Drop the internal copy of "appdirs" to use the vendored one

[pradyunsg]

See #4945 and #4874 (comment) for context.

pkg_resources uses appdirs. pip has it's own copy of appdirs in utils.appdirs. I'll repeat that I think it's a good idea to patch the vendored appdirs to behave the way we want and switch to using that.

[pradyunsg]

This will mostly involve figuring out which patches have been made to pip's copy of appdirs and merging them upstream.

[pradyunsg]

Okay -- figured those things out. Now, things to do are:

• propose PRs for 4000, 4302 upstream
• Minor typo fixes, if relevant
• figure out how to do 4100 without any upstream changes

---

Notes:

#3275 -- Use os.path.join; instead of os.sep.join([])
#4000 -- Unicode fixes
#4100 -- Added "~/.config/pip/pip.conf" on MacOS
#4302 -- Escape sequences
#3790 -- Typo fixes (merged in ActiveState/appdirs#60)
#3136 -- can revert in pip; all latest minor versions fixed the bug.
#3231 -- typos (appdirs' code unchanged)
#3099 -- no action needed; removed a not-used function
#2467 -- Remove dependency on pywin32 as ctypes should always be available (this has been proposed already; pip currently carries a patch for this)
f63d1e3 -- I don't think any changes to upstream here.

[pradyunsg]

More context: The relevant changes have been proposed upstream and, well, upstream appdirs doesn't seem to have much activity.

@uranusjr has filed #7501, patching our vendored appdirs instead, with our enhancements. IMO that's the better route to go here.

[uranusjr]

It’s a little unfortunate how little interest upstream has to various improvements. I guess the library is probably “finished” from their perspective.

[uranusjr]

So, with #7501 done, should we eliminate pip._internal.utils.appdirs completely? I actually kind of like how the module encapsulates pip’s usage to appdirs in a self-documenting way—no appauthors, different roaming preferences, return multiple site config dirs, etc.

[pradyunsg]

No, I don't think so, for the same reason as you pointed out. Let's close this issue. :)

Thanks a lot for tackling this @uranusjr! ^>^