`Callable` is a class

[KotlinIsland]

Summary

from collections.abc import Callable

t: type = Callable  # Object of type `typing.Callable` is not assignable to `type`

this stems from an issue with typeshed

[AlexWaygood]

This is only true for collections.abc.Callable, not for typing.Callable:

>>> import collections.abc, typing
>>> isinstance(collections.abc.Callable, type)
True
>>> isinstance(typing.Callable, type)
False
>>> collections.abc.Callable.__mro__
(<class 'collections.abc.Callable'>, <class 'object'>)
>>> typing.Callable.__mro__
Traceback (most recent call last):
  File "<python-input-4>", line 1, in <module>
    typing.Callable.__mro__
  File "/Users/alexw/.pyenv/versions/3.13.1/lib/python3.13/typing.py", line 1366, in __getattr__
    raise AttributeError(attr)
AttributeError: __mro__. Did you mean: '__or__'?

According to typeshed, typing.Callable and collections.abc.Callable are the same object, and it would require quite deep special casing from us to override typeshed on that point. So I'm not sure this is worth fixing.

[KotlinIsland]

hmmm

from collections.abc import Callable

class A(Callable[int]): # no error
    pass

• Callable not validated properly as a base class #1220

[KotlinIsland]

@AlexWaygood would it be worth updating the vendored typeshed? if so i'd be happy to do it

[AlexWaygood]

@AlexWaygood would it be worth updating the vendored typeshed? if so i'd be happy to do it

I think that would be likely to either break existing type checkers, or else have somewhat unexpected effects on their handling of Callable. I'm not sure it's worth the disruption to the ecosystem.

Is there real-world code where this causes issues? What's the use case needing to consider collections.abc.Callable (but not typing.Callable) an instance of type

[KotlinIsland]

I think that would be likely to either break existing type checkers

i mean, ty's vendored typeshed, not the upstream one

Is there real-world code where this causes issues?

i have run into it in the past where the typing definition of Callable has caused issues, i think there is some special handling in ty already to avoid this as class C(Callable) isn't reported. imo it's better to iron out this incorrect definition at the root

[AlexWaygood]

i mean, ty's vendored typeshed, not the upstream one

I see — no, I'd strongly prefer it if we can avoid patching upstream typeshed unless we have a very good reason for it. The patches mypy has applied to typeshed have been a real maintenance headache for them and made syncing typeshed much more complicated for them than it should be because of the frequent merge conflicts. I don't want for us to get into a similar situation.

Our handling of special forms is quite different to most other type checkers, as you've already observed, so I think it's unlikely that we'll have the same bugs that you've observed with mypy. #1220 relates to a branch of code in ty that is already marked as TODO. We actually infer a Todo type currently (which is similar to Unknown, but is explicit about the fact that there's still some missing logic in ty) when Callable is present as a base class.

So all in all, I don't see sufficient motivation to make any changes with regards to this one. But thanks for the report!

[KotlinIsland]

okay, i agree

i've opened: python/typeshed#14761