Fix annotations for pynput event callbacks

[huntfx]

The event callbacks are designed to accept any number of arguments up to the maximum amount. Also, currently, the injected argument is omitted from all the pyi files, so it'll flag up with an error if you use it.

For example, on_click(), on_click(x, y) and on_click(x, y, button, pressed, injected) are all perfectly valid callbacks.

I've done a union with all possible signatures for the event callbacks, it's a little more verbose, but I'm not aware of another way to handle this.

For reference, here are the full signatures for each callback:

• on_move(x: int, y: int, injected: bool)
• on_click(x: int, y: int, button: Button, pressed: bool, injected: bool)
• on_scroll(x: int, y: int, dx: int, dy: int, injected: bool)
• on_press(key: Key | KeyCode | None, injected: bool)
• on_release(key: Key | KeyCode | None, injected: bool)

The test passes when running python tests/regr_test.py pynput. Fixes #15855.

[srittau]

Not a full review, just two things I noticed:

• We only need/want tests for particular tricky typing situations (for various reasons). In this case, the annotations are a bit unusual, but not particularly tricky, so the tests are not needed.
• I know this was a pre-existing problem, but callbacks returning None is usually not correct. If the return value of the callback is ignored, we use object instead.

(Also I respect the voodoo magic going on that supports callbacks with varying amounts of arguments.)

[huntfx]

Alright cheers, I'll double check the return type and commit those changes then

[huntfx]

Awkward one with the return value actually. If it's specifically False then it'll raise a StopIteration, but anything else and it works fine.

Not actually sure if that's still best left to object, or something more specific in this case?

https://github.com/moses-palmer/pynput/blob/afc64577b069a51a91474a235f0efbb0e7e0357d/lib/pynput/_util/__init__.py

        def wrapper(f):
            def inner(*args):
                if f(*args) is False:
                    raise self.StopException()

            return inner

[srittau]

Interesting. Maybe using bool | None is best in this case? While technically returning anything else would work (which is why we usually use object), in this case I feel that disallowing arbitrary return types would be best.

[huntfx]

Yeah that makes sense I reckon - I also realised there was a mismatch with the keyboard ones so I've just updated them to match.

[srittau]

Thanks!

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉