subversion checkout URL

You can clone with
Download ZIP
create static websites from YAML data and string templates
Haskell CSS HTML Other
Fetching latest commit...
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


yst - static websites from YAML and string templates

yst is a tool for generating a static website by filling string templates with data taken from YAML or CSV text files or SQLite3 file based databases. This approach combines the speed, security, and ease of deployment of a static website with the flexibility and maintainability of a dynamic site that separates presentation and data.

Installing yst

yst is written in Haskell. The easiest way to install yst is by using Haskell's cabal install tool. The best way to get this tool is to install the Haskell platform, which includes a complete installation of the GHC compiler and the cabal executable.

Once you have cabal, you can install yst with two commands:

cabal update
cabal install yst

(Note that by default, cabal installs the yst executable into a special directory: ~\.cabal\bin on unix systems. You will need to make sure that directory is in your system path.)

If you get the error "yst-0.2.3 depends on template-haskell- which failed to install," then try the following:

cabal install syb-with-class-0.6
cabal install yst

Getting started

To get started with yst, use the command:

yst create mysite

This will create a directory mysite and populate it with the files needed for a sample site. Change to this directory and run yst with no arguments to create the site:

cd mysite

The site will be created in the site directory. Open up site/index.html to take a look.

The configuration file index.yaml tells yst which pages to build, and from which templates and data files. Let's take a look, so we can see what went into index.html. The file is a YAML list of YAML hashes (name/value pairs). The first item is

- url      : index.html
  title    : Home
  template :
  requires :
  data     :
    recentevents : FROM events.yaml ORDER BY date DESC LIMIT 2

This says: build the page index.html from the string template (and subtemplate and data from events.yaml. Sort this data (a list of events) by date in descending order, and discard all but the first two items. Put the resulting data in the string template attribute recentevents. Give the page the title "Home."

Now take a look at events.yaml, the data source. Again it is a YAML list of YAML hashes, each item in the list representing one event. The first is:

- date: 2009-06-28
  speaker:  Sam Smith
  title: Building a static website

Pretty self-explanatory! Try adding a new event, then rebuild the site by typing yst and see how it looks.

If you have problems, make sure events.yaml is a valid YAML file. Structure is shown through indentation, so make sure things line up right. And occasionally you may need to use quotation marks around string values---for example, when a title contains a colon.

There's one more ingredient in our recipe---the string templates. Take a look at, the template from which index.html is built:

# Welcome

Here's our website. Have a look around.

Our last two events:


For a complete list of events, see the [events](events.html) page.

The first thing to notice is that this is in markdown format (or, to be precise, markdown with pandoc extensions). So, for example, the first line is a level-one header, and there is a hyperlink to the events page on the last line.

The second thing to notice is the bit between dollar signs. This is a placeholder for some formatted data. The rendered template will include the list recentevents (remember, this was populated from events.yaml after some transformations---see above). And each element of this list will be formatted by a subtemplate called

- $if(it.speaker)$$it.speaker; separator=" and "$, $endif$ *$it.title$*.

Let's break this down. The whole line will print a bulleted list item. it here refers to the event that is being processed by the template. So the first part says: if this event has a speaker field, print the speaker, or, if the speaker field is a list, print all the speakers separated by the word "and", followed by a comma. And the second part says, print the contents of the title field, surrounding it with asterisks (which is the markdown way of making it emphasized).

(See the string template documentation for details of template syntax, and examples.)

If you look at index.html, you'll see this rough structure, but in an HTML file, not a markdown file. yst converts the rendered markdown template to HTML (using pandoc), and inserts it into a "layout" file called If you look at this file, you'll see that it's an HTML file with blanks for $contents$ and $nav$. The $contents$ blank gets filled by the rendered template, converted to HTML, and the $nav$ blank gets filled by an HTML navigation menu (an unordered list with links).

To summarize our example, then: yst sorts and filters the data in events.yaml according to our instructions; inserts this data into the template, formatting each item using the template; uses pandoc to convert the rendered template to HTML; constructs a navigation menu; and puts the contents and navigation menu in the layout template The result is our page, index.html.


The yst command


yst                    # rebuilds site, using default config.yaml
yst -f myconf.yaml     # rebuilds site, using myconf.yaml as config
yst create newsite     # creates a starter (demo) site in newsite directory

When run without arguments, yst looks at index.yaml to determine the dependencies of each page, and rebuilds only the pages whose dependencies have changed since the last build.

In order for this to work properly, you must be sure to list all subtemplates included recursively in the main page template using the requires field. This field takes a single filename or a YAML list, which may be in either of two formats:

requires: [,]



If you don't list all the subtemplates needed to render a page under requires, yst will still work, but it might sometimes fail to rebuild a page when one of these subtemplates has been changed.


The configuration file specifies the following:

  • indexfile: the filename of the index file (default: index.yaml)
  • title: the title of the whole site
  • sourcedir: the directory containing all the templates and page sources (default: .)
  • datadir: the directory containing yaml data files (default: .)
  • filesdir: the directory containing static files (default: files)
  • layout: the default layout template for the site, relative to sourcedir (default:

index.yaml and submenus

The index file is a YAML list of pages. Each page may have the following fields:

  • url: the relative URL of the page to be built
  • title: the title of the page
  • template: the main string template from which the page will be built
  • source: the markdown source from which the page will be built
  • requires: other files changes to which should trigger a page rebuild (primarily subtemplates of the main page template)
  • data: string template attributes, data sources, and transformations (see below)
  • layout: a layout template to use, if other than the site default
  • inmenu: if 'no', the page will not be included in the site navigation menu

Each page must have at least url, title, and either template or source. Values for template, source, and layout are relative to sourcedir specified in config.yaml.

The pages may be organized into a tree-like hierarchy, which will be reflected in the site navigation menu. It is easiest to see how this works by example:

- Rooms:
  - url      : room101.html
    title    : Room 101
    source   : room101.txt

  - url      : room203.html
    title    : Room 203
    source   : room203.txt

Here we have a subtree called "Rooms" with two pages under it. Subtrees can contain other subtrees. Just be consistent about indentation.

The data field

The data field in index.yaml can populate any number of stringtemplate attributes with data from YAML or CSV files or SQLite3 databases. The syntax is easiest to explain by example (note that the keywords do not have to be in ALL CAPS, although they may, and the query doesn't have to end with a semicolon, though it may):

    events:  from events.yaml order by date desc group by title then location
    people:  from people.csv order by birthday then lastname where
              birthstate = 'CA' limit 5
    beststudents: from students.sqlite 
              query "select * from students where grade > 5"
              order by name

First we have the name of the stringtemplate attribute to be populated (say, events). Then, after the colon, we have the data source (events.yaml). If the data source is an SQLite3 database, it should be followed by a query that is a quoted string.

The data source is followed by one or more transformations, which will be applied in order. Here are the possible transformations. In what follows, brackets denote an optional component, | denotes alternatives, and * indicates that the component may be repeated several times:

ORDER BY field [ASC|DESC] [THEN field [ASC|DESC]]*

Sorts a list by comparing the value of field. ASC (the default) means "ascending", and DESC means "descending". The keyword THEN is used to separate fields that will be compared in order. So, if we are ordering by birthday then lastname, we will compare birthdays, and if these are equal, we will break the tie by comparing last names.

GROUP BY field [THEN field]*

Converts a list into a list of lists, where each sublist contains only items with the same value for field. So, for example, group by date takes a list of events and produces a list of lists of items, where each sublist contains events occuring at a single date. GROUP BY date THEN venue would produce a list of lists of lists, and so on.


Removes all but the n top items from a list. n must be a number.

WHERE condition

Selects only items that meet a condition.

A condition in a WHERE statement is a Boolean combination (using NOT, AND, OR, and parentheses for disambiguation) of basic conditions. A basic condition is of the form value op value, where value may be either a fieldname or a constant. Note that all constants must be enclosed in quotes. op may be one of the following: =, >=, <=, >, < and 'contains'.

The basic condition arg1 contains arg2 succeeds if and only if arg1 is a fieldname whose value is a list containing the value of arg2.

Note that the order of transformations is significant. You can get different results if you use LIMIT before or after ORDER BY, for example.

If you want to specify an attribute's value directly, rather than reading it from a file, just omit the "FROM":

  deadline: 11/20/2009

Any YAML value can be given to an attribute in this way.

Static files

Any file or subdirectory in the files directory (or whatever is the value of filesdir in config.yaml) will be copied verbatim to the site. So this is the place to put javascripts, css files, images, PDFs, and the like.

Date fields

yst will recognize date fields in data files automatically, if the dates are in one of the following formats:

  • the locale's standard date format
  • MM/DD/YYYY (e.g. 04/28/1953)
  • MM/DD/YY (e.g. 04/28/53)
  • YYYY-MM-DD (e.g. 1953-04-28)
  • DD MON YYYY (e.g. 28 Apr 1953)

Dates may be formatted in templates using a stringtemplate "format" directive. There's an example in the demo file

$it; format="%B %d, %Y"$

The following codes may be used in format strings (taken from Haskell's Date.Time.Format documentation):

  • %D : same as %m/%d/%y
  • %F : same as %Y-%m-%d
  • %x : as dateFmt locale (e.g. %m/%d/%y)
  • %Y : year
  • %y : last two digits of year, 00 - 99
  • %C : century (being the first two digits of the year), 00 - 99
  • %B : month name, long form (fst from months locale), January - December
  • %b, %h : month name, short form (snd from months locale), Jan - Dec
  • %m : month of year, leading 0 as needed, 01 - 12
  • %d : day of month, leading 0 as needed, 01 - 31
  • %e : day of month, leading space as needed, 1 - 31
  • %j : day of year for Ordinal Date format, 001 - 366
  • %G : year for Week Date format
  • %g : last two digits of year for Week Date format, 00 - 99
  • %f : century (first two digits of year) for Week Date format, 00 - 99
  • %V : week for Week Date format, 01 - 53
  • %u : day for Week Date format, 1 - 7
  • %a : day of week, short form (snd from wDays locale), Sun - Sat
  • %A : day of week, long form (fst from wDays locale), Sunday - Saturday
  • %U : week number of year, where weeks start on Sunday (as sundayStartWeek), 00 - 53
  • %w : day of week number, 0 (= Sunday) - 6 (= Saturday)
  • %W : week number of year, where weeks start on Monday (as mondayStartWeek), 00 - 53

Lists as values

In some cases, a field may have one or several values. For example, an event might occur at a date or a date range, and an article may have one author or a list of authors.

An elegant way to handle these cases is to let the field take either a scalar or a list value, and use stringtemplate's "separator" directive to format the result appropriately. So, for example, in our events.yaml we have:

- date: 2009-06-28
  speaker:  Sam Smith
  title: Building a static website

- date:  2009-04-15
    - Sam Smith
    - '[Jim Jones]('
  title: Advantages of static websites

- date:
    - 2009-04-20
    - 2009-04-22
  title: Seminar on web security

- date: 2009-04-15
  speaker: Jim Jones
  title:  XSS attacks for dummies

Note that the date field is sometimes a single date, sometimes a list (with start and end date of a range), and the speaker field is sometimes a single speaker, and sometimes a range.

Here is how we handle the date in

**$first(it).date:date(); separator=" - "$**

Here first(it).date is the raw data, which may be a single date or a list. first(it).date:date() is the result of formatting each date using the template (discussed above). And first(it).date:date; separator=" - " is the result of taking this list of formatted dates and concatenating them, separated by a hyphen. When there is just one date, we just get a date. When there are two, we get a date range.

We can use the same trick in the case of speaker. If it is an event record, then it.speaker; separator=" and " will be either a single speaker (if the value is not a list) or a list of speakers separated by "and" (if it is a list).

In sorting lists with order by, yst compares two lists by comparing the first members, then (in case of a tie) the second members, and so on. If one item is a list and the other a scalar, the scalar is compared to the first item of the list. So, in the example above, Seminar on web security will be sorted an earlier than an event with date 2009-04-21, and later than an event with date range 2009-04-20 - 2009-04-21.

YAML gotchas

If you have a colon in a YAML value, be sure to enclose it in quotes, or you'll get an error. So,

title:  "Cheever: A Life"


title:  Cheever: A Life

Or (especially if the string is long), use > or | for a wrapped or unwrapped multiline string literal:

title: |
  A very long string that
  goes on and on.

  You can even have blank lines,
  but be sure to maintain indentation.

Using CSV files instead of YAML

If you like, you can use a CSV file instead of YAML for your data source. Just give it the extension .csv. In index.yaml, you'd have:

  events: from events.csv order by date desc

This can be handy if you're using existing data, because spreadsheets and databases can easily be dumped to CSV. In the case of a SQL database, you can use a query like this to get the CSV:

SELECT * INTO OUTFILE 'result.csv'
FROM my_table;

(Thanks to

Using a SQLite database as data source

You can also get the data directly from a database. Just give the file name from the database followed by a query quoted in "'s

In this way you can do joins and other advanced operations on your data before handing them over to yst:

  meetings : FROM data.sqlite 
             QUERY "select * from meetings 
               left outer join persons
               on meetings.speaker ="

Using HTML in the templates

Markdown allows raw HTML to be used, so you can embed HTML in templates. Pandoc's extended markdown is different from standard markdown in that it parses text within HTML block elements as markdown. So, for example, you can include a section in <div> tags, or use raw <ul> and <li> tags instead of markdown formatting.

Layout templates

Layout files are also string templates, but they are not treated as markdown by default. They should use a double extension to indicate the format. So, for example, an HTML layout could be, and a LaTeX layout could be yst will convert the page contents appropriately for the format of the layout template. Here are the supported formats and extensions:

  • HTML:,
  • LaTeX:,
  • ConTeXt:
  • Groff man:
  • Rich text format:
  • Texinfo:
  • DocBook:
  • OpenDocument XML:
  • Plain text (markdown):,

The demo site shows how you can use yst to produce a LaTeX document from the same data sources you use to produce HTML pages.

The following stringtemplate attributes are defined when layouts are rendered:

  • $contents$: the result of rendering the page and converting to the layout's format
  • $nav$: an HTML navigation menu created from index.yaml
  • $gendate$: the date the page was generated
  • $sitetitle$: the site title from config.yaml
  • $pagetitle$: the page title as defined in index.yaml
  • $root$: the path to the website's root, relative to the page being rendered. So, for example, if we are rendering rooms/room503.html, $root$ will have the value ../. Put $root$ in front of relative URLs in your layout file, so that the links aren't broken on pages in subdirectories.

Previewing a site

If you use only relative URLs in your site, you can preview it by opening any of the HTML files in site in your web browser. If you use absolute links, this won't work, but you can use Jinjing Wang's simple static web server maid:

cabal update
cabal install maid

To use maid to preview your site, just change to the site directory and start maid:

cd site

The site will appear at http://localhost:3000. If you want to serve it at another port, just pass the port number as an argument to maid:

maid 5999


Source code

yst's source code lives on github at You can clone the repository with


To install the development code once you've checked it out, just do

cabal install

(But please stick to the released version if you don't like things to break unexpectedly!)

Reporting bugs

If you find a bug, please report it using the issue tracker on yst's github page.

Something went wrong with that request. Please try again.