gh-128335: Make slice generic at runtime

[Gobot1234]

• Issue: Make slice generic #128335

[decorator-factory]

This docstring seems misleading, as PEP 585 doesn't mention the slice class.

(Maybe PEP 585 could refer to https://typing.readthedocs.io/en/latest/guides/modernizing.html#generics-in-the-typing-module? Right now it links to documentation for __class_getitem__ that on its own isn't a replacement for PEP 585 from a reader's perspective)

[Gobot1234]

Agreed it is maybe a little silly to be putting PEPs in these things when they aren't active docs.

[AlexWaygood]

PRs to the PEPs repo improving things like this are welcome!

[Gobot1234]

Curious if @AlexWaygood / @JelleZijlstra have opinions on this. This is done in other places. I seem to have time travelled here

[AlexWaygood]

I do agree that our docstrings probably also shouldn't be referencing PEPs, which are, as you both note, historical documents recording decisions that were made at one point in time rather than living documentation. But all the other __class_getitem__ docstrings say this currently, so I think it makes sense to follow suit here and change all the __class_getitem__ docstrings to something better all at once as a followup

[brianschubert]

I think this calls for a whatsnew entry. Something like

* The :class:`slice` type now supports subscription,
  making it a :term:`generic type`.
  (Contributed by James Hilton-Balfe in :gh:`128335`.)

We should probably also add a versionchanged to the docs noting that slice is now a generic type.

[JelleZijlstra]

This looks good to me but agree it needs to be documented.

[adamjstewart]

Is something like slice[float, float, int] also supported? This is necessary for, e.g., pandas.

[AlexWaygood]

Yes that should work fine. The number of arguments passed in isn't usually checked at runtime.