is it intended behavior that an attribute shows up in __annotations__ if it is later redefined by a function?

[zzzeek]

someone here stumbled upon this oddity:

from __future__ import annotations

import annotationlib

class Foo:
    baz: int
    def baz(self) -> bob:
        pass

print(Foo.__annotations__)
print(annotationlib.get_annotations(Foo))

Because Foo.baz is typed as int, it shows up in __annotations__ and get_annotations() even though it was replaced by a function:

{'baz': 'int'}
{'baz': 'int'}

is that....a defined behavior? Can a library give this pattern to users and rely upon it working? use case is we are creating dataclasses where some of the attributes come from a descriptor, which normally is not in __annotations__ so we have to stuff that value into a custom __annotations__, which we'd rather not do.

[TeamSpen210]

That's perfectly valid runtime-wise, though it's a type error (a function is not assignable to int). It's the same as this sort of code:

value: list[str] | None
if something:
    value = ["a", "b"]
else:
    value = None

Assignments don't replace annotations. That said this probably isn't the best idea since type checkers are going to complain.

[zzzeek]

ah right it is like an assigment, so OK makes sense

yeah im not going to actually use this for anything at this point, just seemed curious thanks!