DEPR: tz kwarg in Period.to_timestamp

[jbrockmendel]

• closes #xxxx
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

inconsistent with PeriodArray/PeriodIndex methods

[jreback]

pls add this to the deprecation removal issue

[jorisvandenbossche]

Is there a motivation?

[jbrockmendel]

Is there a motivation?

As mentioned in the OP, its inconsistent with the PeriodArray/PeriodIndex methods.

[jorisvandenbossche]

Whoops, sorry, missed that sentence.

Now, this keyword doesn't seem really a burden to me, so I still don't find that a fully convincing reason (could it be added to the array method?)

Also, it seems we are using this feature internally in pandas, according to the test failure.

[jorisvandenbossche]

Also, it seems we are using this feature internally in pandas, according to the test failure.

It's in timeseries.py. It can of course easily be fixed there to use tz_localize

[jbrockmendel]

Now, this keyword doesn't seem really a burden to me, so I still don't find that a fully convincing reason (could it be added to the array method?)

Either way works. My default is to prefer a smaller API surface, but not a strong preferenc.

Note also that Period.to_timestamp incorrectly ignores the tz arg in the how == 'end' case, so i see it as deprecating a broken feature.

[jreback]

agree with @jbrockmendel here; simpler is better, we always convert to naive, which is inline the model of Periods which are naive by definition.

[jorisvandenbossche]

OK, if it's only partly implemented, that's certainly another reason to deprecate it