BUG: dt64.astype(period) allowed but not vice-versa

[jbrockmendel]

import pandas as pd

dti = pd.date_range("2016-01-01", periods=3)
pi = dti.to_period("D")

dti.astype(pi.dtype)  # <- OK
pi.astype(dti.dtype)  # <- works but deprecated in #44398

pd.Series(dti).astype(pi.dtype)  # <- OK
pd.Series(pi).astype(dti.dtype)  # <- raises

In #44398 we deprecated PeriodIndex.astype(dt64). The motivation was that PeriodIndex.astype should be consistent with Series[Period].astype. I chose to deprecate the PeriodIndex behavior instead of allow it in Series.astype bc there's an optional keyword in PeriodIndex.astype that we're deprecating anyway.

I'm having seconds thoughts based on the fact that we do support Series[dt64].astype(period), and it seems like these should be able to round-trip.

[jbrockmendel]

Added blocker label bc if we want to un-deprecate it will need to be before the release.

[jreback]

i think fine to undeprecate and match the series behavior here.

[jbrockmendel]

i think fine to undeprecate and match the series behavior here.

So we're clear: the existing deprecation, when enforced, will change the PeriodIndex behavior to match the Series[Period] behavior. This issue is suggesting we 1) un-deprecate and 2) change the Series[Period] behavior to match the PeriodIndex behavior.

[jbrockmendel]

putting together a branch now to Just Do It.