Make #novermin comments work better with other linter comments

[gousaiyang]

Sometimes we need to use multiple linter comments in one line (see also pylint-dev/pylint#2485). Consider the following code snippet, which has three linter comments in one line:

import os

if hasattr(os, 'sched_getaffinity'):
    CPU_CNT = len(os.sched_getaffinity(0))  # type: ignore[attr-defined] # pylint: disable=no-member # novermin
else:
    ...

It would be good if everything works no matter how these comments are ordered, which is the current behavior of Pylint and Flake8. But for vermin, currently it only works (i.e. correctly exclude the line from analysis) when # novermin is the last comment. # novermin # type: ignore[attr-defined] # pylint: disable=no-member and # type: ignore[attr-defined] # novermin # pylint: disable=no-member do not work (for some reason Mypy requires the type comment to be the first comment, but that's not a vermin problem.)

I see a relevant change in the master branch. A caveat is that extracting comments with this regular expression may introduce false positive # novermin comments if # novermin happens to be in a string literal. Maybe using the tokenize module to detect COMMENT tokens is a better way.

Environment

• Vermin version 0.10.3

[netromdk]

Thanks, that is an interesting proposal with tokenize!

I've thought about this before; mainly that it's a mess when using several solutions relying on comments at the same time. I do think it's a solution if the different "contexts" are defined as code ((# comment) ..)

[netromdk]

I got it working. And also double-checked with your input case.