Proposal: Make it easier to specify (and comment) a list of illegals

[joshgoebel]

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

Yes, illegal can become hard to read when it includes multiple items.

Any alternative solutions you considered...

I saw this in routeros which brought this to my attention (again).

      { // illegal constructs
        variants: [
          { begin: /^@/, end: /$/ },               // dns
          { begin: /\/\*/, end: /\*\// },          // -- comment
          { begin: /%%/, end: /$/ },               // -- comment
          { begin: /^'/, end: /$/ },               // Monkey one line comment
          { begin: /^\s*\/[\w-]+=/, end: /$/ },    // jboss-cli
          { begin: /\/\//, end: /$/ },             // Stan comment
          { begin: /^\[</, end: />\]$/ },          // F# class declaration?
          { begin: /<\//, end: />/ },              // HTML tags
          { begin: /^facet /, end: /\}/ },         // roboconf - лютый костыль )))
          { begin: '^1\\.\\.(\\d+)$', end: /$/ }   // tap
        ],
        illegal: /./
      },

But this is very verbose and you have to think about it for a minute to realize what it's doing.

Note: Do not focus on if this is a GOOD idea for routeros to have such a long list (it may not be)... that's not a concern here... there are definitely cases where a list of 2-3 illegals makes very good sense. I only show this as one example of how you'd do this today (get a more readable illegal list) without any core library changes.

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

Instead I'd like to formalize that as:

{
    illegal: [
      /^@/,      // dns
      /\/\*/,    // C style multi-line comment
      /%%/,      // some other comment...
      // ...
    ],
}

And these would be auto-joined with regex.either... The older single regex or string format would still be supported. But if we saw an array here when we're call regex.either on it...

[joshgoebel]

Very much related: #2833

This is exactly the kind of thing this related proposal would make trivial:

// mode grammar plugin
(mode, parent) => {
  if (typeof mode.illegal === "object") {
    mode.illegal = regex.either(...mode.illegal)
  }
} 

A ~5 line plugin PLUS the tests for this would also be trivial since we only need to test the transform itself, not the whole parser.

I mean obviously that isn't a super complex example, but I think it shows how clean and easy it would be to add tiny components in a modular fashion vs just piling more and more internal code into compileMode.