Moving monkeypatches into celery

[dlucredativ]

The monkeypatches mentioned in https://github.com/sbdchd/celery-types/blob/main/README.md#install and #157 (comment) are necessary to get rid of Need type annotation for "task" [var-annotated] typecheck errors. There is an open pull-request to get rid of the monkeypatches in the long run, but as I mentioned in that PR, I wonder why the method should not return a Generic.

I suspect that the correct monkeypatch would have been

Celery.__class_getitem__ = classmethod(types.GenericAlias)  # type: ignore[attr-defined]
Task.__class_getitem__ = classmethod(types.GenericAlias)  # type: ignore[attr-defined]
...

Can someone please clarify?

[9en9i]

I think this option was just the first one that came to mind, and it turned out to be workable. It doesn't break anything in 99.99% of use cases.
But yes, according to the documentation (https://docs.python.org/3/reference/datamodel.html#class-getitem-versus-getitem), this method should return GenericAlias.

[brianhelba]

I don't understand the rationale for the list of monkeypatched classes in the celery-types docs. For instance, FallbackContext isn't generic.

[dlucredativ]

I don't understand the rationale for the list of monkeypatched classes in the celery-types docs. For instance, FallbackContext isn't generic.

You mean semantically? celery-types alleges that FallbackContext is generic by declaring it as such in https://github.com/sbdchd/celery-types/blob/main/celery-stubs/utils/objects.pyi#L23-L26

[9en9i]

FallbackContext implements at least the AbstractContextManager protocol, which is generic, so FallbackContext itself is also generic.

[brianhelba]

@dlucredativ @9en9i Thanks for the correction. I think I was looking in too many places and confused myself.

[brianhelba]

I've created #214 to fully resolve this issue.