[ruff] IO operations performed on closed IO objects (RUF050)

[InSyncWithFoo]

Summary

Resolves #10517.

RUF050 checks for references of IO objects whose binding is created by a with statement and reports if those references might trigger ValueError: I/O operation on closed file at runtime.

It recognizes the following patterns:

• Method references: f.read()/_ = f.write
• Contain checks: _ in f
• for loops: for _ in f: ...

Test Plan

cargo nextest run and cargo insta test.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+2 -0 violations, +0 -0 fixes in 1 projects; 54 projects unchanged)

astropy/astropy (+2 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --no-fix --output-format concise --preview

+ astropy/io/ascii/ui.py:1103:9: RUF050 IO operation performed on closed IO object
+ astropy/io/ascii/ui.py:1104:9: RUF050 IO operation performed on closed IO object

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
RUF050	2	2	0	0	0

[InSyncWithFoo]

There are a lot of false positives of this kind:

with open() as f:
	f.read()  # A reference to the next `f`

with open() as f:  # Same name
	...

with open() as f:  # Same name
	f.write()  # A reference to either of the previous `f`s

This seems to be a bug in the semantic model rather than a problem with this rule.

[MichaReiser]

Thanks for working on this.

As mentioned in the original PR, we have to restrict the rule to objects that are files. This should also help to remove the following false positive

https://github.com/apache/airflow/blob/8c9e0b2a8ec0e6f2842883898370741e71c0e802/airflow/models/dagbag.py#L317C12-L317C29

[MichaReiser]

I'm concerned about the examples you mentioned as well as if f gets deleted.

It seems that the semantic model copies over all references when shadowing an existing binding.

ruff/crates/ruff_linter/src/checkers/ast/mod.rs

Lines 2048 to 2062 in fe516e2

	let references = shadowed.references.clone();
	let is_global = shadowed.is_global();
	let is_nonlocal = shadowed.is_nonlocal();
	// If the shadowed binding was global, then this one is too.
	if is_global {
	self.semantic.bindings[binding_id].flags |= BindingFlags::GLOBAL;
	}
	// If the shadowed binding was non-local, then this one is too.
	if is_nonlocal {
	self.semantic.bindings[binding_id].flags |= BindingFlags::NONLOCAL;
	}
	self.semantic.bindings[binding_id].references = references;

I think we have to make the rule shadowed-binding aware and skip over references that come from a shadowed binding.

[InSyncWithFoo]

I think we have to make the rule shadowed-binding aware and skip over references that come from a shadowed binding.

I'm not sure how to do this, especially if all references are simply copied over.

[MichaReiser]

Yeah, me neither. We'd have to look at other rules to see how to handle it. I don't think I've the time right now to do that. Maybe @charliermarsh knows?