BUG: to_datetime re-parsing Arrow-backed objects

[lithomas1]

• closes PERF: to_datetime doesn't recognise Arrow date time dtypes and converts them again #52545
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

Ideally, we could push this down into DTA/DTI, but that doesn't seem to support Arrow arrays ATM.

[mroeschke]

Could use a whatsnew too.

Also, does something similar happen to to_timedelta?

[lithomas1]

Also, does something similar happen to to_timedelta?

I think so, I'll try to get at it at a followup.

I think in the future, it might also make sense for DF inputs with Arrow backed year, month, etc. columns to convert to Arrow Backed dt columns.

e.g.

>>> import pandas as pd
+ /Users/thomasli/opt/miniconda3/envs/pandas-dev/bin/ninja
[1/1] Generating write_version_file with a custom command
>>> df = pd.DataFrame({'year': [2015, 2016],
...                    'month': [2, 3],
...                    'day': [4, 5]})
>>> df = df.astype("int64[pyarrow]")
>>> pd.to_datetime(df)
0   2015-02-04
1   2016-03-05
dtype: datetime64[ns]

should return something with dtype: timestamp[ns][pyarrow]

Main thing I'm worried about with this PR is doing the wrong thing for utc=True (e.g. localizing instead of converting and vice versa).

(cc @phofl in case any thoughts and to double check my logic).

[phofl]

very small comment

[phofl]

I'd rather use pd.ArrowDtype(pa...) here

[jbrockmendel]

.astype shouldn't allow converting a tz-naive to tzaware (or vice-versa). i think this should use tz_convert/tz_localize instead

[jbrockmendel]

(btw if .astype works on ArrowEA to convert tznaive to tzaware we probably want to change that to match the non-arrow raising behavior)

[lithomas1]

(btw if .astype works on ArrowEA to convert tznaive to tzaware we probably want to change that to match the non-arrow raising behavior)

Sure, I'll raise a PR after this one gets in.

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this.

[MarcoGorelli]

is tz_localize always the right one, or do we need tz_convert if the input was timezone-aware to begin with?

In the test you've added, dti_arrow is always timezone-naive - would we make a timezone-aware example too please, just to check this?

[lithomas1]

Thx for the catch.

I messed up the test (it does parametrize over UTC and US/Central), but I forgot to propagate the tz information in the tests.

[MarcoGorelli]

Well done, looks good to me!

[mroeschke]

Thanks @lithomas1

[jbrockmendel]

Does the to_datetime docstring need to be updated to reflect non-DTI return?