Add `--add-ignore` for adding `ruff:ignore` comments by ntBre · Pull Request #26346 · astral-sh/ruff

ntBre · 2026-06-24T23:49:38Z

Summary

This PR adds the new --add-ignore flag. In preview, this adds ruff:ignore comments with human-readable names instead of noqa comments with codes. Without preview enabled, using the flag emits an error. --add-noqa itself is still available in both preview and stable mode.

In cases where an existing noqa or ruff:ignore comment is present, I opted to preserve the
existing comment and to append the new comment. So something like this:

import math  # noqa: RUF100

becomes:

import math  # noqa: RUF100 # ruff:ignore[unused-import]

instead of also converting the existing noqa comment. I think we can defer to one of our planned
lint rules for the conversion (and joining) instead.

As also shown in this example, the flag adds human-readable names rather than rule codes and prefers
the same placement as a noqa comment, at the end of the line. The one exception to this is an
existing ruff:ignore comment on its own line. In that case, we will reuse and extend the existing
comment instead of adding an additional trailing comment, which I think is a pretty nice balance.

Test Plan

Many new tests exercising the various cases Codex and I could come up with. As the server test
shows, this actually applies to the LSP as well, where the Disable for this line action now also
inserts a ruff:ignore comment with a rule name instead of a noqa comment.

Ecosystem check

I also ran some tests against ecosystem projects as described in my comment below. These revealed some additional discrepancies between ruff:ignore and noqa (and some bugs in noqa/--add-noqa itself), but I think it makes sense to leave these as follow-ups. I've already opened (rough) drafts of them in:

We could land the shebang fixes separately, but the other two are stacked on this PR. With all three of these applied, the ecosystem check showed parity with --add-noqa.

This also seems like an interesting check to add to our ecosystem analysis in general.

astral-sh-bot · 2026-06-24T23:55:06Z

`ruff-ecosystem` results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

ntBre · 2026-06-25T16:38:29Z

As part of this PR, I'm checking how well --add-noqa works on our ecosystem projects, first running check --select ALL to collect a baseline, then adding noqa comments, and then checking again. One immediate thing I notice is a huge number of added I001 diagnostics (5278 on all ecosystem projects). These aren't particularly alarming, as they seem to be a result of import lines becoming too long with the new comment.

Then there are 923 new RUF100 diagnostics from a variety of rules that seem not to be suppressible. One example is RUF101: https://play.ruff.rs/45d0e92f-fe87-437e-af87-9a83b099ff5b. Another is PLR2044: https://play.ruff.rs/9bcf9da7-4a50-494b-b6d0-cda53188b876. Another is SIM109: https://play.ruff.rs/922b0cf5-b97e-4c9f-8bdc-e300283f35fe.

We can also introduce syntax errors, mostly via UP031 from what I've seen so far:

❯ cat try.py
print("A long string \  # noqa: UP031
   with a continuation on line %d" % 2)

❯ python try.py
  File "try.py", line 1
    print("A long string \  # noqa: UP031
          ^
SyntaxError: unterminated string literal (detected at line 1)

There are only 10 of these, but we can also disrupt (our detection of?) script metadata and trigger INP001:

# /// script  # noqa: CPY001
# requires-python = ">=3.11"
# dependencies = [
#     "uv==0.11.13",

I'll plan to open a separate issue or issues for these, but for now I'll just try to make sure --add-ignore doesn't make the situation any worse, at least excluding the isort rules, since the longer ruff:ignore[rule-name] will obviously exacerbate that.

Summary -- This PR adds the `--add-ignore` flag as an alias to `--add-noqa`. In preview, this adds `ruff:ignore` comments with human-readable names instead of `noqa` comments with codes. I initially drafted this PR assuming that `--add-ignore` would be a separate flag that only functioned in preview mode, while `--add-noqa` would be available in both preview and stable, but after talking with Micha realized that we probably want to go ahead and push preview users toward `--add-ignore` entirely rather than giving them the choice. I think the one downside of the alias approach is that `--add-noqa` still exists in preview but will still add `ruff:ignore` comments. In cases where an existing `noqa` or `ruff:ignore` comment is present, I opted to preserve the existing comment and to append the new comment. So something like this: ```py import math # noqa: RUF100 ``` becomes: ```py import math # noqa: RUF100 # ruff:ignore[unused-import] ``` instead of also converting the existing `noqa` comment. I think we can defer to one of our planned lint rules for the conversion (and joining) instead. As also shown in this example, the flag adds human-readable names rather than rule codes and prefers the same placement as a `noqa` comment, at the end of the line. The one exception to this is an existing `ruff:ignore` comment on its own line. In that case, we will reuse and extend the existing comment instead of adding an additional trailing comment, which I think is a pretty nice balance. Test Plan -- Many new CLI tests exercising the various cases Codex and I could come up with. As the server test shows, this actually applies to the LSP as well, where the `Disable for this line` action now also inserts a `ruff:ignore` comment with a rule name instead of a `noqa` comment.

MichaReiser · 2026-06-26T18:38:13Z

but after talking with Micha realized that we probably want to go ahead and push preview users toward --add-ignore

Hmmm, I don't remember this conversation. Given recent feedback. I'm leaning towards preserving --add-noqa to add noqa comments and only have --add-ignore add ruff ignore comments. We could show a warning if a user has the "dont use noqa" rule enabled and hint them towards --add-ignore, but I don't see a reason to take away --add-noqa (it's still useful to everyone who doesn't want to use ruff:ignore and the worst that happens is that you can run --fix, which will fix all noqa comments to ruff:ignore)

ntBre · 2026-06-26T18:40:02Z

Maybe I misremembered, but it was a couple of weeks ago. Anyway, happy to preserve both flags.

MichaReiser · 2026-06-26T18:43:25Z

Sorry, to be clear. I'm not saying you made this up. I just don't remember it, which doesn't mean that I didn't say something in that direction

MichaReiser

This looks good to me as a first version. We can iterate on it based on user feedback. I do suggest not making --add-noqa an alias of --add-ignore. Instead, --add-noqa adds noqa comments, and --add-ignore uses ruff:ignore. Making it an alias is a problem for users who want to continue using noqa codes because they now loose this functionality.

I do suggest replacing the noqa naming in add_noqa with suppression and rename the corresponding function and module.

Unfortunately, my codex interrogation on noqa and ruff:ignore semantics uncovered a few more bugs in ruff:ignore. I don't think they need to block this PR, but we should look into them.

if True:
   first = 1  # ruff:ignore[indentation-with-invalid-multiple]
   second = 2  # ruff:ignore[indentation-with-invalid-multiple]

This does not work, but the corresponding noqa: E111 works

(I must admit, these specific ignore comments look rather verbose)

The same for

import logging

logger = logging.getLogger(__name__)
name = "world"
logger.error(
    f"""Hello {
        name
    }""",  # noqa: G004
)

logger.error(
    f"""Hello {
        name
    }""",  # ruff:ignore[logging-f-string]
)

and

#!/usr/bin/env python3
from __future__ import annotations  # ruff:ignore[undocumented-public-module]

This also does not work with ruff:ignore, but works with noqa

value = 1  # noqa: F401  # ruff:ignore[unused-noqa]

MichaReiser · 2026-07-06T12:54:41Z

+        suppression_kind,
    );
-    build_noqa_edits_by_diagnostic(comments, locator, line_ending, None)
+    build_noqa_edits_by_diagnostic(comments, locator, line_ending, None, suppression_kind)


should this be renamed to say build_suppression_edits_by_diagnostic. We may want to do so in the entire module, including the module itself.

Good idea. I'm renaming the relevant noqa references to suppression, including the module.

Hmm, the ruff:ignore comments are actually already in a module called suppression. As a quick fix, I was going to rename that module to suppression_comments, but I'm wondering if we might want a different structure. Something like a top-level suppression with nested noqa and ignore modules could work, if we don't mind lumping enable/disable under the ignore name. A lot of the code in the noqa module is still specific to noqa comments, but these shared functions could live in the parent suppression module.

I think I'll stick with the function renames here and then follow up on the modules if you think that sounds like a good idea.

MichaReiser · 2026-07-06T13:05:43Z

-    match directive {
-        None => {
+    match (directive, suppression_kind) {
+        (Some(ExistingDirective::Noqa(codes)), SuppressionKind::Noqa) => {


It may be worth adding some inline comments here (what are these different arms)

Add code to an existing noqa

Add code to an existing ruff:ignore

No existing comment or "wrong" comment

MichaReiser · 2026-07-06T13:06:38Z

+) -> (TextRange, bool) {
+    let prefix = locator.slice(TextRange::new(line_range.start(), directive_start));
+    if prefix.trim_whitespace().is_empty() {
+        (TextRange::new(directive_start, line_range.end()), true)


Can we document what this method returns. What's the meaning of the second bool

MichaReiser · 2026-07-06T13:08:22Z

+        // Prefer extending an existing `ruff:ignore` over an existing `noqa` directive.
+        let directive = suppressions
+            .find_applicable_ignore(message)
+            .map(ExistingDirective::Ignore)
+            .or(existing_noqa);


Reading the code below. It seems to be that we never reuse an existing suppression comment if it doesn't match the requested suppression style. This comment sort of suggests the opposite to me. Or is it, if we have two suppression comments on the same line, find the first that matches the expected suppression style?

Ah yeah, I'll rephrase this. I was trying to say something more like "Prioritize an existing ruff:ignore over an existing noqa directive on the same line." For example, we want to extend ruff:ignore in this case instead of appending another ruff:ignore just because we found noqa first:

import math # noqa: F401 # ruff:ignore[unused-import]

Actually, I think this logic will have to change when the flag is no longer an alias, though. In the current state of the PR, it's assumed that if a ruff:ignore comment exists and is active, then preview is obviously enabled and --add-noqa/--add-ignore is thus adding an ignore comment. As you said, what really needs to be prioritized is the requested suppression style.

ntBre · 2026-07-06T13:42:59Z

Unfortunately, my codex interrogation on noqa and ruff:ignore semantics uncovered a few more bugs in ruff:ignore. I don't think they need to block this PR, but we should look into them.

Thanks! Yeah, I had done a similar investigation and spun those bug fixes off into separate (draft) PRs to avoid blocking this one. It looks like we mostly found the same things.

if True:
   first = 1  # ruff:ignore[indentation-with-invalid-multiple]
   second = 2  # ruff:ignore[indentation-with-invalid-multiple]
This does not work, but the corresponding noqa: E111 works

(I must admit, these specific ignore comments look rather verbose)

This should be covered by #26412.

The same for

import logging

logger = logging.getLogger(__name__)
name = "world"
logger.error(
    f"""Hello {
        name
    }""",  # noqa: G004
)

logger.error(
    f"""Hello {
        name
    }""",  # ruff:ignore[logging-f-string]
)

Codex says this one is "partially" fixed by #26412 by adjusting the comment placement for ruff:ignore instead of making the comment work in that placement. I'll make a note to double check this case.

#!/usr/bin/env python3
from __future__ import annotations  # ruff:ignore[undocumented-public-module]

This case should be addressed by #26373.

This also does not work with ruff:ignore, but works with noqa
value = 1  # noqa: F401  # ruff:ignore[unused-noqa]

This should also be covered by #26412.

ntBre · 2026-07-07T00:05:14Z

I moved the alias to a separate flag that conflicts with --add-noqa and causes an error if it's used when preview is disabled. This also pushed the suppression_kind check a little earlier, which meant I could restore the more specific messages like Added 1 noqa directive instead of the generic Added 1 suppression comment.

I also moved most of the tests to the noqa module with a ty-style add_suppressions_in helper, but I probably want to take one more look at that since Codex included some weird closures. They otherwise look nice, though.

ntBre added cli Related to the command-line interface preview Related to preview mode features labels Jun 24, 2026

ntBre force-pushed the brent/add-ignore branch from 31c5406 to c355300 Compare June 24, 2026 23:55

ntBre mentioned this pull request Jun 25, 2026

Handle additional ruff:ignore edge cases after a shebang #26373

Draft

ntBre force-pushed the brent/add-ignore branch 2 times, most recently from 3b9f7c7 to ac10c98 Compare June 26, 2026 13:39

This was referenced Jun 26, 2026

Preserve trailing content in --add-noqa and --add-ignore #26409

Draft

Several ruff:ignore fixes #26412

Draft

ntBre force-pushed the brent/add-ignore branch from ac10c98 to b141602 Compare June 26, 2026 17:49

ntBre marked this pull request as ready for review June 26, 2026 17:58

MichaReiser self-requested a review June 26, 2026 18:10

ntBre mentioned this pull request Jun 29, 2026

quick fix for ignoring an error should use the new human readable # ruff: ignore syntax instead #26447

Closed

ntBre linked an issue Jun 29, 2026 that may be closed by this pull request

quick fix for ignoring an error should use the new human readable # ruff: ignore syntax instead #26447

Closed

MichaReiser approved these changes Jul 6, 2026

View reviewed changes

ntBre added 5 commits July 6, 2026 12:21

Merge branch 'main' into brent/add-ignore

c9f942b

noqa -> suppression in many function and variable names

49dfb94

add comments for generate_suppression_edit branches

cdcab1f

document suppression_edit_range

267ece9

separate flag instead of alias

13e2df2

ntBre force-pushed the brent/add-ignore branch from 37ee648 to 13e2df2 Compare July 6, 2026 22:32

ntBre added 2 commits July 7, 2026 09:53

move to add_suppressions_in tests

77b001d

Merge branch 'main' into brent/add-ignore

054b174

ntBre added 2 commits July 7, 2026 10:16

avoid tempdir in tests

5d91d57

minor simplification

a1bffc0

ntBre force-pushed the brent/add-ignore branch from 3f8b39d to a1bffc0 Compare July 7, 2026 14:38

ntBre added 2 commits July 7, 2026 11:18

migrate a few additional tests and restore a preexisting one

506bb35

restore --preview flags now that --add-noqa still works in preview

c0c76d5

ntBre merged commit 63b82b0 into main Jul 7, 2026
47 checks passed

ntBre deleted the brent/add-ignore branch July 7, 2026 16:26

BrewTestBot mentioned this pull request Jul 9, 2026

ruff 0.15.21 Homebrew/homebrew-core#292272

Merged

Uh oh!

Conversation

ntBre commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test Plan

Ecosystem check

Uh oh!

astral-sh-bot Bot commented Jun 24, 2026

ruff-ecosystem results

Linter (stable)

Linter (preview)

Uh oh!

ntBre commented Jun 25, 2026

Uh oh!

MichaReiser commented Jun 26, 2026

Uh oh!

ntBre commented Jun 26, 2026

Uh oh!

MichaReiser commented Jun 26, 2026

Uh oh!

MichaReiser left a comment

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

ntBre Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

ntBre Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

MichaReiser Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

ntBre Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

ntBre commented Jul 6, 2026

Uh oh!

ntBre commented Jul 7, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ntBre commented Jun 24, 2026 •

edited

Loading

`ruff-ecosystem` results