Change imports to a one-per-line style

[xavfernandez]

This would prevent useless merge conflicts.

I think it would only need to add force_single_line = true to our isort configuration.

Originally posted by @pradyunsg in #6729 (comment)

And for the record, I'm in favor :)

[xavfernandez]

cc @chrahunt @cjerdonek @dstufft @pfmoore for your opinion :)

[pradyunsg]

OT: that list is almost alphabetical and I don't enjoy that. 🙂

edit: I actually edited the list in the comment above to be alphabetical. Judge me internet. :P

[pradyunsg]

FWIW, we'd probably use a style like:

from pip._internal.utils.misc import (
    _make_build_dir,
    ask_path_exists,
    backup_dir,
    call_subprocess,
    display_path,
    dist_in_site_packages,
    dist_in_usersite,
    ensure_dir,
    get_installed_version,
    redact_password_from_url,
    rmtree,
)

I'm almost certain that isort supports this as a mode in it -- 3 or 5.

(I typed this on Monday morning, on my mobile, while walking to class. I'm weird.)

I see there was some discussion at #6729 so I figured it'd help to clarify.

/cc @pfmoore

[pradyunsg]

Filed #6764 for this.

[pfmoore]

I still don't like it, sorry.

[pradyunsg]

Do you dislike it enough, to not be willing to work with it?

As such, this would prevent merge conflicts due on the import statements, which AFAICT is the most common kind of merge conflict we see in pip nowadays -- keyring PR had conflicts a lot of times and it was almost always only the imports.

[pfmoore]

I'm not looking to wield a one-man veto, no.

But my personal workflow means that I can't routinely run a tool like isort on changes I make, so something like this just adds an extra "manual edit, submit PR, get lint failure, manually fix it, rinse and repeat" cycle to my coding. So I want to register my preference, that's all.

[cjerdonek]

I've never noticed / given any thought to the merge conflicts this is meant to address, so I don't personally see a need. But I can adapt to work with anything.

[pradyunsg]

Thanks for clarifying @pfmoore! I wasn't intending to imply that, but then again, an explicit confirmation isn't a bad thing. :)

@cjerdonek, good to know! :)

---

I'm inclined to say let's do this. It's a less painful for external contributors, since their PRs are more likely to stay open for longer, which increases the likelihood of a conflict like this (eg. see the keyring PR or the 3.10 PR, which are really the main motivators for me supporting this change).

To that end, to keep us from stalling here, I'll wait for another couple of days and unless someone raises a "don't do this; it's bad" concern, I'll merge #6764.

[pfmoore]

Good point re reducing the pain for external contributors. I'll learn to live with the change, might even persuade me to improve my workflow, who knows? :-)