TST: Test resampling with NaT

[evectant]

• Closes Unclear error produced by resampling with NaT in Index #13020
• Adds test_generic.py:TestDataFrame.test_resample_with_nat
• Passes tests and flake8
• No documentation needed

[sinhrks]

Thanks. Can you also add TimedeltaIndex and PeriodIndex?

[jreback]

move all of these to tests/tseries/test_resample.py

[codecov-io]

Current coverage is 84.12%

Merging #13164 into master will decrease coverage by <.01%

@@             master     #13164   diff @@
==========================================
  Files           138        138          
  Lines         50388      50380     -8   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
- Hits          42386      42379     -7   
+ Misses         8002       8001     -1   
  Partials          0          0          

1. 2 files (not in diff) in pandas/tseries were modified. more
   • Misses +1
   • Hits -2
2. 2 files (not in diff) in pandas/tools were modified. more
   • Hits -4
3. 4 files (not in diff) in pandas were modified. more
   • Misses -2
   • Hits -1

Powered by Codecov. Last updated by 4e4a7d9...6d47c87

[jreback]

@evectant can you rebase & update

[evectant]

@sinhrks Those don't seem to be very tolerant to NaT. I'm not sure what behavior we should pin down here.

In [2]: dti = pd.DatetimeIndex(['1970-01-01 00:00:00', pd.NaT, '1970-01-01 00:00:02'])

In [3]: pd.DataFrame([2, 3, 5], dti).resample('1s').mean()
Out[3]: 
                       0
1970-01-01 00:00:00  2.0
1970-01-01 00:00:01  NaN
1970-01-01 00:00:02  5.0

In [4]: tdi = pd.to_timedelta(['0s', pd.NaT, '2s'])

In [5]: pd.DataFrame([2, 3, 5], tdi).resample('1s').mean()
...
MemoryError:

In [6]: pi = pd.PeriodIndex(['1970-01-01 00:00:00', pd.NaT, '1970-01-01 00:00:02'], freq='S')

In [7]: pd.DataFrame([2, 3, 5], pi).resample('1s').mean()
...
ValueError: Start and end must not be NaT

[jreback]

if you move these to Base and use .create_series() you will get testing for free for all of the sub-classes (DTI, PeriodIndex, TimedeltaIndex)

[evectant]

Would it be helpful here? All three index types seem to handle NaT differently (see examples above).

[sinhrks]

@evectant Thx to confirm. It's gr8 if you can fix, otherwise keep #13020 open or submit new issue for tracking.

[evectant]

@sinhrks @jreback I think there are three different issues here:

• Resampling with NaT in DatetimeIndex raises ValueError (fixed in master, pinned down here)
• Resampling with NaT in TimedeltaIndex raises MemoryError (entered Resampling with NaT in TimedeltaIndex raises MemoryError #13223, will look into)
• Resampling with NaT in PeriodIndex raises ValueError (entered Resampling with NaT in PeriodIndex raises ValueError #13224, will look into)

So can this one be merged and #13020 closed?

[jreback]

thanks @evectant

merged.

when you work on the other too resample-with-nat issues, ideally we move this test to Base (so it cycles thru all index types); then these are all in 1 place.