Discuss: Restricting language subset used for sublanguage detection

[chrisvfritz]

When limiting auto highlight to a language subset, it only works for top-level languages. So when parsing some HTML with JavaScript inside with:

hljs.highlightAuto(
  code,
  [ 'html', 'css', 'javascript' ]
).value

the JavaScript may still be interpreted as ActionScript, for example. 😕

[isagalaev]

Good catch, thanks. And that should be fixable in a fairly straightforward way. I'll try to get to this one before the next release.

[chrisvfritz]

😄 Thanks!

[joshgoebel]

I think the problem here is misstated - or at least the implied solution is not correct. If the same filters used for language were applied to sublanguage someone could easily break many grammars without realizing it. I don't think this is how users would expect things to behave in general.

Perhaps this merely requires a documentation fix, but I'm not so sure. Many languages are vitally dependent on particular sublanguages and if I only specified the parent language in the subset I would NOT expect that to restrict the sublanguage subset - hence badly breaking the parent language.

Some examples:

• xml/php
• xml/vbscript_html
• many of the templating languages that require xml
• javascript/xml (for JSX support)

I'm not sure what the proper solution here is, but I wanted to add these notes.

[joshgoebel]

the JavaScript may still be interpreted as ActionScript, for example.

Was that the real problem here? I wonder if we couldn't simply remove ActionScript completely (from the autodetect list for <script> tags in xml)... isn't flash pretty much dead now? No one else has commented on this issue so I'd be happy to lose actionscript as a sublanguage for script tags and close this out with that.

Thoughts?

CC @egor-rogov @marcoscaceres

[egor-rogov]

I wonder if we couldn't simply remove ActionScript completely (from the autodetect list for <script> tags in xml)... isn't flash pretty much dead now? No one else has commented on this issue so I'd be happy to lose actionscript as a sublanguage for script tags and close this out with that.

I think it can be a solution for this particular issue.

In general we lack a mechanism to limit sublanguages (outside from the grammar). I agree that

if I only specified the parent language in the subset I would NOT expect that to restrict the sublanguage subset

but could a separate restriction of sublanguage subset be convenient in some cases?

[joshgoebel]

but could a separate restriction of sublanguage subset be convenient in some cases?

Perhaps. Though I think it'd be very hacky and still subject to the caveats I mentioned above (far more likely to break things than to fix things). If you're fine will removing ActionScript from the auto-detection for <script> then I say we do that for now, close this - and see if it ever comes up again or not.

If so perhaps when/if it does we'll have more information and perhaps that'll lead to better ideas.

[egor-rogov]

I'm fine with it.