Generate `Fix` for suppressing violations

[MichaReiser]

This is not fully-fledged, it's more a description of an idea that I would like to explore.

Currently, Ruff serializes the noqa_row as part of the JSON messages and the VS code plugin uses that row number to insert the noqa comment at the right place. The issue that I have with the noqa_row is that it leaks internal details and complicates the LSP because it has to know how to correctly add a noqa comment.

What if ruff would instead generate a second Fix to add the noqa suppression comment?

[bluetech]

I had the same thought: https://github.com/charliermarsh/ruff-lsp/pull/76#issuecomment-1454257946

It does feel much cleaner, however these are not regular fixes (+ it makes every lint "fixable"...) so should be distinguished in some way - how to do it?

[MichaReiser]

I had the same thought: https://github.com/charliermarsh/ruff-lsp/pull/76#issuecomment-1454257946

Oh nice, I didn't know about this. Thanks for cross-referencing!

It does feel much cleaner, however these are not regular fixes (+ it makes every lint "fixable"...) so should be distinguished in some way - how to do it?

Hmm, that's a good point. I guess we could either:

• Use the newly introduced Fix::applicability and only show violations as fixable if they have at least one non-manual fix.
• This would be harder to do in Ruff today because it doesn't support this capability, but we could provide this as a code action if the cursor is positioned in a range with a violation.

[bluetech]

I was thinking, since this is not useful for regular cli usage, but only for LSP-like scenarios, that the LSP should just tell Ruff "please generate these", in one way or another.

[T-256]

Since it'd be used mostly by LSP, I'd recommend hidden CLI option --suggest-noqa.

[dhruvmanila]

\cc @snowsignal You might be interested in this as it would easier to implement in the native LSP implementation.

[MichaReiser]

@snowsignal do you think this issue is still relevant after your new LSP implementation?

[snowsignal]

@MichaReiser I think this is still relevant because we still need to support noqa suppression comments in the new extension, which calls into ruff_linter directly. We'll still need to figure out how to generate fixes for noqa comments, though it would probably be a separate method called after the linter is run.

Re: https://github.com/astral-sh/ruff/issues/4307#issuecomment-1540490537

This would be harder to do in Ruff today because it doesn't support this capability, but we could provide this as a code action if the cursor is positioned in a range with a violation.

With the new Ruff LSP, we should be able to support this, though it will require the implementation of diagnostic caching first.