feat: Preserve HTML data attributes (from spans to anchors)

[pawamoy]

Closes #41

(will squash to add you as co-author @oprypin)

[pawamoy]

Ah, adding class to spans will cause issues. That's probably what you meant in your issue comment about still using the attrs HTML parser?

[oprypin]

Could you explain to me, I actually don't know what kind of issues this would be. No, I didn't predict this. I only meant if you need to parse the attributes but apparently that's not even needed 🙂

[pawamoy]

Well, autorefs sets class=... on anchors, so if someone sets class=... on an autorefs span, anchors will end up with two class attributes. Maybe browsers don't mind though? My initial guess is that it's incorrect and only the last class attribute will be kept, therefore cancelling the autorefs-internal/external classes.

[oprypin]

Ah yes for sure multiple class would be super bad

[oprypin]

My first instinct is to still double down on regexes and require the class attribute to come first 🤪 (well, second)
But yes the other option is to parse the attributes

[oprypin]

The reason that is it's still ok to do this is because all usages should be programmatic and we can still have strict requirements on them

[oprypin]

Commit oprypin@10d76d3 works accordingly

[pawamoy]

The reason that is it's still ok to do this is because all usages should be programmatic and we can still have strict requirements on them

I've personally written autoref spans manually on a few occurrences 😄
But yeah it's not documented so we can consider it's not public API.

I've pushed your commit on this branch, thanks!

LGTM so will merge once CI passes.