[anishgirianish]

Summary

Adds unnecessary-regular-expression-compile (RUF078, preview). It flags re.compile(p).match(s) and similar, where the compiled pattern is only ever used right there. In that case re.match(p, s) does the same thing without the extra object. Patterns stored in a variable are left alone, since that usually signals the pattern is meant to be reused.

Closes #14691.

Test Plan

cargo test -p ruff_mdtest --test mdtest -- unnecessary-regular-expression-compile

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+7 -0 violations, +0 -0 fixes in 4 projects; 54 projects unchanged)

PlasmaPy/PlasmaPy (+1 -0 violations, +0 -0 fixes)

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

+ .github/scripts/citation_updater.py:66:29: unnecessary-regular-expression-compile Compiled regular expression is used only once

apache/airflow (+4 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --no-fix --output-format concise --preview --select ALL

+ kubernetes-tests/tests/kubernetes_tests/test_base.py:134:18: unnecessary-regular-expression-compile Compiled regular expression is used only once
+ kubernetes-tests/tests/kubernetes_tests/test_base.py:142:18: unnecessary-regular-expression-compile Compiled regular expression is used only once
+ kubernetes-tests/tests/kubernetes_tests/test_base.py:391:18: unnecessary-regular-expression-compile Compiled regular expression is used only once
+ task-sdk/src/airflow/sdk/definitions/dag.py:1466:25: unnecessary-regular-expression-compile Compiled regular expression is used only once

scikit-build/scikit-build-core (+1 -0 violations, +0 -0 fixes)

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

+ src/scikit_build_core/settings/skbuild_overrides.py:102:17: unnecessary-regular-expression-compile Compiled regular expression is used only once

astropy/astropy (+1 -0 violations, +0 -0 fixes)

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

+ astropy/coordinates/solar_system.py:177:8: unnecessary-regular-expression-compile Compiled regular expression is used only once

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
unnecessary-regular-expression-compile	7	7	0	0	0

[ntBre]

Thanks, this looks good to me overall! As usual, Codex came up with some edge cases that might be worth looking into.

First, I believe the binding.references() count is only counting textual references, so a case like this:

p = re.compile(...)
for x in range(100):
    p.match(...)

only looks like one reference. This occurs in a couple of ecosystem projects:

• https://github.com/pandas-dev/pandas/blob/db9787c75fa4adba75db34f58f05542ec1a640a4/pandas/io/parsers/python_parser.py#L1471-L1480
• https://github.com/zulip/zulip/blob/d101bdb28b66aaa65bf7dfc351d807c948c88672/scripts/lib/check_rabbitmq_queue.py#L147-L159

Somewhat similarly, I don't think the rule is accounting for rebindings currently. So a conditional case like this can be incorrect:

pattern = Matcher()
if condition:
    pattern = re.compile("a")
return pattern.match(value)

This one is the least problematic, or at least seems the most rare, but we might also want to check for side effects in the re.compile arguments because these will only be evaluated once if the rule is followed. Something like this:

re.compile(get_pattern()).match(get_string())

[ntBre]

Thanks, this is looking good. I found a few more edge cases and a possible small simplification.

[anishgirianish]

Thanks, this is looking good. I found a few more edge cases and a possible small simplification.

Hi @ntBre , thank you so much for the review, I have updated pr resolving the feedbacks. Would like to request you for re-review, when ever you get a chance.

Thank you

[ntBre]

Thanks for working on this and dealing with all my comments! I feel bad mentioning this in light of all the work you've put in, but what do you think about restricting the rule to the inline case rather than trying to track assignments and loops and such? This occurred to me today when reviewing again for a few reasons:

• this is the case described in the original issue
• tracking assignments and control flow introduces a lot of complexity into the rule that is hard to get right (Codex found another small bug today)
• Codex pointed out that the re module actually caches compiled regex patterns, even across calls (https://github.com/python/cpython/blob/ebf955df7a89ed0c7968f79faec1de49f61ed7cb/Lib/re/__init__.py#L321-L330). This means the performance justification for the rule is weaker than I thought. It really is more about complexity or style.
• applying the rule to the single-use case seems more opinionated. I think more people would be interested in the inline case, whereas a separate re.compile call feels more intentional

[anishgirianish]

Hi @ntBre, thank you so much for the re-review. I've restricted the rule to the inline
re.compile(...).method(...) form and removed the binding tracking, along with its tests and docs. Whenever you have a moment, would like to request you for your another look

[anishgirianish]

Hi @ntBre, whenever you have a free moment I'd love your thoughts on the latest changes. Thanks so much!

[ntBre]

Thank you! And sorry for the delay. This looks good to me, I just had a couple of small suggestions remaining.

[anishgirianish]

Hi @ntBre, thank you so much for re-review. I have pushed the changes resolving feedback's in the latest commit. Would like to request you for your re-review, whenever you get a chance. Thank you

[ntBre]

Great, thank you!