CLN: Datetimelike._can_hold_na

[sinhrks]

• tests added / passed
• passes git diff upstream/master | flake8 --diff
• no whatsnew, as it is internal fix

datetimelike Index._can_hold_na is all False even though these can store NaT.

idx = pd.DatetimeIndex(['2011-01-01', 'NaT'])
idx.hasnans
# True

idx._can_hold_na
# False

idx._nan_idxs
# array([], dtype=int64)

Because of test_base.py condition below, datetime-like Indexes had skipped some tests until now.

• https://github.com/pydata/pandas/blob/master/pandas/tests/test_base.py#L219

[codecov-io]

Current coverage is 85.28% (diff: 100%)

Merging #13983 into master will increase coverage by <.01%

@@             master     #13983   diff @@
==========================================
  Files           139        139          
  Lines         50543      50547     +4   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          43101      43107     +6   
+ Misses         7442       7440     -2   
  Partials          0          0          

Powered by Codecov. Last update 8fdfa51...5435247

[sinhrks]

Updated not to rely on #13979. _get_unique_index can be removed after #13984.

[sinhrks]

@jreback @jorisvandenbossche can u check? This is needed to make some Index dtype related fixes.

[jorisvandenbossche]

This is looking good to me. @jreback OK?

[jreback]

needs #13979 first

[jreback]

shouldn't this be handled in the super?

[sinhrks]

This is needed to handle NaT. #13979 and #13984 makes _get_unique_index itself unnecessary.

[jorisvandenbossche]

@sinhrks updated to not rely on that one

[sinhrks]

@jreback Now this PR doesn't rely on #13979.

[jorisvandenbossche]

@sinhrks This needs a rebase now I merged the other

[jreback]

lgtm. ping on green.

[sinhrks]

fixed lint error, and now green.

[jorisvandenbossche]

Thanks!