Skip to content

Unexpected behavior with dataclasses and weakref #102069

Closed
@jbaudisch

Description

@jbaudisch

We recently discovered the following error in our project:
TypeError: descriptor '__weakref__' for 'XYZ' objects doesn't apply to a 'XYZ' object

XYZ in our case is a dataclass which uses slots (added v3.10) and weakref_slot (added v3.11) parameters.

We further investigated the error and have come to the following example to reproduce it:

import dataclasses
import weakref


@dataclasses.dataclass(slots=True, weakref_slot=True)
class EntityAsDataclass:
    x: int = 10


class Entity:
    __slots__ = ("x", "__weakref__")

    def __init__(self, x: int = 10) -> None:
        self.x = x


e1 = Entity()
e1_ref = weakref.ref(e1)

e2 = EntityAsDataclass()
e2_ref = weakref.ref(e2)


assert e1.__weakref__ is e1_ref  # works
assert e2.__weakref__ is e2_ref  # fails with "TypeError: descriptor '__weakref__' for 'EntityAsDataclass' objects doesn't apply to a 'EntityAsDataclass' object"

I don't know if this is intended, but we expect EntityAsDataclass and Entity to have the same behavior here.

Thanks!

Linked PRs

Metadata

Metadata

Assignees

Labels

3.11only security fixesstdlibPython modules in the Lib dirtype-bugAn unexpected behavior, bug, or error

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions