[Request] Multi-match for end matchers

[joshgoebel]

Is your request related to a specific problem you're having?

Working on JSX I end up with the following mess:

{
    end: regex.lookahead(/<\/[A-Za-z-]+>|\/>/),
    starts: {
      contains: [
        { match: /\/>/, className: "tag", endsParent: true },
        { match:
          [/<\//, /[A-Za-z-]+/,/>/],
          className: {1: "tag", 2: "name", 3:"tag"},
          endsParent: true },
      ]
}

Once we find the end (with a look-ahead) we have to jump into a new mode (with starts) and two rules using multi-class on match/begin to highlight the individual pieces of tag as well as using endParent to prevent those same modes from gobbling up anymore then a SINGLE end tag.

And of course the look-ahead also prevents END_SAME_AS_BEGIN since we have no capture.

The solution you'd prefer / feature you'd like to see added...

Not sure.

{
  begin: ...,
  end: [ /</, /[A-Za-z-]+/, />/],
  scope: { 
    1: "tag", 2: ... // scopes for `begin`
    end: {1: "tag.bracket", 2: "tag.name", 3:"tag.bracket" }}
}

This is problematic though because eventually I'd like to use begin and end to allow scoping the ENTIRE begin and end block individually, ie:

{
  begin: /"/, end: /"/
  scope: { begin: "string.quote", end: "string.quote", middle: "string" }

Though I suppose a string vs an object is easy to differentiate at runtime.

A dedicated key is another option:

{
  end: [ /</, /[A-Za-z-]+/, />/ ],
  endScope: { 1: "tag.bracket", 2: "tag.name", 3:"tag.bracket" }
}

Any alternative solutions you considered...

TextMate grammars use captures, beginCaptures, endCaptures. Making a case for a separate key. A fuller example:

{
  begin: [ /</, /[A-Za-z-]+/ ],
  beginScope: {1: "tag.bracket", 2:"tag.name"},
  end: [ /</, /[A-Za-z-]+/, />/ ],
  endScope: { 1: "tag.bracket", 2: "tag.name", 3:"tag.bracket" }
}

With scope and match being sugar for beginScope and begin.

Additional context...

This is the next obvious improvement for multi-match.

[joshgoebel]

CC @highlightjs/core

[joshgoebel]

Oh but we still have the legacy behavior of scope/className as a string referring to the ENTIRE block... so that is problematic.

{
  begin: /"/, 
  end: /"/,
  scope: "string" // refers to the whole shebang: ".*"
}

So here scope vs beginScope would have VERY different output.

[joshgoebel]

I think I'd also be fine going ultra-explicit and all the places we're currently using scope: [] we'd simply change them to beginScope... so then scope/className would retain it's prior behavior... string or nothing, and wrap the entire block. And if you want to get specific, use *Scope.

{
  begin: /"/, 
  end: /"/,
  scope: "string",
  beginScope: "string.delim",
  endScope: "string.delim",
}

"hello world"

<span class='string'><span class="string.delim">&quot;</span>hello world<span class="string.delim">&quot;</span></span>