Fix crash on inference of __len__

[DanielNoord]

Steps

• For new features or bug fixes, add a ChangeLog entry describing what your PR does.
• Write a good description on what the PR does.

Description

This is a weird one.

I found this while working on the primer tests for pylint where running on the codebase of numpy created a crash. I issued pylint-dev/pylint#5244 for this. Initially I thought it had something to do with the brain for numpy as the crash only occurred when the directory name starts with numpy.
However, removing this single line seems to fix the issue. Furthermore, (locally) all tests for astroid and pylint pass with this change. I have looked at the commit that added this line (25384d4) but did not immediately understand why this was added. In any case, no tests seems to have been added or this line is redundant and the removing it is the way forward.

Type of Changes

	Type
✓	🐛 Bug fix

Related Issue

Would close pylint-dev/pylint#5244

[DanielNoord]

I'm marking this as a draft because I'm not 100% sure this is the right way forward, but I would like some input for any of the experienced astroid maintainers. I can't really continue this on my own.

I also have not thought of a good way to test this in astroid. The test in pylint work, but for astroid I do not know what will work best.

[Pierre-Sassoulas]

@brycepg @hippo91 do you have an opinion on this change ?

[hippo91]

@DanielNoord @Pierre-Sassoulas i had a look to this PR. Honestly i do not understand why this line was added.
Maybe @brycepg will be more helpful than me.
IMHO if this line is not tested and all astroid and pylint tests are ok we can remove it.
Maybe we can ask @cdce8p to test it again home-assistant?

[Pierre-Sassoulas]

Maybe we can ask @cdce8p to test it again home-assistant?

We're also putting primer tests in place in pylint-dev/pylint#5173 so maybe we'll get to be more confident in small changes like this in the near future :)

[cdce8p]

I had some time to review this PR. A few notes.

The _proxied comparison is necessary. Otherwise this would also match.

class A:
    def __len__(self) -> int:
        return 42

class Crash:
    def __len__(self) -> int:
        a = A()
        return len(a)

Furthermore, I was able to reproduce the issue without numpy.

class MyClass:
    def some_func(self):
        return lambda: 42

    def __len__(self):
        return len(self.some_func())

Those two tests could probably be added to unittest_brain.py, similar to the existing one.
https://github.com/PyCQA/astroid/blob/b62f243b16b7f435c8be869577959e95a7927a91/tests/unittest_brain.py#L3112-L3126

--
As for numpy, the whole issue was only triggered because self.array was inferred as numpy FunctionDef. Not sure this is correct in this instance or if the numpy brain needs to be updated. Maybe @hippo91 know more about that? Anyway, if it needs to be updated, this should happen in another PR.
https://github.com/PyCQA/astroid/blob/b62f243b16b7f435c8be869577959e95a7927a91/astroid/brain/brain_numpy_core_multiarray.py#L41-L42
https://github.com/PyCQA/astroid/blob/b62f243b16b7f435c8be869577959e95a7927a91/astroid/brain/brain_numpy_utils.py#L90

[DanielNoord]

Fixed the code with @cdce8p's suggestions and added test cases. The last test will fail without the changes in this PR, the first test is to make sure the inference on these kind of functions actually works.

As for the numpy error, I have not looked into it but we might want to open a separate issue to channel discussion. @cdce8p did you happen to find any hints as to why this inference was (probably) wrong? And if so, would you mind opening an issue?

[cdce8p]

Just some last comments. I know that the test names are already descriptive, but adding one more line in the docstring might be useful.

[DanielNoord]

Added docstrings, also for the existing test

[cdce8p]

As for the numpy issue, I don't think it's worth opening an issue for it as long as we don't encounter it in practice.
Maybe @hippo91 knows the answer to it, but not too important if not.

--
File placed in a folder which starts with numpy.

class MyClass:
    def __init__(self, array):
        self.array = array
        self.array  #@

self.array will (wrongly ?) be inferred as FunctionDef.
It seems to be related to this https://github.com/PyCQA/astroid/blob/b62f243b16b7f435c8be869577959e95a7927a91/astroid/brain/brain_numpy_core_multiarray.py#L41-L42
https://github.com/PyCQA/astroid/blob/e1ecb6d3d5a6555a807ab3db0f2f938d8d73e95c/astroid/brain/brain_numpy_core_multiarray.py#L90-L95