ignore error on a per line basis

[jayvdb]

ignore error on a per line basis

[sigmavirus24]

So no(qa|pep8) is easily the biggest regret of the flake8/pep8 projects. We refuse to allow it to silence every error and that decision angers users who like to write terrible code. I'm not sure there's a good reason to add this. It might be nice to have, but the doc-string requirements are relaxed enough that I don't see any particular rule that necessitates this. I work on large projects like OpenStack where noqa is particularly abused and it leads to really inconsistent code-bases where tools like flake8 are supposed to help guide the code and its developers towards a relatively consistent practice.

What particular errors are causing you problems @jayvdb ?

[jayvdb]

noqa is a tool; how it is used depends on the developers.

It is handy when migrating code to be compliant, which is something that pywikibot is slowly doing. We have a mandatory tox rule which we are add files to as we complete them, or they near completion and have a # noqa or two to help them across the line, so there cant be any regressions in compliance.

http://git.wikimedia.org/blob/pywikibot%2Fcore.git/b83e2e9e7dbf01800cc4b3bdef7f31d8ee775fdf/tox.ini#L32

We've encountered a few issues with pep257 itself (e.g. issue #78), some which were not even docstring related. We've had to add ugly hacks to our codebase because of pep257 , but we're all pretty keen on cleaning up the codebase, so it is a net positive ;-).

It would be preferable to be able to use # noqa to workaround pep257 problems, when they occur - until they are fixed and released and deployed onto our build machines, if a more elegant solution isnt available.

http://git.wikimedia.org/blob/pywikibot%2Fcore.git/b83e2e9e7dbf01800cc4b3bdef7f31d8ee775fdf/pywikibot%2F__init__.py#L65
http://git.wikimedia.org/blob/pywikibot%2Fcore.git/b83e2e9e7dbf01800cc4b3bdef7f31d8ee775fdf/pywikibot%2F__init__.py#L91

The main outstanding problem in pywikibot is exceptions which use the class docstring as the exception message. Changing that message is an interface change, so adding a '.' at the end of the class docstring should be a major version version, or we need to implement funky method of removing the '.' at runtime when the exception is raised.

http://git.wikimedia.org/blob/pywikibot%2Fcore.git/b83e2e9e7dbf01800cc4b3bdef7f31d8ee775fdf/pywikibot%2Fexceptions.py (search for noqa)

As another example, yesterday I tried to cleanup ez_setup.py, which is copied into many repos (including pywikibot), and met with resistance, which are mostly to do with personal taste wrt the pep257 fixes.

https://bitbucket.org/pypa/setuptools/pull-request/117/pep8-and-pep257-compliance/diff

IMO it is a good thing to push a module as far towards compliance as a maintainer is happy with, and noqa can help reach that objective - the maintainer has at least accepted that this line has problems in a lot of other peoples minds.

[oryband]

👍 on request. any progress on this?

[Nurdok]

@jayvdb - why isn't flake8's noqa directive satisfying your needs?
@oryband - this issue was opened just a couple of days ago and we haven't decided if it will be done yet.

[oryband]

Any progress?

[jayvdb]

@Nurdok this request is about pep257 and not flake8, which we cant run on pull requests from untrusted devs.

[Nurdok]

@jayvdb I suggest you use a tool like TravisCI which can run flake8 on PRs in a safe environment. Other than that, if the purpose of this issue is to allow you to fix the code bit-by-bit, you can also use the error selection to enable only a closed set of checks, and add to it over time.

While I do understand the request for a per-line ignore and while I have not yet decided against it, it will probably not happen any time soon due to more pressing issues. If the errors you get are localized to few files, I would just ignore the entire files, for now.

[hvdklauw]

As someone who works with django I would like my classes to have docstrings, but it seems a bit silly to have it on the nested meta class.

Ideally I would like to see the errors for nested classes split from top level classes, but a noqa option would solve the issue too.

[Nurdok]

@hvdklauw Seems reasonable. Can you open a separate issue for that?

[jayvdb]

I forgot about this issue.
fwiw, my own need for this evaporated after Wikimedia CI allowed executing flake8, and I created https://github.com/jayvdb/flake8-putty so that I could disable specific codes per file & per line.