Bidirectional inference assigning symbols in other modules

[Andrej730]

pyright 1.1.401

Consider the snippet below - it seems bidirectional inference doesn't work when assigning symbols from other module.

In theory in the example below xxx.callback = lambda x: x.bit_count() should work out similar to callback = lambda x: x.bit_count() - by bydirectional inference, the type of lambda x: x.bit_count() is recognized as type of xxx.callback (Callable[[int], int]), x type to be recognized as int automatically and xxx.callback should keep it's original type after the assignment.

Same for xxx.lst = [].

xxx.py

from typing import Callable

callback: Callable[[int], int] = lambda x: x.bit_count()
lst: list[int] = []

Main script:

from typing import Callable
import xxx

class A:
    a: Callable[[int], int]

    @staticmethod
    def test(a: Callable[[int], int]): ...


callback: Callable[[int], int] = lambda x: x.bit_count()

# OK
A.test(lambda x: x.bit_count())
# OK
A.a = lambda x: x.bit_count()
# OK
callback = lambda x: x.bit_count()

# Bydirectional inference doesn't work.
# Type of "xxx.callback" is "(int) -> int"
reveal_type(xxx.callback)
xxx.callback = lambda x: x.bit_count()
# Type of "xxx.callback" is "(x: Unknown) -> Unknown"
reveal_type(xxx.callback)

# Type of "xxx.lst" is "list[int]"
reveal_type(xxx.lst)
xxx.lst = []
# Type of "xxx.lst" is "list[Unknown]"
reveal_type(xxx.lst)

---

Update: confirm issue as resolved.

[erictraut]

Thanks for the bug report. Your diagnosis was spot on.

This will be addressed in the next release.

[erictraut]

This is addressed in pyright 1.1.402.