Issue 5333 - Allow cms to generate relative pages

Issue 5333 - Allow cms to generate relative pages

[wspee, 2017-10-30 16:40:30]

Still working on tests, the tests in master currently fail and this breaks a few
more but I'm still uploading it so we can discuss the general approach.

As discussed this is a hackish solution. Ideally we would generate the links we
want in the first place. But because most asset links are hard-coded this would
mean changing all the links in all of the websites and is a lot more effort.

One reason this is a hack is because because parsing the results again to
replace the links is inefficient and error prone. But this is done already
anyways so at least things don't get worse. In fact because this replaces
regular expressions with a proper html parser this might actually be a nice
short term improvement.

https://i.redd.it/k6sded6b9mkz.png

[Vasily Kuznetsov, 2017-11-14 17:30:28]

Hi Winsley,

This looks like a workable solution, but it would need tests and a bit of
polishing. A couple of questions/suggestions:

- I think it might make more sense to move this code to converters and also make
it work in the dynamic preview. Is there a particular reason you didn't go this
way?

- Have you tested the performance impact of this filter (compared to the rest of
the CMS)? Is it like 3% or more like 30% or 300%?

Cheers,
Vasily

https://codereview.adblockplus.org/29592634/diff/29592635/cms/bin/generate_st...
File cms/bin/generate_static_pages.py (right):

https://codereview.adblockplus.org/29592634/diff/29592635/cms/bin/generate_st...
cms/bin/generate_static_pages.py:95: def rewrite_link(link):
This seems like it would work but the links would look nicer if we compact the
unnecessary pairs of `../some_path`, where the `some_path` part is the same in
the link and in the current page. Perhaps this is not first priority though and
is more of a cosmetic feature.

https://codereview.adblockplus.org/29592634/diff/29592635/cms/bin/generate_st...
cms/bin/generate_static_pages.py:114: for src_zoom in
elem.attrib['srcset'].split(', '):
Is the space after the comma mandatory? Otherwise we can't just split by ', ',
we need to split by comma.