ENH: Downsampling pyarrow backed DatetimeIndex should potentially return date32 instead of timestamp

[WillAyd]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

In [30]: pd.Series(
    ...:     range(3),
    ...:     index=pd.Index([
    ...:         "2024-01-01 00:00:00",
    ...:         "2024-01-01 12:00:00",
    ...:         "2024-01-02 00:00:00"
    ...:     ], dtype=pd.ArrowDtype(pa.timestamp("s")))
    ...: ).resample("D").asfreq().index.dtype
Out[30]: timestamp[s][pyarrow]

Feature Description

When downsampling to a resolution of day or lower, I think .resample should return a pyarrow date type

Alternative Solutions

status quo

Additional Context

No response

[jbrockmendel]

why? im wary of any silent dtype-changing, which we're moving away from elsewhere

[WillAyd]

I think resampling to days declares the intent of working with dates and not timestamps

[mroeschke]

If a user were to resample by even a lower resolution (e.g. W, ME), would you also expect date32 type?

[WillAyd]

Yes - anything date or lower I would expect a date.