Allow sys.path modifications between imports

[charliermarsh]

Summary

It's common to interleave a sys.path modification between imports at the top of a file. This is a frequent cause of # noqa: E402 false positives, as seen in the ecosystem checks. This PR modifies E402 to omit such modifications when determining the "import boundary".

(We could consider linting against sys.path modifications, but that should be a separate rule.)

Closes: #5557.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+23 -2 violations, +0 -0 fixes in 41 projects)

RasaHQ/rasa (+5 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --preview

+ .github/tests/test_download_pretrained.py:10:29: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ .github/tests/test_mr_generate_summary.py:4:49: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ .github/tests/test_mr_publish_results.py:7:35: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ .github/tests/test_validate_gpus.py:8:22: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ .github/tests/test_validate_gpus.py:9:23: RUF100 [*] Unused `noqa` directive (unused: `E402`)

apache/airflow (+1 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --preview --select ALL

+ docs/conf.py:50:69: RUF100 [*] Unused `noqa` directive (unused: `E402`)

freedomofpress/securedrop (+15 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --preview

+ securedrop/manage.py:22:16: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:23:47: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:29:20: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:30:55: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:31:33: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:32:56: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:33:39: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/manage.py:42:80: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/rqrequeue:11:40: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/rqrequeue:12:46: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/shredder:14:24: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/shredder:15:40: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/shredder:16:28: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/source_deleter:14:24: RUF100 [*] Unused `noqa` directive (unused: `E402`)
+ securedrop/scripts/source_deleter:15:40: RUF100 [*] Unused `noqa` directive (unused: `E402`)

model-bakers/model_bakery (+1 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --preview

+ docs/conf.py:6:37: RUF100 [*] Unused blanket `noqa` directive

pypa/pip (+1 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --preview

+ noxfile.py:16:42: RUF100 [*] Unused blanket `noqa` directive

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

ruff check --no-cache --exit-zero --preview --select ALL

- docs/conf.py:11:1: E402 Module level import not at top of file
- zerver/lib/test_fixtures.py:21:1: E402 Module level import not at top of file

Changes by rule (2 rules affected)

code	total	+ violation	- violation	+ fix	- fix
RUF100	23	23	0	0	0
E402	2	0	2	0	0

[charliermarsh]

I think these are all improvements -- it's nice to see so many # noqa comments being removed.

[charliermarsh]

I gated this to preview for now (ecosystem checks are still updating at time of writing), though honestly I'd be okay with shipping it to stable...

[zanieb]

Nice! I think preview makes sense for this change.

[konstin]

Looking at the ecosystem checks with entrypoint scripts and tests, this makes even more sense

[tooruu]

Why modify sys.path? This sounds like a dirty fix for bad design, not something to encourage.

[zanieb]

Hi @tooruu I'd recommend looking through the ecosystem checks as some of this usage is entirely valid.

We're not encouraging this behavior, we're simply fixing a false positive for a different lint rule. We can introduce a separate rule to suggest not making sys.path modifications to properly discourage this behavior.