B023: Why flake8 returns error only for a temporal variable for for loop?

[Momoko NAGAI]

flake8 causes error: B023 Function definition does not bind loop variable 'chk_key' for the following code.

        ret_df = pd.DataFrame(index=target_df.index)
        for chk_key in [nalod, nlod]:
            ret_df[chk_key] = (
                False if conf_dict[chk_key] is None
                else info_df.apply(
                        lambda x:
                        float(
                            x['INFO.dict'].get(chk_key, np.nan)
                        ) < conf_dict[chk_key],
                        axis=1
                )
            )

I cannot understand the reason. chk_key is only a temporal variable for for loop. How should I fix it?
I'd like to fix it as soon as possible.

I set a value to chk_key at the top of the function for trial. However, flake8 returns the same message.

[Momoko NAGAI]

I think the above code has the same structure as the following code.
However, flake8 does not return error message for this. What is the difference?

a = 'aaa'
b = 'bbb'

for i in [a, b]:
    print(f'i = {i}')

[Tomas R]

Hi, the difference is that in the first example you reference the loop variable inside a lambda which could in theory lead to issues (not in your case though since it's not called outside the loop).
FWIW, there's already an issue regarding B023 false positives: #380

[Momoko NAGAI]

I really thank you for your quick and proper reply!! > @tomasr8

the loop variable inside a lambda

I understand the situation.

Due to another flake8 error(.../python3.9/site-packages/pep8.py:110: FutureWarning: Possible nested set at position 1 EXTRANEOUS_WHITESPACE_REGEX = re.compile(r'[[({] | []}),;:]')) in my current environment, I will try some countermeasures later.

Thank you.

[Momoko NAGAI]

Is it a correct or ideal output for flake8 developers?
Or is it something like a bug or a problem to be fixed?
(I don't blame developers. Just only a question.)

[Tomas R]

Yes it should ideally not be raised in this case, but distinguishing the false positives from true positives is not that straightforward (see the linked issue for more details)

[Momoko NAGAI]

I really thank you for your proper comment.

distinguishing the false positives from true positives is not that straightforward

I understand it.

With your comment, I can avoid B023 error changing the code to lambda x, y=chk_key: ....
Thank you for your help!

[Cooper Lees]

Would accept making the check handle lambdas better if it's possible.