ENH: add fill_value to resample #14591

nchmura4 · 2016-11-05T03:38:35Z

closes ENH: Add fill_value option to resample() #3715
tests added / passed
passes git diff upstream/master | flake8 --diff
whatsnew entry

max-sixty · 2016-11-10T18:08:47Z

If I'm understanding this correctly, this is equivalent to .fillna prior to .resample. Another interpretation of the kwarg is that it fills a value while upsampling.

If that's right, is this worth a kwarg? Why not just leave it to the user to fillna prior to resampling?

nchmura4 · 2016-11-10T19:44:43Z

I think that's a fair question. This is how i understood 3755. An argument for this feature is that you have a fill_value kwarg in reindex, whereas that's equivalent to just running reindex(...).fillna(), right?

max-sixty · 2016-11-10T19:53:24Z

That kwarg reindex only fills empty values created on reindex. (although I also recall that behavior not being consistent):

In [3]: series = pd.Series(range(4)) In [4]: series[2] = pd.np.nan In [5]: series Out[5]: 0 0.0 1 1.0 2 NaN 3 3.0 dtype: float64 In [6]: series.reindex(range(6), fill_value=9) Out[6]: 0 0.0 1 1.0 2 NaN 3 3.0 4 9.0 5 9.0 dtype: float64

nchmura4 · 2016-11-10T21:24:07Z

i see. got it, thanks.

jorisvandenbossche · 2016-11-11T08:59:06Z

@nchmura4 As Maximilian said, I also think that a possible fill_value keyword should be for filling NaNs created during resampling, not to fill NaNs before resample (as IMO a keyword for that should not be worth it as df.fillna(..).resample(..) is just much more explicit that is does that).
Sorry if we weren't clear about that on the issue!

nchmura4 · 2016-11-15T05:46:07Z

no problem, thanks for the feedback. Would this example illustrate the desired behavior?

rng = pd.date_range('1/1/2016', periods=5, freq='4s') ts = pd.Series(np.arange(len(rng)), index=rng) ts2 = pd.Series([1,2,None,4,5], index=rng) df = pd.DataFrame({'one': ts, 'two': ts2}) df one two 2016-01-01 00:00:00 0 1.0 2016-01-01 00:00:04 1 2.0 2016-01-01 00:00:08 2 NaN 2016-01-01 00:00:12 3 4.0 2016-01-01 00:00:16 4 5.0

upsampling with fill_value doesn't fill any NaN's already in the dataframe, but does fill new rows during resampling:
df.resample("2s", fill_value=9).asfreq()

 one two 2016-01-01 00:00:00 0.0 1.0 2016-01-01 00:00:02 9.0 9.0 2016-01-01 00:00:04 1.0 2.0 2016-01-01 00:00:06 9.0 9.0 2016-01-01 00:00:08 2.0 NaN 2016-01-01 00:00:10 9.0 9.0 2016-01-01 00:00:12 3.0 4.0 2016-01-01 00:00:14 9.0 9.0 2016-01-01 00:00:16 4.0 5.0

jorisvandenbossche · 2016-11-15T19:26:44Z

@nchmura4 yes, I think that is a correct example.

Question is here maybe: where does this keyword belong? In resample, or in asfreq ? Or to put it in other words: does such a possible fill_value keyword make sense for all methods on a resample, or only for some?

nchmura4 · 2016-11-19T16:48:05Z

I think this only makes sense on unsampling. I can't think of a use case for downsampling. So then I think it belongs in asfreq(). I'm working on that.

codecov-io · 2016-11-23T13:44:01Z

Current coverage is 85.55% (diff: 100%)

Merging #14591 into master will not change coverage

@@ master #14591 diff @@ ========================================== Files 145 145 Lines 51352 51352 Methods 0 0 Messages 0 0 Branches 0 0 ========================================== Hits 43932 43932 Misses 7420 7420 Partials 0 0

Powered by Codecov. Last update 362e78d...45a9e8e

jreback · 2016-11-23T17:26:44Z

doc/source/whatsnew/v0.20.0.txt


 - ``pd.read_excel`` now preserves sheet order when using ``sheetname=None`` (:issue:`9930`)
-
+- ``DataFrame.asfreq()`` now accepts a fill_value option to fill missing values during resampling (:issue:`3715`).


double-back ticks around fill_value

jreback · 2016-11-23T17:27:29Z

pandas/core/generic.py

+ value to use for missing values, applied during upsampling
+
+ .. version added:: 0.20.0



can you add some Examples here

jreback · 2016-11-23T17:28:04Z

pandas/tseries/tests/test_resample.py

 r.fillna(0)

+ def test_fill_value(self):
+ # setup


add the issue number as a comment, 1-liner expl

jreback · 2016-11-23T17:29:15Z

pandas/tseries/tests/test_resample.py

+ # setup
+ rng = pd.date_range('1/1/2016', periods=10, freq='2S')
+ ts = pd.Series(np.arange(len(rng)), index=rng)
+ ts2 = pd.Series(np.random.rand(len(rng)), index=rng)


can you add a test also to where DataFarme.asfreq is tested (prob not very much :<)

{lamb} grin ".asfreq" pandas\tests\ pandas\tests\frame\test_timeseries.py: 299 : def test_asfreq(self): 300 : offset_monthly = self.tsframe.asfreq(offsets.BMonthEnd()) 301 : rule_monthly = self.tsframe.asfreq('BM') 305 : filled = rule_monthly.asfreq('B', method='pad') # noqa 309 : filled_dep = rule_monthly.asfreq('B', method='pad') # noqa 313 : result = zero_length.asfreq('BM') 316 : def test_asfreq_datetimeindex(self): 320 : df = df.asfreq('B') 323 : ts = df['A'].asfreq('B') pandas\tests\indexes\test_datetimelike.py: 830 : idx.get_loc(idx[1].asfreq('H', how='start'), method), 1) 888 : idx = pd.period_range('2000-01-01', periods=3).asfreq('H', how='start') pandas\tests\plotting\test_datetimelike.py: 319 : bts = tm.makeTimeSeries().asfreq('BM') 740 : ts2 = ts.asfreq('T').interpolate() pandas\tests\series\test_datetime_values.py: 37 : ok_for_period_methods = ['strftime', 'to_timestamp', 'asfreq'] pandas\tests\series\test_timeseries.py: 407 : def test_asfreq(self): 411 : daily_ts = ts.asfreq('B') 412 : monthly_ts = daily_ts.asfreq('BM') 415 : daily_ts = ts.asfreq('B', method='pad') 416 : monthly_ts = daily_ts.asfreq('BM') 419 : daily_ts = ts.asfreq(BDay()) 420 : monthly_ts = daily_ts.asfreq(BMonthEnd()) 423 : result = ts[:0].asfreq('M')

and pls add a tests for series as well

jreback · 2016-11-23T17:29:30Z

pandas/tseries/tests/test_resample.py

+ expected_df = df.resample('1S').asfreq().fillna(9.0)
+ expected_df.loc['2016-01-01 00:00:08', 'two'] = None
+
+ assert_frame_equal(expected_df, actual_df)


add for series as well

jreback · 2016-11-23T17:29:45Z

@nchmura4 looks pretty good.

nchmura4 · 2016-11-25T16:32:37Z

thanks - I added documentation and tests you suggested

jreback · 2016-11-25T16:56:07Z

pandas/core/generic.py

 -------
 converted : type of caller

+ Examples


jreback · 2016-11-25T16:56:33Z

pandas/core/generic.py

 how : {'start', 'end'}, default end
 For PeriodIndex only, see PeriodIndex.asfreq
 normalize : bool, default False
 Whether to reset output index to midnight


can you add a sentence or 2 on the use / diffs of .asfreq() / resample.

jorisvandenbossche · 2016-11-25T21:08:47Z

pandas/core/generic.py

+ 2000-01-01 00:01:30 9.0
+ 2000-01-01 00:02:00 2.0
+ 2000-01-01 00:02:30 9.0
+ 2000-01-01 00:03:00 3.0


Really nice example, thanks!
Can you add the output using the existing method keyword (also for filling NaNs) with this example as well? (pick one of the possible values for the keyword)

jorisvandenbossche · 2016-11-25T21:11:51Z

pandas/core/generic.py

 normalize : bool, default False
 Whether to reset output index to midnight
+ fill_value: scalar, optional
+ value to use for missing values, applied during upsampling


Can you add a similar sentence as for method: "this does not fill NaNs that were already present"?
(can you also capitalize the first word of the sentence?

nchmura4 · 2016-11-29T05:48:56Z

thanks for the feedback

nchmura4 · 2016-12-02T07:07:52Z

sorry, i think i accidentally messed this pull request up :/ was trying to pull new changes locally and accidentally pushed them into this PR. Can someone help me? sorry!

jorisvandenbossche · 2016-12-02T23:17:58Z

@nchmura4 If you follow those steps, it should be solved:

git fetch upstream git checkout nchmura-resample-fill-value git rebase upstream/master git push -f origin nchmura-resample-fill-value

nchmura4 · 2016-12-03T18:41:11Z

@jorisvandenbossche thank you! i see what i did wrong. PR all up to date now :)

jreback · 2016-12-04T17:55:22Z

pandas/core/generic.py

+ 2000-01-01 00:02:30 3.0
+ 2000-01-01 00:03:00 3.0
+
 Notes


can you add a See Also to .reindex (which has fill_value)

jreback · 2016-12-04T17:55:55Z

pandas/tseries/resample.py

+ this does not fill NaNs that already were present).
+
+ .. versionadded:: 0.20.0
+


can you add a See Also to Series.asfreq/DataFrame.asfreq

jreback · 2016-12-04T17:58:20Z

pandas/tseries/tests/test_resample.py

 with self.assertRaises(ValueError):
 r.fillna(0)

+ def test_fill_value(self):


these tests should be in class Base. Which will tests with all index types (you need to tests both Series/DataFrame here)

use self.create_index() to create your index.

@jreback I'm a bit stuck on PeriodIndex. With my current commits you can add a fill_value for PeriodIndexResampler._upsample with kind='timestamp' but it doesn't work for kind='period'. I think I have to build that out. Otherwise, this won't work for PeriodIndex.

Upsampling on a PeriodIndex with kind='period' doesn't work through reindex. I think it would involve adding fill_value to Index.get_indexer (line 901 of tseries/resample.py) Before I start going down that path, I just want to confirm that's its the right thing to do.

jreback · 2016-12-30T21:44:02Z

can you rebase

jreback · 2017-01-21T22:50:26Z

thanks @nchmura4 !

jreback · 2017-01-21T22:51:23Z

closed by 2540d5a

nchmura4 changed the title ~~Nchmura resample fill value~~ ENH: add fill_value to resample Nov 5, 2016

sinhrks added Enhancement Missing-data np.nan, pd.NaT, pd.NA, dropna, isnull, interpolate Resample resample method labels Nov 7, 2016

nchmura4 closed this Nov 9, 2016

nchmura4 reopened this Nov 9, 2016

nchmura4 force-pushed the nchmura-resample-fill-value branch from 58d72bc to 649a56b Compare November 23, 2016 12:57

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Nov 23, 2016

ENH: add fill_value to asfreq pandas-dev#14591

43acefc

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Nov 23, 2016

ENH: pandas-dev#14591 - tests and documentation

a7b394d

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Nov 23, 2016

ENH: pandas-dev#14591 - formatting

649a56b

jreback reviewed Nov 23, 2016

View reviewed changes

jreback reviewed Nov 25, 2016

View reviewed changes

pandas/core/generic.py

-------

converted : type of caller

Examples

Copy link

Contributor

jreback Nov 25, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nice!

jreback reviewed Nov 25, 2016

View reviewed changes

jorisvandenbossche reviewed Nov 25, 2016

View reviewed changes

nchmura4 force-pushed the nchmura-resample-fill-value branch from c8cd7f2 to aab4497 Compare December 3, 2016 18:04

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: add fill_value to asfreq pandas-dev#14591

4df51cf

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: pandas-dev#14591 - tests and documentation

2701ccb

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: pandas-dev#14591 - formatting

9dfb3f6

nchmura4 force-pushed the nchmura-resample-fill-value branch from aab4497 to 8f9097e Compare December 3, 2016 18:37

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: add fill_value to asfreq pandas-dev#14591

7d1c171

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: pandas-dev#14591 - tests and documentation

b343615

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 3, 2016

ENH: pandas-dev#14591 - formatting

717239c

jreback reviewed Dec 4, 2016

View reviewed changes

nchmura4 force-pushed the nchmura-resample-fill-value branch from 8f9097e to 133175d Compare December 21, 2016 06:21

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 21, 2016

ENH: add fill_value to asfreq pandas-dev#14591

54961e0

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 21, 2016

ENH: pandas-dev#14591 - tests and documentation

5ef1909

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 21, 2016

ENH: pandas-dev#14591 - formatting

6aa6d69

nchmura4 force-pushed the nchmura-resample-fill-value branch from 133175d to 36095c6 Compare December 31, 2016 04:40

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 31, 2016

ENH: add fill_value to asfreq pandas-dev#14591

bd704ff

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 31, 2016

ENH: pandas-dev#14591 - tests and documentation

76d5a14

nchmura4 added a commit to nchmura4/pandas that referenced this pull request Dec 31, 2016

ENH: pandas-dev#14591 - formatting

7397472

nchmura4 force-pushed the nchmura-resample-fill-value branch from 36095c6 to 45a9e8e Compare January 18, 2017 04:11

ENH: add fill_value to asfreq, see pandas-dev#3715

08b6b28

nchmura4 force-pushed the nchmura-resample-fill-value branch from 45a9e8e to 08b6b28 Compare January 18, 2017 05:34

jreback added this to the 0.20.0 milestone Jan 21, 2017

jreback closed this Jan 21, 2017


		- ``pd.read_excel`` now preserves sheet order when using ``sheetname=None`` (:issue:`9930`)

		- ``DataFrame.asfreq()`` now accepts a fill_value option to fill missing values during resampling (:issue:`3715`).

		value to use for missing values, applied during upsampling

		.. version added:: 0.20.0

		this does not fill NaNs that already were present).

		.. versionadded:: 0.20.0

Uh oh!

ENH: add fill_value to resample #14591

ENH: add fill_value to resample #14591

Uh oh!

Conversation

nchmura4 commented Nov 5, 2016

max-sixty commented Nov 10, 2016

nchmura4 commented Nov 10, 2016

max-sixty commented Nov 10, 2016

nchmura4 commented Nov 10, 2016

jorisvandenbossche commented Nov 11, 2016

nchmura4 commented Nov 15, 2016

jorisvandenbossche commented Nov 15, 2016

nchmura4 commented Nov 19, 2016

codecov-io commented Nov 23, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Current coverage is 85.55% (diff: 100%)

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 Nov 23, 2016

nchmura4 commented Nov 25, 2016

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

nchmura4 commented Nov 29, 2016

nchmura4 commented Dec 2, 2016

jorisvandenbossche commented Dec 2, 2016

nchmura4 commented Dec 3, 2016

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 Dec 30, 2016

jreback commented Jan 21, 2017

jreback commented Jan 21, 2017

Labels

7 participants

codecov-io commented Nov 23, 2016 •

edited

Loading