Repository navigation
False positive for Final attributes using typevars in dataclasses #21637
Description
Update: The false positive happens for non-dataclasses as well; see #21637 (comment).
Bug Report
In a dataclass, annotating an instance attribute with a Final that depends on a type variable gives a false positive Final name declared in class body cannot depend on type variables [misc] error.
I think that this should be allowed for instance attributes (but not class attributes). Per the spec:
A
Finaldataclass field initialized in the class body is not a class attribute unless explicitly annotated withClassVar.
To Reproduce
from dataclasses import dataclass
from typing import Final
from typing import Generic
from typing import TypeVar
T = TypeVar("T")
@dataclass(frozen=True)
class A(Generic[T]):
value: Final[T] # error: Final name declared in class body cannot depend on type variables [misc]
@dataclass(frozen=True)
class B(Generic[T]):
value: T(https://mypy-play.net/?mypy=2.1.0&python=3.14&gist=996c4b8d68e8b57e292b49a834f4ac24)
Expected Behavior
No error.
Actual Behavior
main.py:11: error: Final name declared in class body cannot depend on type variables [misc]
Your Environment
- Mypy version used: 2.1.0
- Mypy command-line flags: none for reproducer
- Mypy configuration options from
mypy.ini(and other config files): none for reproducer - Python version used: 3.14
Activity
A5rocks commentedon Jun 26, 2026
It seems to me this has nothing to do with dataclasses, the same error happens for the innocuous analog:
I don't see why
Final[int]should be treated as an instance variable butFinal[T]should not...gschaffner commentedon Jul 1, 2026
Finalin a class body (except in a dataclass class body) means final class variable, so AFAIK, in that non-dataclass example, the error is a true positive because a class variable cannot depend on the class's type variable.Relevant snippet of the spec:
A5rocks commentedon Jul 1, 2026
But it's not initialized in the class body!
gschaffner commentedon Jul 1, 2026
Oh, right! Oops.
[-]False positive for `Final` attributes using typevars in dataclasses[/-][+]False positive for `Final` instance attributes using typevars[/+]gschaffner commentedon Jul 1, 2026
I'm not sure what maintainers prefer for tracking here: should we close this as a duplicate of #8982 and add a note in #8982 that it's broken for dataclasses as well?
A5rocks commentedon Jul 1, 2026
Hm might as well keep this open -- dataclass
Finalthings probably needs extra code to support. Let's see if I can mark it as a child issue :^)[-]False positive for `Final` instance attributes using typevars[/-][+]False positive for `Final` attributes using typevars in dataclasses[/+]adhavan18 commentedon Jul 16, 2026
A fix for the dataclass case is up in #21643 (scoped per the discussion above — plain generic class bodies still error, tracked by #8982). CI is green, mypy_primer reports no corpus changes, and it includes tests for both the dataclass-allowed and non-dataclass-still-errors cases. Review welcome!