API: Have invalid_comparison return pandas object following `__pandas_priority__`?

[mroeschke]

Currently an invalid comparison (eq and ne only) returns an np.ndarray[bool].

In [5]: i = pd.Index([1, 2, 3], dtype='datetime64[ns]')

In [6]: s = pd.Series([1, 2, 3], dtype='timedelta64[ns]')

In [7]: i == s
Out[7]: array([False, False, False])

Should this follow __pandas_priority__ now and return Series[bool]?

[jbrockmendel]

xref #36759

[mroeschke]

xref #36759

Thanks. Any thoughts on whether EA == other should return a pandas object?

[jbrockmendel]

Any thoughts on whether EA == other should return a pandas object?

I'd assume that we would use __pandas_priority__ to determine whether to return NotImplemented. If not returning NotImplemented, I'd expect to return either ndarray[bool] or EA[bool-ish]