member lookup with Generic base class in 2.5.3

[farmio]

Steps to reproduce

from abc import ABC
from typing import Generic, TypeVar

Anything = TypeVar("Anything")
MoreSpecific = TypeVar("MoreSpecific", str, int)


class A(ABC, Generic[Anything]):
    def a_method(self) -> None:
        print("hello")


class B(A[MoreSpecific]):
    pass


class C(B[str]):
    pass


c = C()
c.a_method()

Current behavior

pylint fails with 22:0: E1101: Instance of 'C' has no 'a_method' member (no-member) when using astroid 2.5.3 but doesn't when using astroid 2.5.2

Expected behavior

python -c "from astroid import __pkginfo__; print(__pkginfo__.version)" output

When failing:

> pylint --version
pylint 2.7.4
astroid 2.5.3
Python 3.9.1 (default, Feb 15 2021, 10:36:56)
[Clang 10.0.0 (clang-1000.10.44.4)]

Without failing:

> pylint --version
pylint 2.7.4
astroid 2.5.2
Python 3.9.1 (default, Feb 15 2021, 10:36:56)
[Clang 10.0.0 (clang-1000.10.44.4)]

[sodul]

We are experiencing the same issue with Astroid 2.5.3, rolling back to 2.5.2 addresses it.

We suspect it was introduced by: cd0f896

[cdce8p]

I can confirm the issue. #931 might indeed be related but I don't believe it's the sole reason.

After some investigation, I found that the ancestors of C are calculated incorrectly. Here is a test case for astroid that illustrates the problem:

from astroid import builder

def test_ancestors_with_generics():
    tree = builder.parse(
        """
    from typing import TypeVar, Generic
    T = TypeVar("T")
    class A(Generic[T]):
        def a_method(self):
            print("hello")
    class B(A[T]): pass
    class C(B[str]): pass
    """
    )
    node_b = tree["B"]
    inferred_b = next(node_b.infer())
    ancestors_b = list(inferred_b.ancestors())
    assert [cdef.name for cdef in ancestors_b] == ["A", "Generic", "object"]

    node_c = tree["C"]
    inferred_c = next(node_c.infer())
    ancestors_c = list(inferred_c.ancestors())
    assert [cdef.name for cdef in ancestors_c] == ["B", "A", "Generic", "object"]  # fails!

test_ancestors_with_generics()

At the moment, only B is inferred as ancestors which then results in the no-member error.

@hippo91 Do you have an idea how to fix that? It's probably related to the changes in ClassDef.getitem we did to support __class_getitem__.

[hippo91]

@cdce8p i just had a look to the pb. In fact the ancestors method failed at line

for grandpa in baseobj.ancestors(recurs=True, context=context):

in fact this is the context̀ obj which is the origin of the pb.
An interesting fact is that #927 seems to solve the pb.
However using pylint and astroid with #927 leads to another false positive:

E1136: Value 'A' is unsubscriptable (unsubscriptable-object)

I will try to investigate at bit more ASAP, but a sure thing is that, IMHO, we definitely need #927

[nelfin]

@hippo91 and @cdce8p: both your test and the E1136 pass after merging #927 and #946 (I had a suspicion)

(venv39) (venv39) ~/p/o/a/fixtures (master) $ cat issue942.py
# pylint: disable=missing-docstring,invalid-name,too-few-public-methods,no-self-use
from abc import ABC
from typing import Generic, TypeVar

Anything = TypeVar("Anything")
MoreSpecific = TypeVar("MoreSpecific", str, int)


class A(ABC, Generic[Anything]):
    def a_method(self) -> None:
        print("hello")


class B(A[MoreSpecific]):
    pass


class C(B[str]):
    pass


c = C()
c.a_method()
(venv39) (venv39) ~/p/o/a/fixtures (master) $ python -m pylint -- issue942.py 
************* Module issue942
issue942.py:23:0: E1101: Instance of 'C' has no 'a_method' member (no-member)

------------------------------------------------------------------
Your code has been rated at 6.15/10 (previous run: 6.15/10, +0.00)

(venv39) (venv39) ~/p/o/a/fixtures (master) $ git co fix/926-strong-mutable-ref-in-context-clone 
Switched to branch 'fix/926-strong-mutable-ref-in-context-clone'
(venv39) (venv39) ~/p/o/a/fixtures (fix/926-strong-mutable-ref-in-context-clone) $ 
python -m pylint -- issue942.py 
************* Module issue942
issue942.py:14:8: E1136: Value 'A' is unsubscriptable (unsubscriptable-object)

------------------------------------------------------------------
Your code has been rated at 6.15/10 (previous run: 6.15/10, +0.00)

(venv39) (venv39) ~/p/o/a/fixtures (fix/926-strong-mutable-ref-in-context-clone) $ 
git co fix/942-merge 
Switched to branch 'fix/942-merge'
(venv39) (venv39) ~/p/o/a/fixtures (fix/942-merge) $ python -m pylint -- issue942.py   # <-- here they pass

-------------------------------------------------------------------
Your code has been rated at 10.00/10 (previous run: 6.15/10, +3.85)

I've pushed my this to https://github.com/nelfin/astroid/tree/fix/942-merge on my fork if you'd like to try something else.

[hippo91]

@nelfin nice catch!
Thanks. I hope both PR will be merged soon.

[cdce8p]

This issue is fixed in the current astroid master branch and will be included in the next pylint release.