Fix astroid error for custom next method

[clavedeluna]

Type of Changes

	Type
✓	🐛 Bug fix

Description

#7610 provided code that shows an edge case. When someone writes a method called def next():, pylint assumed this func will always have args.

Closes #7610

[coveralls]

Pull Request Test Coverage Report for Build 3377462790

• 3 of 3 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 95.381%

---

Totals
Change from base Build 3367355028:	0.0%
Covered Lines:	17241
Relevant Lines:	18076

---

💛 - Coveralls

[DanielNoord]

Why don't we fix L1142? That if seem to disregard the len(args) == 0 case?

[clavedeluna]

The way the code is written is pretty subtle.. let me reproduce and explain:

            has_sentinel_value = len(node.args) > 1
            if (
                isinstance(frame, nodes.FunctionDef)
                and frame.is_generator()
                and not has_sentinel_value
                and not utils.node_ignores_exception(node, StopIteration)
                and not _looks_like_infinite_iterator(node.args[0])
            ):

has_sentinel_value would be true if there are 2+ args, and False if there are 0 or 1 args. In our specific case of args 0, because this check and not has_sentinel_value is True, then we continue on to check the other two checks of the if statement and hit and not _looks_like_infinite_iterator(node.args[0]), which will fail because we just said args is len 0.

Adding or not node.args: higher up just short circuits everything cleanly.

If you have nicer code that would perform the same thing, do let me know.

[DanielNoord]

Okay, but this changes still doesn't really fix the issue that is described. It just prevents the issue from happening by checking for another heuristic.

Why don't we actually check the qname of inferred to be builtins.next? That seems much more robust?

[clavedeluna]

Because I hadn't thought of it :)

Good point! Seems doable, no?

[DanielNoord]

Yeah I think an isinstance check and then a qname call on that line should be doable!

[Pierre-Sassoulas]

I thought it was impossible to do that because of this: #7622 (comment)

[DanielNoord]

I think we should use an isinstance as well, to make sure the qname method actually exists.

[DanielNoord]

This can be reverted.

[github-actions]

🤖 Effect of this PR on checked open source code: 🤖

Effect on pytest:
The following messages are now emitted:

Details

1. inconsistent-return-statements:
   Either all return statements in a function should return an expression, or none of them should.
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/pytester.py#L301
2. no-name-in-module:
   No name 'NOSE_SUPPORT_METHOD' in module '_pytest.deprecated'
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/python.py#L62
3. no-name-in-module:
   No name 'PytestReturnNotNoneWarning' in module '_pytest.warning_types'
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/python.py#L81
4. inconsistent-return-statements:
   Either all return statements in a function should return an expression, or none of them should.
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/python.py#L1085
5. inconsistent-return-statements:
   Either all return statements in a function should return an expression, or none of them should.
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/fixtures.py#L130
6. inconsistent-return-statements:
   Either all return statements in a function should return an expression, or none of them should.
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/python_api.py#L798
7. no-name-in-module:
   No name 'NOSE_SUPPORT' in module '_pytest.deprecated'
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/nose.py#L5
8. no-name-in-module:
   No name 'warn_explicit_for' in module '_pytest.warning_types'
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/config/__init__.py#L62
9. no-member:
   Module '_pytest.deprecated' has no 'HOOK_LEGACY_MARKING' member
   https://github.com/pytest-dev/pytest/blob/50b232b0cb39490776da2658ae69c669ce263050/src/_pytest/config/__init__.py#L369

This comment was generated for commit abfd9ee

[Pierre-Sassoulas]

LGTM 👍