Add overload for DataFrameGroupBy.groupby("size") return Series

[ljmc-github]

Fairly straightforward, I had to add ignores for mypy and pyright since "size" is included in AggFuncTypeFrame's str.

I made as few changes as possible leaving agg = aggregate which seems to work, although I saw the SeriesGroupBy.agg is a redeclaration rather than assignment, I am happy to switch if necessary.

pandas-stubs/pandas-stubs/core/groupby/generic.pyi

Lines 53 to 60 in 6fe90bb

	@overload
	def aggregate(self, func: list[AggFuncTypeBase], *args, **kwargs) -> DataFrame: ...
	@overload
	def aggregate(self, func: AggFuncTypeBase, *args, **kwargs) -> Series: ...
	@overload
	def agg(self, func: list[AggFuncTypeBase], *args, **kwargs) -> DataFrame: ...
	@overload
	def agg(self, func: AggFuncTypeBase, *args, **kwargs) -> Series: ...

• Closes DataFrameGroupBy.aggregate can return Series with argument "size", typed as only returning DataFrame #736
• Tests added: Please use assert_type() to assert the type of any return value

[Dr-Irv]

Thanks for the PR. You wrote:

although I saw the SeriesGroupBy.agg is a redeclaration rather than assignment, I am happy to switch if necessary.

Actually, this was an oversight from a previous PR 10 months ago. I think I'd rather use assignment in SeriesGroupBy, so can you make that change there so we are consistent? You don't have to add a test for that.

[Dr-Irv]

Can you move these tests to the function test_types_groupby_agg() ?

[ljmc-github]

Done, what about the other aggregates/transforms in test_types_groupby ?

pandas-stubs/tests/test_frame.py

Lines 870 to 881 in 6fe90bb

	df1: pd.DataFrame = df.groupby(by="col1").agg("sum")
	df2: pd.DataFrame = df.groupby(level="ind").aggregate("sum")
	df3: pd.DataFrame = df.groupby(by="col1", sort=False, as_index=True).transform(
	lambda x: x.max()
	)
	df4: pd.DataFrame = df.groupby(by=["col1", "col2"]).count()
	df5: pd.DataFrame = df.groupby(by=["col1", "col2"]).filter(lambda x: x["col1"] > 0)
	df6: pd.DataFrame = df.groupby(by=["col1", "col2"]).nunique()
	df7: pd.DataFrame = df.groupby(by="col1").apply(sum)
	df8: pd.DataFrame = df.groupby("col1").transform("sum")
	s1: pd.Series = df.set_index("col1")["col2"]
	s2: pd.Series = s1.groupby("col1").transform("sum")

[Dr-Irv]

If you want, you could add extra tests in another PR, but let's leave this PR to focus on just the one issue. Thanks.

[Dr-Irv]

Better to say # error: overload 1 overlaps overload 2 because of different return types

[Dr-Irv]

thanks @ljmc-github