Suboptimal autofix when combining PLC0415,I001,RUF100 with a long line

[tpgillam]

Summary

Using ruff 0.12.10.

Start with the following code in moo.py:

def moo():
    from very.long.nonsense.indeed.need.line.to.be.full.to.cause.breakage import foo

    return foo()

Running ruff with (at least) these three rules enabled, currently PLC0415 triggers:

$ uvx ruff==0.12.10 check --select PLC0415,I001,RUF100 moo.py
moo.py:2:5: PLC0415 `import` should be at the top-level of a file

Because there are good reasons that the import is local, we disable (in practice, this disable is done in an editor via the ruff language server's suggested code action):

def moo():
    from very.long.nonsense.indeed.need.line.to.be.full.to.cause.breakage import foo  # noqa: PLC0415

    return foo()

but now the line being too long causes I001 to trigger:

$ uvx ruff==0.12.10 check --select PLC0415,I001,RUF100 moo.py
moo.py:2:5: I001 [*] Import block is un-sorted or un-formatted

but if we trigger the autofix with RUF100 enabled, then what the user sees is that the noqa they added disappears, and we get back to:

def moo():
    from very.long.nonsense.indeed.need.line.to.be.full.to.cause.breakage import foo

    return foo()

Fixing without RUF100, we can see that we get this intermediate state:

def moo():
    from very.long.nonsense.indeed.need.line.to.be.full.to.cause.breakage import (
        foo,  # noqa: PLC0415
    )

    return foo()

... but the # noqa: PLC0415 is now the line below where it needs to be, so RUF100's autofix will remove it.

Version

0.12.10

[ntBre]

Thanks for the report and the neat reproduction! I think this might be a duplicate of #12179, where the root I001-RUF100 conflict is the same.