notifications: Switch to use make_links_absolute() from lxml library.

[HarshitOnGitHub]

Instead of using custom regexes for converting relative URLs to
absolute URLs switch to using make_links_absolute() function
from lxml library.

[rht]

@HarshitOnGitHub is there any performance gain observed? Regardless, I prefer simpler codebase :> $$\therefore$$ LGTM.

[timabbott]

I think we should just do import lxml.html or import lxml; no reason to use an abbreviation here.

[timabbott]

Posted one comment, otherwise this looks good, since we have unit tests for that function so we can pretty confident there isn't something minor broken here (right?).

[HarshitOnGitHub]

@timabbott Agree. Updated! Yeah we have a good test suite there so it is very unlikely that this may have broken something minor.
@rht I didn't checked for performance gains but I think this new approach should be faster than previous one. Since this is not a performance critical path I would have replaced the older approach event if this approach was a little bit slower than the older one since this is cleaner and more accurate. :)

[timabbott]

Yeah we should definitely not waste time trying to see performance impact. It probably is a bit slower in a completely irrelevant way.

[timabbott]

Merged, thanks @HarshitOnGitHub!

[timabbott]

Oh, one note; do you think there's a way to handle the file uploads part with lxml as well? Might be a good idea to convert that as well.

[HarshitOnGitHub]

@timabbott I think we want abbreviation there because of this: https://stackoverflow.com/a/13093714/7516461 otherwise mypy will complain.
Which file uploads part are you referring to?

[timabbott]

I meant the second half of relative_to_full_url.

This works fine:

In [1]: import lxml
In [2]: lxml.html.parse
Out[2]: <function lxml.html.parse>

but you're right that mypy's annotations for lxml are broken; I posted on python/typeshed#525 about the issue (and added some type: ignores to our codebase).