fix np.interp type #5231

[asukaminato0721]

Summary

Fixes #5231

add missing __float__ and np.interp

• I am a human, and I fully understand the code I am submitting.

Test Plan

add test

[stroxler]

Overall this looks like a good fix, thanks!

Back to you with a suggested improvement in overload structure; also the PR import didn't work so pushing an updated version will probably help me merge it :)

[stroxler]

Can we change this overload to

@overload
def interp[N: IntVar, Rest: _Shape = []](
    x: ndarray[[N, *Rest]]
    | RegularNestedList[[N, *Rest], _ArrayScalar],
    ...
) -> ndarray[[N, *Rest]]: ...

so that we rule out matching an empty shape here?

I think if we make that change it might also be able to combine the first and last overloads into just one, because we will have solved the the overlapping-overloads problem.

The current diff uses ordering to try to fix the overlap, but python overload resolution is only partially and ambiguously order-dependent, so it's much cleaner if we can just make the overloads non-overlapping.

[stroxler]

No auto-import unfortunately :/

I'll try to get this reviewed and merged, thanks!

[meta-codesync]

@stroxler has imported this pull request. If you are a Meta employee, you can view this in D125328983. (Because this pull request was imported automatically, there will not be any future comments.)

[rchen152]

Review automatically exported from Phabricator review in Meta.

[meta-codesync]

@stroxler merged this pull request in 0a9dca8.