CLI: Show fixes by default in ruff check

[zanieb]

Requires #7350 for opt-out of display.

ruff check should display fixes that Ruff can apply. Currently this is only possible for the whole file with --diff.

Some fixes may modify content across a file, e.g. when renaming a binding. We should limit the display of fixes to roughly match the range of source context displayed (#7349). If we do not display the entire fix, we should include a notice of the number of additional lines that will be modified if the fix is applied. Naively, we may be able to implement this by only selecting a portion of the fix. In a more comprehensive implementation, we may need to add a new diagnostic range attribute for this display and address it in each problem case.

Clippy has a nice example user experience of showing fixes:

error[E0308]: mismatched types
--> crates/ruff_python_semantic/src/model.rs:224:21
    |
224 |             source: &self.stmt_id,
    |                     ^^^^^^^^^^^^^ expected `Option<NodeId>`, found `&Option<NodeId>`
    |
    = note:   expected enum `std::option::Option<node::NodeId>`
            found reference `&std::option::Option<node::NodeId>`
help: consider removing the borrow
    |
224 -             source: &self.stmt_id,
224 +             source: self.stmt_id,
    |

For more information about this error, try `rustc --explain E0308`.

[pjonsson]

I came here from #8677 because I wanted a non-zero exit code when there are things to be fixed in my source code (even if Ruff can't fix them).

Based on the description of this issue, I don't understand the details but I get the impression it will somehow eliminate the need for --diff, will ruff check <without --diff?> --output-format junit -o ruff-check-report.xml exit with a non-zero exit code and produce junit reports after this is fixed?

[zanieb]

I'm honestly not sure... this is different than --diff in that we just show fixes for the relevant source context. I don't think this would have any affect on the exit code.

[zanieb]

A note regarding the implementation.. we probably need to gate this with preview to match #7349

[pjonsson]

If this has no effect on exit code, I feel that the comments in the tickets concerning exit codes that are linking here (#4093 (comment), #8677 (comment), #8584 (comment)) are confusing. I'm not a native speaker, but to me, all those tickets were filed because the user was expecting a non-zero exit code when Ruff is unhappy. I think part of the reason people are surprised is because it's so easy to get the non-zero exit code with ruff format --diff.

Since this ticket does not help with exit codes, should the non-zero exit code part of this issue be dealt with by #8677?

[zanieb]

Ah I see what you're saying:

1. You want to run Ruff and see all the fixes it has to offer
2. The only way to do this now is with --diff
3. But when you use --diff the exit code is zero

Yes this should address your use-case, in that:

1. We will show fixes by default so --diff is not necessary
2. The exit code semantics should be the same as Ruff without --diff
3. If any violations are remaining, Ruff will exit with a non-zero code