Categorical type #16015

TomAugspurger · 2017-04-16T03:11:31Z

We extended the CategoricalDtype to accept optional categories and ordered
argument. CategoricalDtype is now part of the public API. This allows users to
specify the desired categories and orderedness of an operation ahead of time.
The current behavior, which is still possible with categories=None, the
default, is to infer the categories from whatever is present.

This change will make it easy to implement support for specifying categories
that are know ahead of time in other places e.g. .astype, .read_csv, and the
Series constructor.

Closes #14711
Closes #15078
Closes #14676

Dunno if we would want to do this for 0.20 or not. It ended up not being too big a change, so maybe.

jreback · 2017-04-16T04:23:32Z

.dtype on a Categorical/CI should construct this

Categorica.is_dtype_equal should just work

jreback · 2017-04-16T04:32:47Z

can u run asv on this to make sure nothing much changing perf wise

TomAugspurger · 2017-04-16T15:30:56Z

.dtype on a Categorical/CI should construct this

Ha, that's why it felt like so much less code this time, because I forgot to do the other half :)

I'll work on this a bit more today, but will probably push to 0.21

jreback · 2017-04-16T17:15:55Z

yeah this feels like it should be in 0.21.0. maybe we also expose this in pandas.api.types.CategoricalDtype rather than at the top-level.

jreback · 2017-04-16T18:47:01Z

pandas/core/dtypes/dtypes.py

@@ -131,6 +171,33 @@ def __eq__(self, other):

        return isinstance(other, CategoricalDtype)

+    @staticmethod
+    def _hash_categories(categories):
+        from pandas.tools.hashing import hash_array, _combine_hash_arrays


you can just usehash_pandas_object. Its an ordered hash of the data.

gfyoung · 2017-04-17T03:23:25Z

pandas/tests/core/dtypes/test_dtypes.py

@@ -356,6 +357,70 @@ def test_not_string(self):
        self.assertFalse(is_string_dtype(PeriodDtype('D')))


+class TestCategoricalDtypeParametrized(object):


I think we can just call this TestCategoricalDtype (naming consistency)

gfyoung · 2017-04-17T03:25:48Z

pandas/tests/series/test_constructors.py

+        result = pd.Series(['a', 'b'], dtype=pd.CategoricalDtype(['b', 'a']))
+        assert is_categorical_dtype(result)
+        tm.assert_index_equal(result.cat.categories, pd.Index(['b', 'a']))
+        assert result.cat.ordered is False


What's more idiomatic: assert ... is False or assert not ... ?

if you are asserting that the result is a boolean (False) , then the first is more appropriate. If its just not True then the 2nd is fine (e.g. imagine 0 satisfies 2nd but not 1st). In this case we do want to assert a boolean result.

gfyoung · 2017-04-17T03:26:38Z

pandas/tests/core/dtypes/test_dtypes.py

+    def test_categories(self):
+        result = CategoricalDtype(['a', 'b', 'c'])
+        tm.assert_index_equal(result.categories, pd.Index(['a', 'b', 'c']))
+        assert result.ordered is False


Same comment as here

xref pandas-dev#16015 xref apache/arrow#585

* API: expose dtypes in pandas.api.types xref #16015 xref apache/arrow#585 xref #16042 xref #15541 (comment)

* API: expose dtypes in pandas.api.types xref pandas-dev#16015 xref apache/arrow#585 xref pandas-dev#16042 xref pandas-dev#15541 (comment)

pep8speaks · 2017-07-17T03:21:33Z

Hello @TomAugspurger! Thanks for updating the PR.

Cheers ! There are no PEP8 issues in this Pull Request. 🍻

Comment last updated on September 23, 2017 at 16:34 Hours UTC

TomAugspurger · 2017-08-12T11:16:49Z

pandas/tests/dtypes/test_dtypes.py

@@ -618,3 +618,8 @@ def test_categorical_equality(self, ordered, other, expected):
        c1 = CategoricalDtype(['a', 'b'], ordered)
        result = c1 == other
        assert result == expected
+
+    def test_mixed(self):


This is going to fail at the moment. hash_object_array tries to work on the mixed array first, and if that fails it falls back to hash_object_array(a.astype(str)), which will cause a "collision" with the section set of categoricals. Not sure how I'll handle this yet.

jreback

looks pretty good. I don't like the MultiIndex change (already commened on that PR). instead I would simply just change it (rather than adding a kw); the small api break is fine.

jreback · 2017-08-12T17:47:56Z

doc/source/categorical.rst

+CategoricalDtype
+----------------
+
+.. versionadded:: 0.21.0


versionchanged; it exists now.

jreback · 2017-08-12T17:49:32Z

pandas/core/api.py

@@ -6,6 +6,7 @@

 from pandas.core.algorithms import factorize, unique, value_counts
 from pandas.core.dtypes.missing import isna, isnull, notna, notnull


I don't think exposing this is a good idea here. We don't directly do this with other dtypes. instead the are in pandas.api.types

Hmm, I could go either way on this. It would be the only type at the top level. On the other hand pd.api.types.CategoricalDtype(values) is quite long :)

I'll think about it some more, but you're probably right.

jreback · 2017-08-12T17:53:58Z

pandas/core/dtypes/common.py

-        dtype = CategoricalDtype()
+        categories = getattr(dtype, 'categories', None)
+        ordered = getattr(dtype, 'ordered', False)
+        dtype = CategoricalDtype(categories=categories, ordered=ordered)


this is not really necessary if you allow CategoricalDtype(dtype) IOW, you can check whether the incoming categories is an instance of a CagetegoricalDtype and act accordingly.

I'm not sure I like overloading the categories like that.

jreback · 2017-08-12T17:54:16Z

pandas/core/dtypes/dtypes.py

+    Type for categorical data with the categories and orderedness,
+    but not the values.
+
+    .. versionadded:: 0.20.0


versionchanged

jreback · 2017-08-12T17:54:37Z

pandas/core/dtypes/dtypes.py

+    Examples
+    --------
+    >>> t = CategoricalDtype(categories=['b', 'a'], ordered=True)
+    >>> s = Series(['a', 'a', 'b', 'b', 'a'])


I wouldn't show anything not related to the actual dtype here.

jreback · 2017-08-12T17:55:09Z

pandas/core/dtypes/dtypes.py

+
+    def __new__(cls, categories=None, ordered=False, fastpath=False):
+        from pandas.core.indexes.base import Index
+        if categories is not None:


do the import inside the function

jreback · 2017-08-12T17:55:22Z

pandas/core/dtypes/dtypes.py

+            # validation
+            cls._validate_categories(categories, fastpath=fastpath)
+            cls._validate_ordered(ordered)
+            # We have a choice when hashing pandas unordered categoricals


use blank lines to indicate different thoughts

jreback · 2017-08-12T17:56:03Z

pandas/core/dtypes/dtypes.py

+
+    def __new__(cls, categories=None, ordered=False, fastpath=False):
+        from pandas.core.indexes.base import Index
+        if categories is not None:


categories could be a CategoricalDtype (see my comment above)

jorisvandenbossche · 2017-08-18T22:26:39Z

@TomAugspurger just to check: this is ready for review?

TomAugspurger · 2017-09-15T21:14:04Z

@jorisvandenbossche Agreed about not having a fastpath in the public API. see 91752fe

Initially we used __new__ for initialization, since we cached the instance (even though it didn't take parameters). That commit cleans things up a bit.

jreback

look good. a few changes.

jreback · 2017-09-17T14:28:48Z

doc/source/api.rst

@@ -646,7 +646,10 @@ strings and apply several methods to it. These can be accessed like
 Categorical
 ~~~~~~~~~~~

-If the Series is of dtype ``category``, ``Series.cat`` can be used to change the the categorical
+.. autoclass:: api.types.CategoricalDtype
+   :members: categories, ordered


yep, we should use the auto* things more readily in other places. Maybe make an issue about this.

I don't think we in general need to do this, as for most functions/methods, we already have the generated pages to link to

I just did this instead of autosummary since there are a bunch of unrelated methods that are just there for NumPy duck-typing.

jreback · 2017-09-17T14:30:02Z

doc/source/categorical.rst

+
+A categorical's type is fully described by
+
+1. its categories: a sequence of unique values and no missing values


maybe show as

1. ``categories``: .... 2. ``ordered``: .....

as these are the actual kwargs

jreback · 2017-09-17T14:30:42Z

doc/source/categorical.rst

+expects a `dtype`. For example :func:`pandas.read_csv`,
+:func:`pandas.DataFrame.astype`, or in the Series constructor.
+
+As a convenience, you can use the string ``'category'`` in place of a


maybe put this in a warning/note as this is the backward compat aspect here. (and then refer to this in the whatsnew)

what is back incompat about doing dtype='category' ?

I think Jeff was saying that was the only way to do things in the past (backwards compatible). But I don't want to give the impression that we're thinking about deprecating dtype='category'.

Yes, but you didn't do that actual deprecation (or not yet)?

Yes, but you didn't do that actual deprecation (or not yet)?

I don't think this is desirable, 'category' is a nice convenience

Sorry, I was speaking about astype(dtype='category', categories=[...]). I suppose I misread Jeff's comment (as it was about backwards compatibility, I thought it was about that part, as this is the only part that might be back incompat, if we deprecate categories=.. (not dtype='category')

jreback · 2017-09-17T14:31:21Z

doc/source/whatsnew/v0.21.0.txt

@@ -10,6 +10,8 @@ users upgrade to this version.
 Highlights include:

 - Integration with `Apache Parquet <https://parquet.apache.org/>`__, including a new top-level :func:`read_parquet` and :func:`DataFrame.to_parquet` method, see :ref:`here <io.parquet>`.
+- New user-facing :class:`pandas.api.types.CategoricalDtype` for specifying
+  categoricals independent of the data (:issue:`14711`, :issue:`15078`)


I would add the link the sub-section (and then list the issues there instead)

jreback · 2017-09-17T14:31:31Z

doc/source/whatsnew/v0.21.0.txt

+expanded to include the ``categories`` and ``ordered`` attributes. A
+``CategoricalDtype`` can be used to specify the set of categories and
+orderedness of an array, independent of the data themselves. This can be useful,
+e.g., when converting string data to a ``Categorical``:


jreback · 2017-09-17T14:39:29Z

pandas/core/dtypes/dtypes.py

+            data = self.categories._format_data(name=self.__class__.__name__)
+        return tpl.format(data, self.ordered)
+
+    def __repr__(self):


jreback · 2017-09-17T14:39:38Z

pandas/core/dtypes/dtypes.py

+            categories = list(categories)  # breaks if a np.array of categories
+            cat_array = hash_tuples(categories)
+        else:
+            if categories.dtype == 'O':


jreback · 2017-09-17T14:40:49Z

pandas/core/indexes/category.py

+        # we want to reuse self.dtype if possible, i.e. neither are
+        # overridden.
+        if dtype is not None and (categories is not None or
+                                  ordered is not None):


jreback · 2017-09-17T14:42:27Z

pandas/tests/dtypes/test_common.py

+    (pd.Categorical(['a', 'b']), CategoricalDtype(['a', 'b'])),
+    (pd.CategoricalIndex(['a', 'b']).dtype, CategoricalDtype(['a', 'b'])),
+    (pd.CategoricalIndex(['a', 'b']), CategoricalDtype(['a', 'b'])),
+    (CategoricalDtype(), CategoricalDtype()),


leave original pd.Categorical(['a', 'b']).dtype, CategoricalDtype()) which should still work

jreback · 2017-09-17T14:43:48Z

pandas/tests/dtypes/test_dtypes.py

+        assert result == expected
+
+    def test_invalid_raises(self):
+        with tm.assert_raises_regex(TypeError, 'ordered'):


might add a test for passing just a string for the first argument (should fail)
e.g.
CategoricalDtype('category') (see my comment about a from_string above, which we should add for compat.

TomAugspurger · 2017-09-18T10:59:53Z

Correct, and I don't plan to deprecate `dtype'category'`.

…

On Mon, Sep 18, 2017 at 5:55 AM, Joris Van den Bossche < ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In doc/source/categorical.rst <#16015 (comment)>: > +should be inferred from whatever is present in the data when the +:class:`pandas.Categorical` is created. + +.. ipython:: python + + from pandas.api.types import CategoricalDtype + + CategoricalDtype(['a', 'b', 'c']) + CategoricalDtype(['a', 'b', 'c'], ordered=True) + CategoricalDtype() + +A :class:`~pandas.api.types.CategoricalDtype` can be used in any place pandas +expects a `dtype`. For example :func:`pandas.read_csv`, +:func:`pandas.DataFrame.astype`, or in the Series constructor. + +As a convenience, you can use the string ``'category'`` in place of a Yes, but you didn't do that actual deprecation (or not yet)? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#16015 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHItygdUADe7vz66Heen2kIDzUzdl9ks5sjkwVgaJpZM4M-hoW> .

jreback · 2017-09-18T13:51:11Z

Yes, but you didn't do that actual deprecation (or not yet)?

I don't think this is desirable, 'category' is a nice convenience

We extended the CategoricalDtype to accept optional categories and ordered argument. ```python pd.CategoricalDtype(categories=['a', 'b'], ordered=True ``` CategoricalDtype is now part of the public API. This allows users to specify the desired categories and orderedness of an operation ahead of time. The current behavior, which is still possible with categories=None, the default, is to infer the categories from whatever is present. This change will make it easy to implement support for specifying categories that are know ahead of time in other places e.g. .astype, .read_csv, and the Series constructor. Closes pandas-dev#14711 Closes pandas-dev#15078 Closes pandas-dev#14676

Get a valid instance of `CategoricalDtype` as early as possible, and use that throughout.

TomAugspurger · 2017-09-20T15:56:49Z

416d1d7 has the changes to Categorical.__init__ that you requested @jreback. It's a bit messy since there are basically 3 ways for specifying the categories

passing dtype=CategoricalDtype(...)
passing categories=...
passing a Categorical for values

I think all three are useful and (hopefully) well-tested now. I'll update the docstring to indicate the priority (basically moving that comment to the docstring)

jreback

just a couple of comments, which can be put in a followup. let's merge on green.

jreback · 2017-09-23T00:14:33Z

doc/source/categorical.rst

+
+   Since ``dtype='category'`` is essentially ``CategoricalDtype(None, False)``,
+   and since all instances ``CategoricalDtype`` compare equal to ``'`category'``,
+   all instances of ``CategoricalDtype`` compare equal to a ``CategoricalDtype(None)``


so you should say CategoricalDtype(None, False) at the end.

jreback · 2017-09-23T00:22:18Z

pandas/core/dtypes/dtypes.py

+            categories = list(categories)  # breaks if a np.array of categories
+            cat_array = hash_tuples(categories)
+        else:
+            if categories.dtype == 'O':


neh hash collisions basically don't happen in the space of things we are doing (normally). however you can collide by construction, e.g. code doing something wrong..

In any event we should move this code (can be a followup, maybe make an issue for things to do)

jreback · 2017-09-23T00:24:51Z

pandas/core/dtypes/dtypes.py

+            return hash(self) == hash(other)
+
+    def __unicode__(self):
+        tpl = u'CategoricalDtype(categories={}ordered={})'


we have this hacky thing in Categorical to repr only 10 categories. CategoricalIndex actually correctly uses the option max_categories, so should do that here (again can do as a followup)

jreback · 2017-09-23T00:28:38Z

pandas/tests/test_categorical.py

+        tm.assert_index_equal(result.categories, expected)
+        assert result.ordered is False
+
+    def test_constructor_tuples_datetimes(self):


this is janky. I don't think we should actually allow it. Where did this example come from? So we allow tuples as categories, I guess so.

This was extracted from a groupby test that was failing when I improperly handled tuples of level-1. Agreed that it's weird, but I'm inclined to support it.

jreback · 2017-09-23T00:31:31Z

doc/source/whatsnew/v0.21.0.txt

+expanded to include the ``categories`` and ``ordered`` attributes. A
+``CategoricalDtype`` can be used to specify the set of categories and
+orderedness of an array, independent of the data themselves. This can be useful,
+e.g., when converting string data to a ``Categorical`` (:issue:`14711`, :issue:`15078`):


there is 1 more issue you listed in the top of the PR

TomAugspurger · 2017-09-23T11:09:58Z

Thanks. I fixed up the doc comments. I'll merge on green and do the other followups later today.

…

On Fri, Sep 22, 2017 at 7:31 PM, Jeff Reback ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In doc/source/whatsnew/v0.21.0.txt <#16015 (comment)>: > @@ -89,6 +91,30 @@ This does not raise any obvious exceptions, but also does not create a new colum Setting a list-like data structure into a new attribute now raise a ``UserWarning`` about the potential for unexpected behavior. See :ref:`Attribute Access <indexing.attribute_access>`. +.. _whatsnew_0210.enhancements.categorical_dtype: + +``CategoricalDtype`` for specifying categoricals +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +:class:`pandas.api.types.CategoricalDtype` has been added to the public API and +expanded to include the ``categories`` and ``ordered`` attributes. A +``CategoricalDtype`` can be used to specify the set of categories and +orderedness of an array, independent of the data themselves. This can be useful, +e.g., when converting string data to a ``Categorical`` (:issue:`14711`, :issue:`15078`): there is 1 more issue you listed in the top of the PR — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#16015 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIlwjFqIXM4vMTgsVZ20tYOzp9CQSks5slFFogaJpZM4M-hoW> .

jreback · 2017-09-23T18:56:23Z

nice patch @TomAugspurger !

Closes pandas-dev#14711 Closes pandas-dev#15078 Closes pandas-dev#14676

jreback reviewed Apr 16, 2017

View reviewed changes

gfyoung reviewed Apr 17, 2017

View reviewed changes

jreback added the Categorical Categorical Data Type label Apr 17, 2017

TomAugspurger added this to the 0.21.0 milestone Apr 17, 2017

jorisvandenbossche mentioned this pull request Apr 18, 2017

API: scope of the public types module #16042

Closed

jreback added a commit to jreback/pandas that referenced this pull request Apr 23, 2017

API: expose dtypes in pandas.api.types

eb775e6

xref pandas-dev#16015 xref apache/arrow#585

jreback mentioned this pull request Apr 23, 2017

API: expose dtypes in pandas.api.types #16099

Merged

jreback added a commit to jreback/pandas that referenced this pull request Apr 24, 2017

API: expose dtypes in pandas.api.types

e6a927c

xref pandas-dev#16015 xref apache/arrow#585

jreback added a commit that referenced this pull request Apr 24, 2017

API: expose dtypes in pandas.api.types (#16099)

844013b

* API: expose dtypes in pandas.api.types xref #16015 xref apache/arrow#585 xref #16042 xref #15541 (comment)

TomAugspurger force-pushed the categorical-type branch from 5a7c498 to d10db1a Compare May 19, 2017 18:27

TomAugspurger mentioned this pull request May 20, 2017

Categorical handles imaginary values incorrectly #16399

Closed

pcluo pushed a commit to pcluo/pandas that referenced this pull request May 22, 2017

API: expose dtypes in pandas.api.types (pandas-dev#16099)

a4709e2

* API: expose dtypes in pandas.api.types xref pandas-dev#16015 xref apache/arrow#585 xref pandas-dev#16042 xref pandas-dev#15541 (comment)

TomAugspurger force-pushed the categorical-type branch from d10db1a to b731138 Compare May 30, 2017 16:43

TomAugspurger mentioned this pull request Jun 13, 2017

BUG: Fix Categorical comparsion with Series of dtype 'category' #16667

Closed

4 tasks

chris-b1 mentioned this pull request Jun 21, 2017

API: provide a to_enum method for Categorical dtypes? #16749

Closed

TomAugspurger force-pushed the categorical-type branch from b731138 to 04a77d3 Compare July 17, 2017 03:21

TomAugspurger mentioned this pull request Aug 10, 2017

DISCUSSION: Add format parameter to .astype when converting to str dtype #17211

Open

TomAugspurger force-pushed the categorical-type branch from 04a77d3 to b79ae6e Compare August 12, 2017 11:09

TomAugspurger commented Aug 12, 2017

View reviewed changes

jreback requested changes Aug 12, 2017

View reviewed changes

TomAugspurger force-pushed the categorical-type branch 2 times, most recently from d5eb041 to f069158 Compare August 18, 2017 17:39

TomAugspurger force-pushed the categorical-type branch from 7ef02b4 to 91752fe Compare September 15, 2017 21:11

TomAugspurger force-pushed the categorical-type branch from 8364e7c to cd5f4e2 Compare September 17, 2017 12:05

jreback requested changes Sep 17, 2017

View reviewed changes

TomAugspurger mentioned this pull request Sep 17, 2017

API: Deprecate Categorical fastpath argument #17562

Closed

TomAugspurger added 3 commits September 19, 2017 15:53

update

ed5c814

Consistent CategoricalDtype use in Categorical init

416d1d7

Get a valid instance of `CategoricalDtype` as early as possible, and use that throughout.

TomAugspurger force-pushed the categorical-type branch from a9e0538 to 416d1d7 Compare September 20, 2017 15:48

PEP8 fixes

e6c05a0

jreback approved these changes Sep 23, 2017

View reviewed changes

jreback reviewed Sep 23, 2017

View reviewed changes

jreback mentioned this pull request Sep 23, 2017

DEPR: .astype(..., categories=) #17636

Closed

Final doc fixups

41172ce

TomAugspurger mentioned this pull request Sep 23, 2017

Accept CategoricalDtype in read_csv #17643

Merged

TomAugspurger added 2 commits September 23, 2017 07:56

Fixup set_categories inplace test

141e509

PEP8

43f90cc

jreback added the Dtype Conversions Unexpected or buggy dtype conversions label Sep 23, 2017

jreback mentioned this pull request Sep 23, 2017

WIP: prototype for unit support #10349 #17153

Closed

4 tasks

jreback merged commit e57f189 into pandas-dev:master Sep 23, 2017

TomAugspurger deleted the categorical-type branch September 24, 2017 14:00

alanbato pushed a commit to alanbato/pandas that referenced this pull request Nov 10, 2017

Categorical type (pandas-dev#16015)

da7ad15

Closes pandas-dev#14711 Closes pandas-dev#15078 Closes pandas-dev#14676

No-Stream pushed a commit to No-Stream/pandas that referenced this pull request Nov 28, 2017

Categorical type (pandas-dev#16015)

3bb7929

Closes pandas-dev#14711 Closes pandas-dev#15078 Closes pandas-dev#14676

jbrockmendel mentioned this pull request Apr 20, 2020

REF: remove single-tuple special case for Categorical.__hash__ #33678

Closed

jameslamb mentioned this pull request Jun 23, 2023

[python-package] add 'pandas' extra microsoft/LightGBM#5937

Merged

This pull request was closed.

		@@ -356,6 +357,70 @@ def test_not_string(self):
		self.assertFalse(is_string_dtype(PeriodDtype('D')))


		class TestCategoricalDtypeParametrized(object):

		@@ -6,6 +6,7 @@

		from pandas.core.algorithms import factorize, unique, value_counts
		from pandas.core.dtypes.missing import isna, isnull, notna, notnull


		A categorical's type is fully described by

		1. its categories: a sequence of unique values and no missing values

Categorical type #16015

Categorical type #16015

Conversation

TomAugspurger commented Apr 16, 2017

jreback commented Apr 16, 2017

jreback commented Apr 16, 2017

TomAugspurger commented Apr 16, 2017

jreback commented Apr 16, 2017

Choose a reason for hiding this comment

gfyoung Apr 17, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gfyoung Apr 17, 2017 • edited Loading

Choose a reason for hiding this comment

pep8speaks commented Jul 17, 2017 • edited Loading

Comment last updated on September 23, 2017 at 16:34 Hours UTC

Choose a reason for hiding this comment

jreback left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jorisvandenbossche commented Aug 18, 2017

TomAugspurger commented Sep 15, 2017

jreback left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TomAugspurger commented Sep 18, 2017 via email

jreback commented Sep 18, 2017

TomAugspurger commented Sep 20, 2017

jreback left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TomAugspurger commented Sep 23, 2017 via email

jreback commented Sep 23, 2017

gfyoung Apr 17, 2017 •

edited

Loading

gfyoung Apr 17, 2017 •

edited

Loading

pep8speaks commented Jul 17, 2017 •

edited

Loading