itertuples

[twoertwein]

• Closes With the most recent pandas-stubs, mypy now complains about every call to itertuples() #834 (Replace xxxx with the Github issue number)
• Tests added: Please use assert_type() to assert the type of any return value

[Dr-Irv]

I'm wondering if we could do something like this:

class _PandasNamedTuple(tuple[Any,...]):
    def __getattr__(self, field: str) -> Any: ...

Then have itertuples() return Iterable[_PandasNamedTuple]

Then at least you know you're getting a tuple.

[twoertwein]

Nice solution!

[Dr-Irv]

Needs to come from typing_extensions. Failed under python 3.9

[Dr-Irv]

Can you look at test_types_itertuples() in test_frame.py and see if that should be modified? Maybe introduce check(assert_type in there just to be sure.

[Dr-Irv]

I'm thinking we should return Scalar here, because we also return Scalar when someone does df.loc[3, "a"]

While it's true that some non-scalar value could be an individual element of a DataFrame, I've taken the philosophy of limiting the types to what is "normal" usage, and if you put a funky type in a DataFrame or Series, then you can do a cast to fix it. I've done that in some of our application code when we have lists or other objects inside a Series or DataFrame.

[Dr-Irv]

thanks @twoertwein