Add explicit type parameters to arguments of df.apply

[JanEricNitschke]

• Closes #xxxx (Replace xxxx with the Github issue number)
• Tests added: Please use assert_type() to assert the type of any return value

Adresses: #952

Currently just added the explicit Any annotations everywhere, moved ListLikeExceptSeriesAndStr from a TypeVar to a Union and verified that more tests pass pyright strict mode.

However i still have a confusion about the overloads with Mapping. Isnt that just any generic object?

[Dr-Irv]

thanks @JanEricNitschke

[Dr-Irv]

However i still have a confusion about the overloads with Mapping. Isnt that just any generic object?

I don't understand your question here.

[JanEricNitschke]

Here for example:

    @overload
    def apply(
        self,
        f: Callable[..., ListLikeExceptSeriesAndStr | Series[Any] | Mapping[Any, Any]],
        axis: Axis = ...,
        raw: _bool = ...,
        args: Any = ...,
        *,
        result_type: Literal["expand"],
        **kwargs: Any,

If i do

def returns_dict(x: pd.Series) -> dict[str, int]:
    return {"col4": 7, "col5": 8}

df = pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4], "col3": [5, 6]})
print(df.apply(returns_dict))

col1    {'col4': 7, 'col5': 8}
col2    {'col4': 7, 'col5': 8}
col3    {'col4': 7, 'col5': 8}

I just get a series where each entry is this dict object. The mappings there just made me feel like it should be something like Mapping[S1, S1] -> Series[S2] or somethinglike that, but it Mapping -> Series[Mapping]

[Dr-Irv]

I just get a series where each entry is this dict object. The mappings there just made me feel like it should be something like Mapping[S1, S1] -> Series[S2] or somethinglike that, but it Mapping -> Series[Mapping]

You could do Mapping[Any, S1] and have it return Series[S1], but you'd need to then have to separate the overloads for Mapping from the existing ones, and also you'd have to add another overload of Mapping[Any, Any] -> Series[Any] as well.

The key idea in the stubs is that we define S1 to be the "regular" types of things that we see in Series, and where it is useful to make a differentiation. While it is possible to put any object into a Series, we didn't include object in S1. That might be a change to make for cases like these. You'd possibly end up with overlapping overloads that mypy won't be happy with, but we can just #type: ignore statements for that.