Resolve relative URLs in e-* HTML content

[snarfed]

Inspired by snarfed/bridgy-fed#390 (comment), open parsing issue microformats/microformats2-parsing#38, related implementation details in #138.

[angelogladding]

From the HTML attributes table referenced in @Zegnat's proposal I see the following URL attributes:

• cite, data, href, src, poster
• ping
• action, formaction
• itemid, itemprop, itemtype

Do we need to support ping, forms or microdata?

Can this wait until after the transition to py3?

[snarfed]

Thanks! I expect it's fine to ship an initial implementation of this with just the first line: href, src, cite, data, poster, probably in that priority order. The rest seem very low priority and/or unwanted entirely.

(Also mf2py migrated to python 3 ages ago, that's not a blocker here.)

[angelogladding]

Oops, I meant to say "transition away from py2" or "transition to py3 only".

With py3 wheels we can install lxml without having to compile and thus it's possible to add it as a core dependency.

If that were to be the case we could use their machinery: https://lxml.de/lxmlhtml.html#working-with-links

Here's their code: https://github.com/lxml/lxml/blob/11b33a83ad689bd16bd0a98c14cda51a90572b73/src/lxml/html/__init__.py#L435

It's a bit of a stretch but also a nod toward feature freeze in anticipation of deprecation of py2.

[angelogladding]

lxml does not make relative srcset items absolute and I see now that other properties are already being made absolute.

I'll work on support for cite, data, href, src, poster. I imagine srcset on an img found inside the e-* should be made absolute as well.

[JKingweb]

I went through this exercise a few weeks ago, and compiled a list of URL-holding attributes from the HTML, MathML 3, and SVG Tiny specifications:

• action on form
• altimg on math:math
• cdgroup on math:math
• cite on blockquote, del, ins, q
• data on object
• definitionURL on math:*
• formaction on button, input
• handler on svg:listener
• href on a, area, base, link, math:*
• itemid on html:*
• itemprop on html:*
• itemtype on html:*
• ping on a, area
• src on audio, embed, iframe, img, input, script, source, track, video, math:mglyph, math:share, math:annotation, math:annotation-xml
• xlink:href on svg:*, math:*
• xlink:role on svg:*, math:*
• xlink:arcrole on svg:*, math:*
• xml:base on svg:*, math:*

Then you also have image candidate strings (comma-separated):

• imagesrcset on link
• srcset on img, source

And finally, URL token-lists (whitespace-separated):

• macros on math:math
• requiredExtensions on svg:a, svg:animate, svg:animateColor, svg:animateMotion, svg:animateTransform, svg:animation, svg:audio, svg:circle, svg:desc, svg:discard, svg:ellipse, svg:foreignObject, svg:g, svg:image, svg:line, svg:metadata, svg:path, svg:polygon, svg:polyline, svg:rect, svg:set, svg:switch, svg:tbreak, svg:text, svg:textArea, svg:title, svg:tspan, svg:use, svg:video
• requiredFeatures on svg:a, svg:animate, svg:animateColor, svg:animateMotion, svg:animateTransform, svg:animation, svg:audio, svg:circle, svg:desc, svg:discard, svg:ellipse, svg:foreignObject, svg:g, svg:image, svg:line, svg:metadata, svg:path, svg:polygon, svg:polyline, svg:rect, svg:set, svg:switch, svg:tbreak, svg:text, svg:textArea, svg:title, svg:tspan, svg:use, svg:video

Of all these href on base is a special case (its own value is not considered for resolution), as is xml:base. Attributes on foreign elements should probably also perform xml:base resolution in addition to HTML base resolution.

[angelogladding]

@JKingweb what do you think about supporting the minimal case for now? https://github.com/microformats/mf2py/pull/201/files#diff-746fcc4c80a0a67fedb71d5bed24ae4a0e4925083edd63a8cef66eadc69a4cb3R100