False positive Instance of '_CountingAttr' has no 'append' member with attrs

[peendebak]

Bug description

Pylint is generating warnings on dataclasses defined with attrs.
In particular on lists annotated with field, e.g. values: list[str] = field(factory=list).

With a.py equal to

# Minimal example 

from attrs import define, field

@define
class TestResult:
    """ Test class """

    name: str
    values: list[str] = field(factory=list)
    values2: list[str] = []


    def add_value(self, string : str) -> None:
        """ Ädd value """
        self.values.append(string)

    def hello(self):
        """ Print hello """
        print('hi')
        self.values2.append('hi')

Command used

pylint a.py

Pylint output

************* Module qi_analysis.experiments.x
x.py:16:8: E1101: Instance of '_CountingAttr' has no 'append' member (no-member)

------------------------------------------------------------------
Your code has been rated at 5.00/10 (previous run: 4.00/10, +1.00)

Expected behavior

No pylint errors

Pylint version

pylint 2.15.7
astroid 2.12.13
Python 3.10.8 (tags/v3.10.8:aaaf517, Oct 11 2022, 16:50:30) [MSC v.1933 64 bit (AMD64)]

[alonme]

This seems to be originating from the way astroid detects if a class is decorated by attrs see.

This means that it will not detect attrs when using it as is shown in the official docs .

So - by changing your code to use attrs.define as the decorator the problem is solved

# Minimal example

import attrs
from attrs import define, field

@attrs.define
class TestResult:
    """ Test class """

    name: str
    values: list[str] = field(factory=list)
    values2: list[str] = []


    def add_value(self, string : str) -> None:
        """ Ädd value """
        self.values.append(string)

    def hello(self):
        """ Print hello """
        print('hi')
        self.values2.append('hi')

[alonme]

I opened a PR with a POC to improve the way that astroid detects attrs,
However i think that if we don't get that in, we should add some pointer in the docs / warning to tell users to use @attrs.define to make it work

[alonme]

@Pierre-Sassoulas Which should way should we go here?

1. Add hints for which attrs decorators are "valid" to the docs?
2. My Astroid PR?
3. Add flags to enable more attrs decorators to be treated?

[Pierre-Sassoulas]

I'm not super up to date on this issue, but creating/upgrading an astroid brain is almost always the easy way to fix things, so I would say your MR.