FURB148 generates invalid fix when arg is a generator

[Skylion007]

a = (b for b in range(1, 100))
for i, _ in enumerate(a):
    print(i)

gets autofixed to:

a = (b for b in range(1, 100))
for i in range(len(a)):
    print(i)

but a is a generator, and len, surprisingly cannot take a generator (len needs a SizedSequence). Running the code in this autofix will generate the following error:

Traceback (most recent call last):
  File "test_file.py", line 2, in <module>
    for i in range(len(a)):
TypeError: object of type 'generator' has no len()

So while this would work if object was a SizedSequence like a tuple, list, dict, or set, it does not work in this case and will generate invalid.

As an aside, using enumerate to generate an index here when it's a generator is rather nifty, not sure there is an easier way to actually get a range length on generator, using enumerate might actually be most efficient in this case.

The command run is ruff test_file.py --select FURB148 --preview --fix
on ruff 0.0.291

[charliermarsh]

Thanks!

[dhruvmanila]

This will probably require type inference as the variable might be defined outside the enumerate call site as is in this example.

[Skylion007]

@dhruvmanila True, but in the meantime we shouldn't suggest or force the change because if the arg is a generator, enumerate might strictly be better than range (I can't think of a cleaner way to do it).

[tjkuson]

Suppressing the fix would still have the rule trigger, which is less destructive but still frustrating for the user who has to figure out it's a false negative and suppress the rule. Limiting the rule to known safe sequences (for example, using ruff_python_semantic::analyze::typing::is_list) would eliminate this, but result in a whole load of false negatives. We could do that to eventually promote the rule out of the preview category, and then create an issue for generator discrimination when Ruff has better type inference?

[charliermarsh]

Yeah, I think my suggestion would be to only trigger the enumerate case if is_list or is_set or is_dict or is_tuple (i.e., it's a known container) is truthy.

[tjkuson]

Feel free to assign to me, I can work on a merge request tomorrow.