BUG: value_counts may raise OutOfBoundsDatetime

[sinhrks]

• closes series with datetime raises pandas.tslib.OutOfBoundsDatetime on value_counts() #13663
• tests added / passed
• passes git diff upstream/master | flake8 --diff
• whatsnew entry

The root cause looks normal Index raises OutOfBoudsDatetime. I think it should be coerced to object dtype because user specifies no dtype.

pd.Index([datetime(9999, 1, 1)])
# on current master
# OutOfBoundsDatetime: Out of bounds nanosecond timestamp: 9999-01-01 00:00:00

# this PR
Index([9999-01-01 00:00:00], dtype='object')

[codecov-io]

Current coverage is 84.57% (diff: 100%)

No coverage report found for master at b60e42b.

Powered by Codecov. Last update b60e42b...533bba3

[TomAugspurger]

Gave a quick look, everything looks good.

Slightly OT, do we have a convention on import datetime vs. from datetime import datetime? You changed it to the second version, which seems to be slightly more common in the codebase? Might be nice to standardize that everywhere.

[jreback]

thanks @sinhrks

yeah @TomAugspurger there is a bit of differences on import, though generally not back; we mostly from datetime import datetime which is the correct idiom here. Could change if someone wants to put in the effort :>

pls make an issue.

[sinhrks]

yeah, from datetime import datetime looks commonto meet with < 80 chars. Though i don't have strong opinion about it.