BUG: Raise TypeError when subracting DateTimeArray and other date types

[KevsterAmp]

• closes BUG: provide better error message for pd.Timedelta - pd.Series[Timestamp] #59571 (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

[KevsterAmp]

@rhshadrach - can you verify if my implementation here is correct? I'm not sure of this. A bit of guidance could be useful. Thanks

[rhshadrach]

I think this is looking good, CI failures look unrelated. They'll be fixed outside of this PR.

[KevsterAmp]

        from pandas.core.arrays import DatetimeArray

        datetime_result = self - other
        if isinstance(datetime_result, DatetimeArray):
            raise TypeError(
                "TypeError: unsupported operand type(s) for -: "
                f"'{type(self).__name__}' and '{type(other).__name__}'"
            )
        return -(datetime_result)

I added an outside top-level import because I am experiencing Unbound errors with Datetimearray in this line exactly:

        if isinstance(datetime_result, DatetimeArray):

Also, I replaced the messages with the variable type as any wildcard .*. Because I had trouble matching the exact dtype of the variable vs the error output. Seems like the dtype gets converted. For example, there are instances where timedelta variable error message contains timedelta64 but there are cases where it returns timedelta as normal.

            r"TypeError: unsupported operand type\(s\) for -: 'DatetimeArray' and '.*'"

[rhshadrach]

lgtm

[KevsterAmp]

@mroeschke - could you review this. Thanks

[mroeschke]

I'm a little nervous enforcing a particular type out of a rsub operation.

Could we just catch the TypeError and reraise it with a better message?

[KevsterAmp]

Gotcha, I agree that it would be a better implementation

[KevsterAmp]

My implementation raises re-raises new message in TypeError on DateTimeArrays only.

Since there are already other error messages for other dtypes like

Cannot subtract tz-naive and tz-aware datetime-like objects.

If I implement it as

try:
    return -(self - other)
except TypeError as e:
    raise TypeError(
        "Unsupported operand type(s) for -: "
        f"'{type(self).__name__}' and '{type(other).__name__}'"
    ) from e

I would have to fix all tests related to raising tz-naive/tz-aware datetime-like objects. Thoughts?

[KevsterAmp]

Looks like my previous commit affects other datetime related operations.

Could you give me ideas on catching the error? From my understanding, when a datetime objects go into this block

return -(self - other)

It will always raise a TypeError. So my previous implementation catches it before doing the operation.

        # We get here with e.g. datetime objects
        from pandas.core.arrays import DatetimeArray

        datetime_result = self - other
        if isinstance(datetime_result, DatetimeArray):
            raise TypeError(
                "Unsupported operand type(s) for -: "
                f"'{type(self).__name__}' and '{type(other).__name__}'"
            )
        return -(datetime_result)

Though from checking the CI failures, I can't seem to trace/ I'm not familiar how my latest commit affected them.

Guidance would be helpful. Thank you!

[mroeschke]

I don't think you need this if statement here. We'd always want to raise the exception