ClassVar[Final[T]] should be allowed

[ben-alexander]

Pyright doesn't allow the construction Final[ClassVar[T]], typing.ClassVar states:

Changed in version 3.13: ClassVar can now be nested in Final and vice versa.

from typing import ClassVar, Final

class A:
    var: ClassVar[str] = 'a'

class B(A):
    var: Final[ClassVar[str]] = 'b'

# "ClassVar" is not allowed in this context Pylance(reportInvalidTypeForm)

There was a previous report #7959 specific to dataclasses from before the release of 3.13. I'm not sure where this restriction to dataclasses comes from as the typing spec doesn't mention this in 3.12.

Running Pylance in VS Code. Version: 2025.5.1

[erictraut]

Thanks, I agree this is a bug. There's actually two bugs demonstrated in your code. The first is that pyright generates a "ClassVar not allowed in this context" diagnostic. The second is that pyright doesn't generate a diagnostic for attempting to override a non-final attribute with a final attribute. That combination violates the Liskov Substitution Principle.

[Viicos]

My understanding is that using ClassVar with Final was explicitly allowed for dataclasses in python/typing#1669, then followed by runtime changes to allow this in python/cpython#116096. Finally, it was documented in python/cpython#116686 (mentioning the same issue, scoped to dataclasses).

Should the spec be updated to explicitly describe what it means for Final and ClassVar to be used together, in dataclasses and arbitrary classes?

[erictraut]

Thanks for the reminder. The typing spec was updated to accommodate a combination of Final and ClassVar within dataclasses. It says here:

Type checkers should infer a final attribute that is initialized in a class body as being a class variable, except in the case of Dataclasses, where x: Final[int] = 3 creates a dataclass field and instance-level final attribute x with default value 3; x: ClassVar[Final[int]] = 3 is necessary to create a final class variable with value 3. In non-dataclasses, combining ClassVar and Final is redundant, and type checkers may choose to warn or error on the redundancy.

So pyright is actually correct here in indicating that ClassVar can't be used in conjunction with Final because Final implies ClassVar in this situation. If you remove the redundant ClassVar, then pyright does correctly detect and report the attempt to override a non-Final ClassVar with a Final one.

That means pyright is conforming with the spec here. I'm therefore going to close the issue.

Should the spec be updated to explicitly describe what it means for Final and ClassVar to be used together, in dataclasses and arbitrary classes?

As mentioned above, the spec is already pretty clear on this point.