BUG: Fix some PeriodIndex resampling issues #16153

winklerand · 2017-04-26T20:46:01Z

closes #15944
xref partially #12884
closes #13083
closes #13224

This PR addresses some of the issues related to PeriodIndex resampling.
As I'm new to the pandas codebase, I appreciate any advice.

jreback · 2017-04-26T22:56:10Z

pandas/tests/test_resample.py

-
-        with pytest.raises(NotImplementedError):
-            df.resample('2D', level='d')
+        for freq in ['H', '12H', '2D', 'W']:


these tests are more clear with parametrize
but we still inherit from TestCase so can't change this

but you can start a new test class that does this

again if it's more clear

not sure if I understood correctly - pushed 48e324f as an example, are you thinking along those lines of using pytest fixtures/mark.parametrize decorators?

commented directly on the commit

jreback · 2017-04-27T12:35:17Z

lmk when you need review.

codecov · 2017-04-27T20:47:45Z

Codecov Report

Merging #16153 into master will decrease coverage by 0.04%.
The diff coverage is 97.61%.

@@            Coverage Diff             @@
##           master   #16153      +/-   ##
==========================================
- Coverage   91.27%   91.23%   -0.05%     
==========================================
  Files         163      163              
  Lines       49765    49774       +9     
==========================================
- Hits        45421    45409      -12     
- Misses       4344     4365      +21

Flag	Coverage Δ
#multiple	`89.02% <97.61%> (-0.03%)`	⬇️
#single	`40.33% <7.14%> (-0.07%)`	⬇️

Impacted Files	Coverage Δ
pandas/core/resample.py	`96.05% <97.61%> (-0.12%)`	⬇️
pandas/io/gbq.py	`25% <0%> (-58.34%)`	⬇️
pandas/plotting/_converter.py	`63.38% <0%> (-1.82%)`	⬇️
pandas/compat/numpy/function.py	`93.12% <0%> (-0.21%)`	⬇️
pandas/core/frame.py	`97.73% <0%> (-0.1%)`	⬇️
pandas/core/sparse/array.py	`91.58% <0%> (-0.01%)`	⬇️
pandas/core/indexes/interval.py	`92.85% <0%> (ø)`	⬆️
pandas/core/tools/datetimes.py	`83.79% <0%> (+0.82%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ad7d051...6084e0c. Read the comment docs.

winklerand · 2017-04-29T04:29:42Z

ok, gave parametrized tests a shot in 68f02f7.
In my opinion, those are much more readable - it's clear at first sight which combinations of parameters/inputs are tested.

If you agree, I'd be happy to help moving/refactoring more of the tests in TestPeriodIndex. I'd like to tackle some more issues with PeriodIndex in a follow-up PR (convention parameter, maybe drop kind parameter altogether, ...) which would require better coverage of all parameter combinations/edge cases.

winklerand · 2017-04-29T04:34:53Z

pandas/tests/test_resample.py

+        return DataFrame({'value': np.arange(len(index))}, index=index)
+
+    @pytest.fixture(params=[Series, DataFrame], scope='class')
+    def pandas_obj(self, request, index):


there is probably a more elegant way to create such "meta-fixtures"? ;-)

I would do it more like this

@pytest.fixture(params=[PI, DTI, TDI]) def index(klass): # construct an index return klass(.....) @pytest.fixture def series(index): return Series(.....) @pytest.fixture def frame(index): return DataFrame(....)

it will cartesian product the indexes separately for series/frame which would now be the fixture arg to a test function.

tried another approach by keeping the existing inheritance structure - we can inherit fixtures defined in the Base class, index fixture defined in Base dispatches to the subclass-specific type via the existing _index_factory. If you want to take a look:
https://github.com/winklerand/pandas/blob/bbf01b312a2d8be0b3c7681b9a46a039f33fdcc/pandas/tests/test_resample_pytest.py

Seems easier to me than parametrizing the index fixture, there's a lot of index type-specific test logic.

Should we move out this discussion to another issue or PR? This looks like a more general issue on how to (gradually) transition to pytest.

NO

we are removing the base classes soon

sorry u meant the Base class and not TestCase base class

in any event it's not more clear to me

ok, was maybe a bit confusing ;-)

pushed e497fc2, should illustrate what I meant: basically move fixtures to a Base_pytest(object) class which can be inherited by TestPeriodIndex_pytest, and DatetimeIndex_pytest. This resembles the existing test organization structure, but allows for parametrization using the common fixtures.

max-sixty · 2017-04-29T10:57:06Z

pandas/core/resample.py

+
+        if self.loffset is not None:
+            if self.kind == 'period':
+                print('Warning: loffset -> convert PeriodIndex to timestamps')


Use warnings.warn rather than print

was this here before? we don't warn on things like this.

a similar print-style warning is currently in master: https://github.com/pandas-dev/pandas/blob/master/pandas/core/resample.py#L800:L802

I defer to @jreback

removed the warning in 8a4182e - any objections?

jreback

looks pretty good. I have to do another look on the actual tests.

jreback · 2017-05-01T12:00:44Z

pandas/core/resample.py

        elif is_superperiod(ax.freq, self.freq):
+            if how == 'ohlc':
+                # upsampling to subperiods is handled as an asfreq, which works


can you add the issue reference here

jreback · 2017-05-01T12:01:36Z

pandas/core/resample.py

+                            'an instance of %r' % type(ax).__name__)
+
+        memb = ax.asfreq(self.freq, how=self.convention)
+        # NaT handling as in pandas._lib.lib.generate_bins_dt64()


jreback · 2017-05-01T12:03:33Z

pandas/core/resample.py

+        nat_count = 0
+        if memb.hasnans:
+            import warnings
+            with warnings.catch_warnings():


you don't need this

memb._isnan already has this mask.

jreback · 2017-05-01T12:04:13Z

pandas/core/resample.py

+
+        i8 = memb.asi8
+        freq_mult = self.freq.n
+        # when upsampling to subperiods, we need to generate enough bins


jreback · 2017-05-01T12:04:54Z

pandas/core/resample.py

+        freq_mult = self.freq.n
+        # when upsampling to subperiods, we need to generate enough bins
+        expected_bins_count = len(binner) * freq_mult
+        i8_extend = expected_bins_count - (i8[-1] - i8[0])


this will fail if not len(memb)

Some lines above (after masking out NaTs), we check for if not len(memb) and return and empty index, so this case should be caught.
Maybe put in an assert len(memb) to make it explicit?

jreback · 2017-05-01T12:06:12Z

pandas/tests/test_resample.py

+                                 freq=self._index_fixture_freq)
+
+    @pytest.fixture(params=[Series, DataFrame], scope='class')
+    def pandas_obj(self, request, index):


make this 2 fixtures, series and frame, much simpler, and easier to grok.

you can then make another fixture like series_and_frame if you want, though I find that more confusing, rather have you use params.

jreback · 2017-05-01T12:08:31Z

pandas/tests/test_resample.py

@@ -2777,6 +2694,211 @@ def test_evenly_divisible_with_no_extra_bins(self):
        assert_frame_equal(result, expected)


+def assert_series_or_frame_equal(result, expected):


this is prob ok, but move to util.testing

jreback · 2017-05-01T12:09:48Z

pandas/tests/test_resample.py

+    pass
+
+
+class TestPeriodIndex_pytest(Base_pytest):


I would rather name these TestPeriodIndex(Base) and rename the original tests I think (and then we should move them in other PR's)

winklerand · 2017-05-06T07:37:33Z

Rebased on current master, so we can put the parametrized test in the existing test classes (which just inherit from object).
Thanks for feedback, all requested changed should be completed.

jreback · 2017-05-06T23:09:30Z

cc @MaximilianR can you have a look.

max-sixty · 2017-05-07T07:53:04Z

Looks great. Fixes a couple of important issues.

Whether or not we reveal kind on resamples is an open question - I think there already enough tools for switching between DateTimeIndex and PeriodIndex that we don't need more in resample.

But this definitely makes things better! Thanks @winklerand

jreback

really just some minor stylistic changes.

can you fully close all of the mentioned issues at the top of the PR, and then create a new issue with any remnants. (ideally with check boxes).
Also pls create a new issue for test_resample reorg as I suggested. thanks.

jreback · 2017-05-07T13:42:50Z

pandas/core/resample.py

+        bins = memb.searchsorted(rng, side='left')
+
+        if nat_count > 0:
+            # NaT handling as in pandas._lib.lib.generate_bins_dt64()


is this path tested sufficiently, e.g. 0, 1, 2 NaT?

added a test case for consecutive NaTs in the index (1cad7fa)

Should be sufficiently tested, cases covered:

0 NaT: basically all other resampling tests

multiple single NaTs (at beginning, inside and end of index)

consecutive NaTs (at beginning, inside and end of index)

Any ideas for more exhaustive test cases?

jreback · 2017-05-07T13:44:46Z

pandas/tests/test_resample.py

@@ -696,6 +696,32 @@ def create_index(self, *args, **kwargs):
        factory = self._index_factory()
        return factory(*args, **kwargs)

+    _index_fixture_start = datetime(2005, 1, 1)


as a followup PR. test_resample is getting sufficiently long that it makes sense to split this into some sub-dirs e.g.

pandas/tests/resample/test_api, test_grouping, test_datetimeindex, test_period etc. IOW split out the classes to separate files (to start).

we would normally co-locate fixtures into a common.py

jreback · 2017-05-07T13:46:27Z

pandas/tests/test_resample.py

+    _index_fixture_end = datetime(2005, 1, 10)
+    _index_fixture_freq = 'D'
+
+    @pytest.fixture(scope='class')


normally don't use scoping like this (and instead use default function scoping). The issue is that we could have mutating things which will then propogate (we shouldn't in any event).

makes sense, changed to default function scoping (6fe6c0c)

jreback · 2017-05-07T13:46:57Z

pandas/tests/test_resample.py

@@ -696,6 +696,32 @@ def create_index(self, *args, **kwargs):
        factory = self._index_factory()
        return factory(*args, **kwargs)

+    _index_fixture_start = datetime(2005, 1, 1)
+    _index_fixture_end = datetime(2005, 1, 10)
+    _index_fixture_freq = 'D'


these can be fixtures themselves

jreback · 2017-05-07T13:48:02Z

pandas/tests/test_resample.py

+    @pytest.mark.parametrize('freq', ['2D'])
+    @pytest.mark.parametrize('kind', ['period', None, 'timestamp'])
+    def test_asfreq_downsample(self, series_and_frame, freq, kind):
+        # GH 12884, 15944


give a 1-liner about the issue that you are specifiying test when not obvious (e.g. here)

jreback · 2017-05-07T13:49:49Z

pandas/tests/test_resample.py

+        start = obj.index[0].to_timestamp(how='start')
+        end = (obj.index[-1] + 1).to_timestamp(how='start')
+        if kind == 'timestamp':
+            expected = obj.to_timestamp().resample(freq).asfreq()


can this test be combined with prior test? (looks like you can just add to the freq), or is there a reason not?

yes, parametrization makes this obvious. In fact, both tests cover almost the same code path, so I consolidated them into one (f8f4157).
Reason basically was not to change the existing structure - the same methods are defined in the Base class. I consolidated them as well in the Base class - PTAL.

jreback · 2017-05-07T13:50:32Z

pandas/tests/test_resample.py

+        expected_index = self.create_index(df.index[0],
+                                           periods=len(df.index) / 2,
+                                           freq='2D')
+        # loffset coreces PeriodIndex to DateTimeIndex


jreback · 2017-05-07T13:51:03Z

pandas/tests/test_resample.py

+            result_how = df.resample('2D', how=agg_arg, loffset='2H',
+                                     kind=kind)
+        if isinstance(agg_arg, list):
+            expected.columns = pd.MultiIndex .from_tuples([('value', 'mean')])


you have an odd space after pd.MultiIndex

jreback · 2017-05-07T13:51:12Z

pandas/tests/test_resample.py

+        pi = PeriodIndex(start='2000', freq='D', periods=10)
+        s = Series(range(len(pi)), index=pi)
+        expected = s.to_timestamp().resample(freq).ohlc().to_period(freq)
+        # timestamp-based resampling doesn't include all sub-periods


jreback · 2017-05-07T13:52:08Z

pandas/tests/test_resample.py

+        assert_frame_equal(result, expected)
+
+    def test_resample_with_only_nat(self):
+        pi = PeriodIndex([pd.NaT] * 3, freq='S')


issue number for this?

done (not mentioned in the issue explicitly, is just an edge case)

jreback · 2017-05-07T13:55:12Z

Whether or not we reveal kind on resamples is an open question - I think there already enough tools for switching between DateTimeIndex and PeriodIndex that we don't need more in resample.

For now this is fine. Yes I think we should remove this as PI becomes more flexible. But that's for another time.

jreback · 2017-05-07T13:56:48Z

@winklerand biggest thing is going to need a small section in 0.21.0 to show these changes. In fact I find that writing these examples forces organization on describing the actual changes. Then (in this issue or followup), ideally like to example the timeseries.rst docs w.r.t. PI resampling.

winklerand · 2017-05-21T16:01:49Z

pandas/tests/test_resample.py

-        expected = s.reindex(new_index)
-        assert_series_equal(result, expected)
+    @pytest.mark.parametrize('freq', ['2D', '1H'])
+    def test_asfreq(self, series_and_frame, freq):


@jreback
I consolidated the test_asfreq_downsample() and test_asfreq_upsample() tests in the Base class as well (so we don't have to override them in TestPeriodIndex). PTAL if everything's still correct and covered as before.

winklerand · 2017-05-21T16:32:15Z

@jreback sorry for the delay on this one.

I think I made the requested changes, PTAL. Thanks for the great suggestions.

I put a small PeriodIndex resampling paragraph into the v0.21 whatsnew documentation, in the 'Backward incompatible API changes' section - is this the right place?

Going to organise the issues (closing the solved ones, open new one with the remaining issues) and create a test_resample reorg issue later tonight.

gfyoung · 2017-07-13T19:39:55Z

@jreback : Were we waiting for anything on this one?

jreback

looks pretty good. couple of comments.

jreback · 2017-07-13T19:45:24Z

doc/source/whatsnew/v0.21.0.txt

+
+.. code-block:: ipython
+
+   In [1]: pi = pd.period_range('2017-01', periods=12, freq='M')


the fixed things that are not changed (e.g. [1], [2]), do in a separate ipython block above (to avoid repeating in previous/new).

jreback · 2017-07-13T19:45:47Z

doc/source/whatsnew/v0.21.0.txt

+``PeriodIndex`` resampling
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+In previous versions of pandas, resampling a ``Series``/``DataFrame`` indexed by a ``PeriodIndex`` returned a ``DatetimeIndex`` in some cases (:issue:`12884`). Resampling to a multiplied frequency now returns a ``PeriodIndex`` (:issue:`15944`).


is this the only issue?

jreback · 2017-07-13T19:46:43Z

pandas/core/resample.py

+
+        if self.loffset is not None:
+            # Cannot apply loffset/timedelta to PeriodIndex -> convert to
+            # timestamps


should we show a warning for this?

I think we should, given that we kind of did that before (it was print statement with a warning!?)

jreback · 2017-07-13T19:50:44Z

side issue, maybe we should split up test_resample.py similar to groupby, e.g. make a sub-dir with common, test_datetime, test_period, test_timedelta. to make this a bit more understandable (let's make a separate issue for that).

jreback · 2017-08-18T01:04:07Z

can you rebase. this looked really good. need to review once more.

jreback · 2017-09-23T20:15:46Z

@winklerand sorry let this linger. can you rebase.

jreback · 2017-09-28T14:29:34Z

ok I rebased, will look after passing

jreback · 2017-09-28T14:30:42Z

@MaximilianR can you have a look. in particular can you look at the original issues and see if we are fully / partially covering and make a note.

max-sixty · 2017-09-28T19:49:36Z

pandas/core/resample.py

+
+        # if index contains no valid (non-NaT) values, return empty index
+        if not len(memb):
+            binner = labels = PeriodIndex(


You could use _shallow_copy here, but this is OK

I left this, ok for now.

Added tests to cover all code paths, moved the check to _convert_obj() and removed then-redundant check from _upsample() method.

Exceptions: - force conversion to DatetimeIndex by kind='timestamp' param - if loffset is given, convert to timestamps in any case

Removed previously defined helper method assert_series_or_frame_equal(), which behaved identical to assert_almost_equal()

…_asfreq

jreback · 2017-10-01T14:55:39Z

thanks @winklerand nice fixes!

* 'master' of github.com:pandas-dev/pandas: (188 commits) Separate out _convert_datetime_to_tsobject (pandas-dev#17715) DOC: remove whatsnew note for xref pandas-dev#17131 BUG: Regression in .loc accepting a boolean Index as an indexer (pandas-dev#17738) DEPR: Deprecate cdate_range and merge into bdate_range (pandas-dev#17691) CLN: replace %s syntax with .format in pandas.core: categorical, common, config, config_init (pandas-dev#17735) Fixed the memory usage explanation of categorical in gotchas from O(nm) to O(n+m) (pandas-dev#17736) TST: add backward compat for offset testing for pickles (pandas-dev#17733) remove unused time conversion funcs (pandas-dev#17711) DEPR: Deprecate convert parameter in take (pandas-dev#17352) BUG:Time Grouper bug fix when applied for list groupers (pandas-dev#17587) BUG: Fix some PeriodIndex resampling issues (pandas-dev#16153) BUG: Fix unexpected sort in groupby (pandas-dev#17621) DOC: Fixed typo in documentation for 'pandas.DataFrame.replace' (pandas-dev#17731) BUG: Fix series rename called with str altering name rather index (GH17407) (pandas-dev#17654) DOC: Add examples for MultiIndex.get_locs + cleanups (pandas-dev#17675) Doc improvements for IntervalIndex and Interval (pandas-dev#17714) BUG: DataFrame sort_values and multiple "by" columns fails to order NaT correctly Last of the timezones funcs (pandas-dev#17669) Add missing file to _pyxfiles, delete commented-out (pandas-dev#17712) update imports of DateParseError, remove unused imports from tslib (pandas-dev#17713) ...

winklerand · 2017-10-06T17:44:18Z

thanks @jreback for pushing this one over the finish line, completely forgot about it (works for me syndrome ;-)), sorry!

closes pandas-dev#15944 xref pandas-dev#12884 closes pandas-dev#13083 closes pandas-dev#13224

winklerand changed the title ~~BUG: Fix some period index resampling issues~~ BUG: Fix some PeriodIndex resampling issues Apr 26, 2017

jreback reviewed Apr 26, 2017

View reviewed changes

jreback added Period Period data type Resample resample method labels Apr 27, 2017

winklerand commented Apr 29, 2017

View reviewed changes

max-sixty reviewed Apr 29, 2017

View reviewed changes

jreback requested changes May 1, 2017

View reviewed changes

winklerand force-pushed the fix-some-PeriodIndex-resampling-issues branch from 8a4182e to 09d7ac3 Compare May 6, 2017 07:31

jreback requested changes May 7, 2017

View reviewed changes

jreback added this to the 0.21.0 milestone May 7, 2017

winklerand commented May 21, 2017

View reviewed changes

winklerand mentioned this pull request Jun 9, 2017

Upsampling a time-series is missing an option to properly deal with the end #10449

Open

jreback requested changes Jul 13, 2017

View reviewed changes

jreback force-pushed the fix-some-PeriodIndex-resampling-issues branch from 48f1e76 to 0f63098 Compare September 28, 2017 14:29

max-sixty reviewed Sep 28, 2017

View reviewed changes

winklerand and others added 18 commits September 29, 2017 08:15

TST/CLN: raise error when resampling with on= or level= selection

c27f430

Added tests to cover all code paths, moved the check to _convert_obj() and removed then-redundant check from _upsample() method.

BUG: resampling PeriodIndex now returns PeriodIndex (GH 12884, 15944)

390e16e

Exceptions: - force conversion to DatetimeIndex by kind='timestamp' param - if loffset is given, convert to timestamps in any case

BUG: OHLC-upsampling of PeriodIndex now returns DataFrame (GH 13083)

23566c2

BUG: enable resampling with NaT in PeriodIndex (GH 13224)

a82879d

CLN: remove warning on falling back to tstamp resampling with loffset

4b1c740

CLN: use memb._isnan for NaT masking

73c0990

DOC: added issue reference for OHLC resampling

fa6c1d3

STYLE: added blank lines

7ea04e9

TST: convert to parametrized tests / pytest idiom

82a8275

CLN/TST: call assert_almost_equal() when comparing Series/DataFrames

432c623

Removed previously defined helper method assert_series_or_frame_equal(), which behaved identical to assert_almost_equal()

STYLE: added blank lines, removed odd whitespace, fixed typo

c8814fb

TST: add test case for multiple consecutive NaTs in PeriodIndex

486ad67

TST/DOC: added issue number to test case

ad8519f

TST: consolidate test_asfreq_downsample, test_asfreq_upsample -> test…

39fc7e2

…_asfreq

TST: set fixtures to default function scoping

efcad5b

TST: convert constant 'setup-like' values/objects to pytest fixtures

41401d4

DOC: whatsnew v0.21.0 entry (in API changes section)

398a684

fixups

8358c41

jreback mentioned this pull request Sep 29, 2017

ENH: PeriodIndex resampling redux #12884

Open

moar whatsnew

6084e0c

jreback force-pushed the fix-some-PeriodIndex-resampling-issues branch from 0f63098 to 6084e0c Compare September 29, 2017 12:24

jreback approved these changes Sep 29, 2017

View reviewed changes

jreback merged commit 7d4a260 into pandas-dev:master Oct 1, 2017

winklerand deleted the fix-some-PeriodIndex-resampling-issues branch October 6, 2017 17:44

winklerand mentioned this pull request Oct 6, 2017

TST/MAINT: split up test_resample.py to make it more understandable #17806

Closed

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

BUG: Fix some PeriodIndex resampling issues (pandas-dev#16153)

4c3ca19

closes pandas-dev#15944 xref pandas-dev#12884 closes pandas-dev#13083 closes pandas-dev#13224

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

BUG: Fix some PeriodIndex resampling issues (pandas-dev#16153)

e76612f

closes pandas-dev#15944 xref pandas-dev#12884 closes pandas-dev#13083 closes pandas-dev#13224

		@@ -2777,6 +2694,211 @@ def test_evenly_divisible_with_no_extra_bins(self):
		assert_frame_equal(result, expected)


		def assert_series_or_frame_equal(result, expected):


		.. code-block:: ipython

		In [1]: pi = pd.period_range('2017-01', periods=12, freq='M')

BUG: Fix some PeriodIndex resampling issues #16153

BUG: Fix some PeriodIndex resampling issues #16153

Conversation

winklerand commented Apr 26, 2017 • edited by jreback Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Apr 27, 2017

codecov bot commented Apr 27, 2017 • edited Loading

Codecov Report

winklerand commented Apr 29, 2017 • edited Loading

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

winklerand Apr 30, 2017 • edited Loading

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

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

winklerand commented May 6, 2017 • edited Loading

jreback commented May 6, 2017

max-sixty commented May 7, 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

Choose a reason for hiding this comment

jreback commented May 7, 2017

jreback commented May 7, 2017

Choose a reason for hiding this comment

winklerand commented May 21, 2017

gfyoung commented Jul 13, 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

jreback commented Jul 13, 2017

jreback commented Aug 18, 2017

jreback commented Sep 23, 2017

jreback commented Sep 28, 2017

jreback commented Sep 28, 2017

Choose a reason for hiding this comment

jreback Sep 29, 2017 • edited Loading

Choose a reason for hiding this comment

jreback commented Oct 1, 2017

winklerand commented Apr 26, 2017 •

edited by jreback

Loading

codecov bot commented Apr 27, 2017 •

edited

Loading

winklerand commented Apr 29, 2017 •

edited

Loading

winklerand Apr 30, 2017 •

edited

Loading

winklerand commented May 6, 2017 •

edited

Loading

jreback Sep 29, 2017 •

edited

Loading