Changed SeriesGroupBy to return Series

[ramvikrams]

• Closes SeriesGroupBy.value_counts() should return Series not DataFrame #442
• Tests added: Please use assert_type() to assert the type of any return value

[gandhis1]

Why do we need two overloads?

[bashtage]

Should not need two overloads when both return Series.

[ramvikrams]

I updated the file thanks

[Dr-Irv]

Actually there should be two overloads here, and I realized this after merging. @ramvikrams can you open a new PR with this:

    @overload
    def value_counts(
        self,
        normalize: Literal[False] = ...,
        sort: bool = ...,
        ascending: bool = ...,
        bins=...,
        dropna: bool = ...,
    ) -> Series[int]: ...
    @overload
    def value_counts(
        self,
        normalize: Literal[True]
        sort: bool = ...,
        ascending: bool = ...,
        bins=...,
        dropna: bool = ...,
    ) -> Series[float]: ...

The changes from what you initially had are not doing = ... for normalize in the second overload, and adding the subtype as the return type of the two overloads.

In the test, don't do c: pd.Series =, and modify the test to check that it returns Series[int], and then add a test with normalize=True and check thtat it returns Series[float]

[ramvikrams]

yes sir will do that

[gandhis1]

There is a typo here. And why is it in frame tests? There are no similar tests to put it with?

[bashtage]

Also, please use standard python function naming, e.g., test_series_groupby_and_value_counts.

[ramvikrams]

yes sir

[Dr-Irv]

thanks @ramvikrams

IN the future, please name your branches related to what you are fixing

[ramvikrams]

thanks @ramvikrams

IN the future, please name your branches related to what you are fixing

ok sir will surely do that
thank you @Dr-Irv

[MarcoGorelli]

But should copy overloads from DataFrameGroupBy for value_counts() to handle normalize argument correctly.

sorry @Dr-Irv , I'm confused - shouldn't overloads have been added to return Series[float] if normalize=True and Series[int] otherwise?

[Dr-Irv]

But should copy overloads from DataFrameGroupBy for value_counts() to handle normalize argument correctly.

sorry @Dr-Irv , I'm confused - shouldn't overloads have been added to return Series[float] if normalize=True and Series[int] otherwise?

Yes, and I think I was typing the same message while you were typing yours!