Setting ignore rules through the environment

[acdha]

We have many projects which run Ruff using pre-commit, along with a variety of other tools. There are some Ruff checks which we want to run locally when people are working interactively on the code but we don't necessarily want otherwise passing CI pipelines to fail something like a dependency update for something like the upgrade rules (especially with e.g. #14600).

It would be nice if we could do something like add a RUFF_CONFIG_LINT_IGNORE=UP variable to our group's GitLab CI variables so all pipelines would automatically ignore that set of rules when running in CI but developers working locally would still see those warnings when they have time to deal with them.

One alternative would be if there was an easy way to set the config file path so we could have a CI/CD-oriented one with a few ignore rules. The downside to that is that it's another file to manage across dozens of repositories but it would have a fairly significant upside of making it easy to manage more nuanced configurations.

[MichaReiser]

That's an interesting use case and I currently don't see an easy solution to it.

For my understanding, are users mainly using an editor integration, or how will you avoid flooding users with errors if they run ruff check locally (because the extra rules apply)?

You could pipe some shared CLI arguments from a file (using @argument_file) see argfile) but I suspect this would be very painful to implement because you'd have to change every single CI job to at the argfile and it isn't much different from having a ruff check --ignore $ENV

[acdha]

The VSCode Ruff extension is in use but we also use pre-commit.com's hooks so it'll run ruff check any time they commit code, along with all of the other hooks configured (e.g. JSON formatting, whitespace normalization, running the unit tests, etc.). That works fairly well because when someone is working locally they can spend a little time on maintenance cleanup as part of that task and doing it regularly means there isn't a huge backlog.

The scenario where that doesn't work as well is basically Dependabot-style updates, where it's happening in bulk and outside of a local development context but we'd like to have those not break the build when, say, pre-commit autoupdate means that Ruff's update ruleset expands but do otherwise want to run Ruff & apply any automatic fixes.