TYP: sort_index

[twoertwein]

Tested the overloads with mypy and pyright:

import pandas as pd

def bool_() -> bool:
    ...

frame = pd.DataFrame()
reveal_type(frame.sort_index()) # DataFrame
reveal_type(frame.sort_index(inplace=True)) # None
reveal_type(frame.sort_index(inplace=False)) # DataFrame
reveal_type(frame.sort_index(inplace=bool_())) # DataFrame | None

series = pd.Series()
reveal_type(series.sort_index()) # Series
reveal_type(series.sort_index(inplace=True)) # None
reveal_type(series.sort_index(inplace=False)) # Series
reveal_type(series.sort_index(inplace=bool_())) # Series | None

[Dr-Irv]

make the types consistent (bool vs. bool_t) between generic.py and series.py, frame.py

[twoertwein]

as far as I remember, pyright doesn't like self annotations that are broader than the class itself (cannot use NDFrameT).

[Dr-Irv]

Questions about the ascending argument and typing self on all overloads.

[simonjayhawkins]

Thanks @twoertwein generally lgtm.

[simonjayhawkins]

Thanks @twoertwein

The requested changes were to keep the scope of this PR narrow and for consistency with the existing code. (and not necessarily the correct thing to do)

#32638 could potentially discuss changes on how the _constructor mechanism works. We maybe need a more limited discussion from a typing perspective considering a trade-off for false positives and more refined return types for the public api.

Also a similar discussion is maybe needed considering a trade-off for false positives and catching more errors using Literal for the Axis parameter in the public api.

I'll leave it to someone else to champion these issues as I try to backseat on the public api typing issues as it generally needs a different approach to the internal typing of the codebase and is therefore difficult to not give out mixed messages on typing policy.

[jreback]

can you merge master

[twoertwein]

green, except for mypy error also present on main

[jreback]

thanks @twoertwein (and others!)