How to deactivate/exclude a "repository" under some context in a grammar package

#1

I’m now developing a grammar package for a typesetting language named SATySFi. SATySFi has several states, two of which are program state and inline state, and transition between states is made by a dedicated syntax. For example, an opening brace { in a program state starts an inline state and the corresponding } resumes a program state.

In my Atom package, I want to highlight keywords like let in program state but not in inline state because let in inline state is just an English verb. Same can be said for other stuff like integer numerals.

I tried to make a prototype for this context-sensitive feature (see the CSON below). However, this grammar still scopes let in inline state as keyword.other.satysfi. I understand the reason but this is not what I want. I’m looking for how to deactivate/exclude the program state grammar in inline state part and is it possible?

scopeName: "source.satysfi"
name: "SATySFi"
fileTypes: [
  "saty"
]  
patterns: [
  {
    include: "#programState"
  }
  {
    include: "#inlineState"
  }
]
repository:
  programState:
    patterns: [
      {
        begin: "\\{"
        beginCaptures:
          0:
            name: "punctuation.transition.program-to-inline.begin.satysfi"
        end: "\\}"
        endCaptures:
          0:
            name: "punctuation.transition.program-to-inline.end.satysfi"
        patterns: [
          include: "#inlineState"
        ]
      }
      {
        match: "\\blet\\b"
        name: "keyword.other.satysfi"
      }
    ]
  inlineState:
    patterns: [
      # grammar for inline state here.
    ]
#2

I’ll need to test it, but a combination of injection selectors might work. Will update when I experiment.

Edit: is there a demo we can test with? The grammar seems to be working fine as-is. For example, the text let { let } is correctly scoped (the outer one is picked up on, the inner one ignored).

#3

Well, you are right. The reason I get incorrect scope is that I have nested {} strcutures in the actual document and the closing } was parsed incorrectly. I need to develop the grammar more, but that will solve the problem I had today.

Sorry for taking your time and thanks!

#4

And do nested braces mean something important, or are they just like TeX groups when already in inline mode? I think you just need a rule that looks for the opening brace and finishes at a closing brace, to ensure they stay balanced.