Sitelet https://github.com/python/cpython/issues/125316
Skip to content

Using functools.partial to make functions not considered as methods when given as enum values no longer works #125316

Description

@bluthej

Bug report

Bug description:

On python 3.10-3.12, the following snippet prints True, while on Python 3.13.0 it prints False:

from enum import Enum
from functools import partial


def a():
    pass


def b():
    pass


class MyCallables(Enum):
    A = a
    B = partial(b)


print(MyCallables.B in MyCallables)

This is because MyCallables.B is considered to be functools.partial(<function b at 0x756d7f0065c0>) on 3.13.0, when it used to be <MyCallables.B: functools.partial(<function b at 0x71f7cbc407c0>)> before.

This is something I've been using to allow function values in enums not to be considered methods, which I most likely got from this thread. In the example above, A would always be considered a method.

It seems like on 3.13.0, the "wrapper class" approach is the only one that works.

CPython versions tested on:

3.10, 3.11, 3.12, 3.13

Operating systems tested on:

Linux, Windows

Linked PRs

Activity

  1. ericvsmith commented on Oct 11, 2024

    @ericvsmith
    Member
  2. added
    stdlibStandard Library Python modules in the Lib/ directory
    3.13only security fixes
    3.14bugs and security fixes
    on Oct 11, 2024
  3. ZeroIntensity commented on Oct 11, 2024

    @ZeroIntensity
    Member

    Confirmed on current main as well, technically a 3.13 regression.

  4. serhiy-storchaka commented on Oct 11, 2024

    @serhiy-storchaka
    Member

    This is an intentional change, and it is documented. See #121027.

    Using staticmethod() to set a callable attribute without making it a method is more common idiom.

  5. bluthej commented on Oct 11, 2024

    @bluthej
    Author

    @serhiy-storchaka maybe my title was not well formulated. This is what my issue boiled down to in my mind but I realize this is not exactly my problem, my problem is with the behavior I laid out in the snippet.

    It used to be that we could use partial in order to use a function as the value for an enum variant while still having that variant act like one. With the breaking change you mentioned this is no longer possible, the enum variant is considered to be the function itself. But that's just a consequence (maybe even unintended, I don't know).

    At least there is still the wrapper class approach that works! I guess I'll just update the stackoverflow thread I mentioned in case people run into that.

  6. bluthej commented on Oct 11, 2024

    @bluthej
    Author

    I should probably add that using staticmethod in that context doesn't do the trick.

  7. serhiy-storchaka commented on Oct 11, 2024

    @serhiy-storchaka
    Member

    As was mentioned in other answer, there is a better solution -- use enum.member() (I did not know about it!).

  8. bluthej commented on Oct 11, 2024

    @bluthej
    Author

    Thanks for the tip! I had seen that answer but since it wasn't upvoted much I didn't try 😅

  9. added a commit that references this issue on Oct 12, 2024
  10. serhiy-storchaka commented on Oct 12, 2024

    @serhiy-storchaka
    Member

    #125361 should fix this for 3.13. It restores the old behavior of partial in enums and adds a runtime warning. In 3.14 the behavior will be changed in any case.

    We tried to emit a warning about future behavior change, but adding a warning already changed the behavior in some cases.

  11. ZeroIntensity commented on Oct 14, 2024

    @ZeroIntensity
    Member

    @hacscred This kind of feedback isn't helpful--we don't intentionally break code, this is a regression. Please read PEP-387 for our backwards compatibility policy.

  12. added a commit that references this issue on Oct 21, 2024
  13. MegaIng commented on Oct 25, 2024

    @MegaIng

    This kind of feedback isn't helpful--we don't intentionally break code, this is a regression. Please read PEP-387 for our backwards compatibility policy.

    No, this is clearly wrong. #121089 was an intentional break in backwards compatibility, disregarding PEP-387 with the argument that "this will only rarely break code".

  14. ZeroIntensity commented on Oct 25, 2024

    @ZeroIntensity
    Member

    Yes, but this specific case with Enum was a regression and was fixed for this case. If that PR is causing more problems, open a new issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    3.13only security fixesstdlibStandard Library Python modules in the Lib/ directorytype-bugAn unexpected behavior, bug, or error

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions