Fix crash when using enumerate with start and a class attribute

[clavedeluna]

Type of Changes

	Type
✓	🐛 Bug fix

Description

fixes

    start_val = node.value
AttributeError: 'Attribute' object has no attribute 'value'

Closes #7821

[clavedeluna]

in this entire func I'm negotiating between knowing what type of node will be and crashing... would you prefer I make the else the "catch all" instead and ensure we never crash regardless of not knowing what the node could be?

[DanielNoord]

I don't really understand your point and I'm not sure I missed this in an earlier review but the broad else doesn't make a lot of sense.
mypy only allows it because astroid isn't typed and therefore every astroid object is Any as far as mypy can see. Doing node.value is not type safe sadly so we should use instance checks.

[Pierre-Sassoulas]

Oftentimes if a node is not something we expect we return early. I.e. we use the pattern we talked about in #7787 (comment). It's a way to make sure we have false negatives instead of false positive or crashes. To be clearer listing the node we expect to treat is safer than thinking we know what we'll have to treat. It's really hard to think of all possible python construct around a particular situation. Might be a principle that should be documented / checked in review btw.

[clavedeluna]

Alright so... I think this would meet both of your comments?

    def _get_start_value(self, node: nodes.NodeNG) -> tuple[int | None, Confidence]:
        if isinstance(node, (nodes.Name, nodes.Call, nodes.Attribute)):
            inferred = utils.safe_infer(node)
            start_val = inferred.value if inferred else None
            return start_val, INFERENCE
        if isinstance(node, nodes.UnaryOp):
            return node.operand.value, HIGH
        if isinstance(node, nodes.Const):
            return node.value, HIGH

        return None, HIGH

?

[coveralls]

Pull Request Test Coverage Report for Build 3544004706

• 7 of 7 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.005%) to 95.428%

---

Totals
Change from base Build 3543573374:	0.005%
Covered Lines:	17552
Relevant Lines:	18393

---

💛 - Coveralls

[clavedeluna]

Primer stdlib error appears to be unrelated and probably hitting other PRs, too

[Pierre-Sassoulas]

Blocked by #7826

[Pierre-Sassoulas]

Could you rebase/merge main now that #7826 is merged please ? :)

[clavedeluna]

I'm going to add the test case provided in #7821 (comment) and rebase and hopefully we should be good

[Pierre-Sassoulas]

Good job on adding the second test case.

[github-actions]

🤖 According to the primer, this change has no effect on the checked open source code. 🤖🎉

This comment was generated for commit f3e5eab