"is_all_dates" inconsistent across datetimelike types

[toobaz]

xref #19187

Code Sample, a copy-pastable example if possible

In [2]: pd.DatetimeIndex(['2000-02-01', '2000-02-02']).is_all_dates
Out[2]: True

In [3]: pd.DatetimeIndex(['2000-02-01', '2000-02-02']).append(pd.Index(['not a date']))[:2].is_all_dates
Out[3]: True

In [4]: pd.PeriodIndex(['2000', '2001'], freq='A').is_all_dates
Out[4]: True

In [5]: pd.PeriodIndex(['2000', '2001'], freq='A').append(pd.Index(['not a date']))[:2].is_all_dates
Out[5]: False

In [6]: pd.TimedeltaIndex([2000, 2001]).append(pd.Index(['not a date']))[:2].is_all_dates
Out[6]: False

Problem description

is_all_dates is smart for object indexes but only for DatetimeIndex, which is inconsistent since the other datetimelike indexes are also considered is_all_dates.

Expected Output

Out[5]: and Out[6]: should be the same as Out[3]:.

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]

[jreback]

this should be made more like the other is_* methods (in base.py), where it simply checks the inferred type (and then checks for object), needs a doc-string

[giba0]

I'm a newbie here and I'm in search of fun! Can I help you with that? =)

[TomAugspurger]

@gilbertoolimpio yep, that'd be great. Do you know where to start?

[toobaz]

this should be made more like the other is_* methods (in base.py), where it simply checks the inferred type (and then checks for object), needs a doc-string

I suggest then to just deprecate it, since it becomes equivalent to a simple isisntance call, only more misterious.