Allow `# noqa: SIM103` on last return

[JKE-be]

Description

searched: SIM103 Ignore

snippet:

def func(a, b):
    if a == 0:
        return False

    a = a + b / a
    if a != 1:
        return False

    a = a * b
    if a > 1:
        return False

    return True  # noqa SIM103

Command: ruff check /tmp/ruff.py

Output:

/tmp/ruff.py:10:5: SIM103 Return the negated condition directly
   |
 9 |       a = a * b
10 | /     if a > 1:
11 | |         return False
12 | |
13 | |     return True  # noqa SIM103
   | |_______________^ SIM103
   |
   = help: Inline condition

version: ruff 0.9.6

config:

fix = false
show-fixes = true
line-length = 120
output-format = "full"
target-version = "py38"

[lint]
ignore = [
    "E501",
    "E731",
    "RUF012",   # mutable-class-default
]
select = [
    "B",   # flake8-bugbear
    "E",   # pycodestyle
    "F",   # Pyflakes
    "G",   # flake8-logging-format
    "ISC", # flake8-implicit-str-concat
    "PERF",# perflint
    "RUF", # ruff specific rules
    "SIM", # flake8-simplify
    "W",   # pycodestyle
]

Expected (Nice to have feature)

Allow ignoring SIM103 without requiring the # noqa comment on the last if statement itself. The goal is to enable other developers to add another if after the last one without modifying the previous condition.

This is why I don’t want to refactor the last condition as suggested by Ruff. It’s a special case, but I don’t want to ignore SIM103 for the entire file or all files, as I still find it useful in 99% of cases.

Currently, I need to add # noqa on the last ìf to get All checks passed!

def func(a, b):
    if a == 0:
        return False

    a = a + b / a
    if a != 1:
        return False

    a = a * b
-   if a > 1:  # noqa SIM103 
+   if a > 1:
        return False

-   return True
+   return True  # noqa SIM103 

[InSyncWithFoo]

Irrelevant, but the comment should have a colon: # noqa: SIM103. See also PGH004.

[dylwil3]

This is an interesting case, thank you!

As far as I can tell this would require some fairly substantial changes to how error suppression currently works for multi-line diagnostic ranges. Note that, in your example, that final return statement is part of the diagnostic range, it's just that in general the in-line suppression comment is supposed to be on the first line of a diagnostic.

So I think supporting this may require some design discussion to understand whether it should be done and what the ramifications might be.

[MichaReiser]

Ha, @dylwil3 I was just about to write the same.

Red Knot implements your desired behavior (see #15046 (comment)) but I'd prefer to keep the behavior of noqa consistent with flake8. I'd have to check what behavior flake8 uses. Either way, this is a rather significant change.

There are some trade offs involved with matching both the start and end of the range (see red knot PR) and the main mitigation is to strictly use error codes.

[brandonchinn178]

Not sure how easy it would be to detect, but I would actually argue that this should not be a violation of SIM103. The spirit of SIM103 is to replace simple if checks with a boolean expression, but fall through boolean checks is a very common idiom that should be allowed

# GOOD - easy to read and extend
def check(s: str) -> bool:
    if s.startswith("/"):
        return True

    if s.endswith("/"):
        return False

    # TODO: more guards will be added in the future

    return True

# BAD - difficult to read or extend
def check(s: str) -> bool:
    return s.startswith("/") or not s.endswith("/")

The original ask would be a decent mitigation for this, but IMO a better resolution would be to exclude these situations, if possible.