Fix __doc__ on unbound method

[DanielNoord]

Type of Changes

	Type
✓	🐛 Bug fix

Description

Fixes the issue as described in #2847

The reproducer for the crash is:

class Executor:
    def submit(self): ...


class ThreadPoolExecutor(Executor):
    def submit(self): ...

    submit.__doc__ = Executor.submit.__doc__

That could be fixed in pylint (by not calling pytype() on NodeNG subclasses that don't have it), but that seems unnecessary.

Instead we can (I hope/think) better represent the actual object model by letting UnboundMethodModel inherit from FunctionModel. That feels like it would make sense? The test doesn't pass without this fix.

We could add attr___doc__ on UnboundMethodModel but for me this feels like it better fits at least my own mental model of Python objects 😅

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.30%. Comparing base (c75c0a0) to head (08fa1a3).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2958      +/-   ##
==========================================
- Coverage   93.31%   93.30%   -0.01%     
==========================================
  Files          92       92              
  Lines       11292    11295       +3     
==========================================
+ Hits        10537    10539       +2     
- Misses        755      756       +1     

Flag	Coverage Δ
linux	93.16% <100.00%> (-0.01%)	⬇️
pypy	93.30% <100.00%> (-0.01%)	⬇️
windows	93.28% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
astroid/interpreter/objectmodel.py	95.95% <100.00%> (-0.39%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[emmanuel-ferdman]

Hey @DanielNoord! I was just working on this same issue. I should've commented that but no worries 😄 .
I agree that UnboundMethodModel should extend FunctionModel. I would like to suggest this pr to also include fixes to two releated prolems that I think are still happening:

• ObjectModel-only dunders like __eq__, __hash__, __repr__ still return Unknown instead of Uninferable because FunctionModel inherits from ObjectModel and doesn't override those.

class A:
    def test(self): pass

A.test.__eq__  # Returns Unknown, not Uninferable

• super().__doc__ has the same problem.

class Base:
    def method(self):
        return super().__doc__

Base().method()  # Returns Unknown, not Uninferable

I think the fix can be adding Unknown filtering to UnboundMethod.getattr/igetattr and Super.igetattr (same pattern PR #2847 used for BaseInstance and ClassDef) so Unknown placeholders fall through to the proxied object. I have this implementation in a commit on my fork branch - feel free to take the changes if they're relevant to your PR, or let me know if I should open a separate PR to address the additional cases.

[DanielNoord]

@emmanuel-ferdman Let's open a different PR for your fix! It is indeed valuable, but I think we can apply both as they have somewhat different (but good) effects 😄

[jacobtylerwalls]

Looks good!

[DanielNoord]

@jacobtylerwalls Unrelated to this PR, but this action failed: https://github.com/pylint-dev/astroid/actions/runs/22041554054