Make it possible for type checking to detect "The truth value of a DataFrame is ambiguous."

[karolzlot]

I described the problem here:
microsoft/pyright#5024
and also here:
microsoft/pylance-release#4289

[gandhis1]

Maybe annotate __bool__ on pd.DataFrame and pd.Series to return typing.Never ?

[Dr-Irv]

I think the solution here is that DataFrame.__eq__() needs to have a stub in core/frame.pyi returning a DataFrame

[gandhis1]

I think the solution here is that DataFrame.__eq__() needs to have a stub in core/frame.pyi returning a DataFrame

I'm not sure I follow what that will accomplish. This error is triggered when someone does the equivalent of casting a DataFrame to a bool - which calls __bool__, no?

Even pandas itself has it annotated the way I suggested:

    @final
    def __nonzero__(self) -> NoReturn:
        raise ValueError(
            f"The truth value of a {type(self).__name__} is ambiguous. "
            "Use a.empty, a.bool(), a.item(), a.any() or a.all()."
        )

    __bool__ = __nonzero__

[Dr-Irv]

I'm not sure I follow what that will accomplish. This error is triggered when someone does the equivalent of casting a DataFrame to a bool - which calls __bool__, no?

You're right, and actually DataFrame.__eq__() is return a DataFrame already.

So I tried with a simple class, and it didn't work with pyright. See microsoft/pyright#5039

[Dr-Irv]

pyright fixed the issue, in 1.1.306 so we should define __bool__() for Series and DataFrame as NoReturn, and add some tests.