bug: bidirectional inference when using typeguard
- Dominant language
- Python
- Stars
- 15.6k
- Forks
- 1.8k
- Avg merge
- 12h 13m
- Merged PRs (30d)
- 52
Description
I can't see why pyright doesn't complain at `return g` here. I had expected it would say that `Callable[P, R]` isn't necessarily `F`.
```
from typing import Callable
from typing import TypeGuard
class DecoratorProblem[F, **P, R]:
def is_pr(self, f: object) -> TypeGuard[Callable[P, R]]:
return isinstance(f, Callable)
def decorator(self, f: F) -> F:
assert self.is_pr(f)
g = self._decorator(f)
return g # pyright allows Callable[P, R] to be assigned to F? Why?
def _decorator(self, f: Callable[P, R]) -> Callable[P, R]:
def wrapper(*args: P.args, **kwargs: P.kwargs) -> R:
return f(*args, **kwargs)
return wrapper
```
see also: https://github.com/microsoft/pyright/discussions/10724
Contributor guide
Research direction
Use the Python code sample as the minimal reproduction and run it through pyright, focusing on the type accepted at `return g` after the `TypeGuard` assertion. Read the related discussion for context, then trace the bidirectional inference involved and add a regression test that verifies whether the expected diagnostic is reported.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- tooling
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100