BUG: Exception to the boolean frame comparison guarantee

[ResidentMario]

The implementation for core._com_method_FRAME makes the following note:

            # straight boolean comparisions we want to allow all columns
            # (regardless of dtype to pass thru) See #4537 for discussion.

However, I think I stumbled on an exception here.

The root cause is that astype(bool) has a "hole" in it in the case of datetime dtypes. The following:

import pandas as pd
from io import StringIO

# create our surprisingly slippery frame
res = pd.read_csv(StringIO(""",dates1,dates2
0,1970-01-01 00:00:00.000000000,1970-01-01 00:00:00.000000000
1,1970-01-01 00:00:00.000000000,"""), index_col=0)
res['dates1'] = pd.to_datetime(res['dates1'])

res.astype(bool)

Raises:

TypeError: cannot astype a datetimelike from [datetime64[ns]] to [bool]

If you get this frame as a fillna output (not possible on master, but possible in the feature branch I am working on), it will raise this same error when passed through _com_method_FRAME, due to:

return res.fillna(True).astype(bool)

This is a problem in PR#15653.

[ResidentMario]

xref #4537

[ResidentMario]

The obvious solution would be to forcefully upcast any datetime columns to object dtype beforehand.

Highly inelegant, but perhaps necessary.

[ResidentMario]

(closing because I don't think this can happen in master, just wanted to create something I can reference later)