Repository navigation
Regression in 1.2.0: callback protocol with Callable and Concatenate #15065
Description
Bug Report
The code below passed with mypy 1.1.1 but not in 1.2.0
To Reproduce
from typing import Concatenate, Callable, ParamSpec, Protocol
P = ParamSpec("P")
class Decorator(Protocol):
def __call__(self, fct: Callable[Concatenate[str, P], int]) -> Callable[P, int]:
...
def foo(value: int) -> Decorator:
def bar(fct: Callable[Concatenate[str, P], int]) -> Callable[P, int]:
raise NotImplementedError
return barI haven't been able to strip down any further, it seems that it has something to do with both Callable and Concatenate.
Expected Behavior
No error (as in mypy 1.1.1).
Actual Behavior
The following error is displayed:
issue.py:15: error: Incompatible return value type (got "Callable[[Callable[[str, **P], int]], Callable[P, int]]", expected "Decorator") [return-value]
issue.py:15: note: "Decorator.__call__" has type "Callable[[Arg(Callable[[str, **P], int], 'fct')], Callable[P, int]]"
Found 1 error in 1 file (checked 1 source file)
Note that I have tried to change the Decorator definition to make fct parameter positional-only:
class Decorator(Protocol):
def __call__(self, fct: Callable[Concatenate[str, P], int], /) -> Callable[P, int]:
...Which resulted in the error message being even more confusing:
issue.py:15: error: Incompatible return value type (got "Callable[[Callable[[str, **P], int]], Callable[P, int]]", expected "Decorator") [return-value]
issue.py:15: note: "Decorator.__call__" has type "Callable[[Callable[[str, **P], int]], Callable[P, int]]"
Found 1 error in 1 file (checked 1 source file)
Your Environment
- Mypy version used: 1.2.0
- Mypy command-line flags: none
- Mypy configuration options from
mypy.ini(and other config files): none - Python version used: 3.11.3
Activity
Rogdham commentedon Apr 16, 2023
Possibly related to #15037 in which the same error arises without the callback protocol part, like so:
AlexWaygood commentedon Apr 16, 2023
This bisects to 730ba8a (cc. @A5rocks)
AlexWaygood commentedon Apr 16, 2023
But #14903 (by @A5rocks) appears to fix the issue! (The PR has not yet been merged, and looks like it still has a few issues for now -- e.g. there's a crash in the latest
mypy_primerdiff.)Rogdham commentedon Apr 16, 2023
Thank you @AlexWaygood for your analysis!
In that case it seems to confirm that it is a duplicate of #15037. But at least it gives @A5rocks an other case (the example from the description of the PR uses a
Protocol) to test their PR with!Reconsider constraints involving parameter specifications (#15272)
abid-mujtaba commentedon Aug 10, 2023
Looks like the fix was reverted: #15272 (comment). Can we please re-open this issue so that we can keep track of the problem? Thanks.
A5rocks commentedon Aug 10, 2023
Fortunately, the PR it was reverted for fixes this too.
Namely, #15837