DataFrame.__getitem__ returns wrong type for a dict_keys object

[gandhis1]

import pandas as pd

some_columns = {
    "a": 1,
    "b": 2,
}
x  = pd.DataFrame.from_dict(some_columns)
y = x[some_columns.keys()]
reveal_type(y)  # note: Revealed type is "pandas.core.series.Series[Any]"

__getitem__ with a Generator is already annotated correctly, not clear to me why it doesn't use that overload here.

[twoertwein]

Generator requires a few different methods. I assume dict.keys() is Hashable and therefore the first matching overload is the Series overload.

Assuming str is the only iterable that we want to special case (@Dr-Irv), we could first have an overload for str, and then use Iterable (covers, lists, generators and dict.keys) but add an ignore comment (overlapping overloads).

There might be more places were we could first special case the exceptions, e.g., str, to then replace large list-like Unions with Iterable+ignore.

[Dr-Irv]

Assuming str is the only iterable that we want to special case (@Dr-Irv), we could first have an overload for str, and then use Iterable (covers, lists, generators and dict.keys) but add an ignore comment (overlapping overloads).

No issue for me on that. PR welcome