How can I prioritize breaking out of a "repeat", when continued repeating is also valid?

[Hubro]

I'm trying to improve the Robot tree-sitter grammar by adding support for parsing the built-in conditional keywords.

Basically I'm trying to parse this:

Example
    IF    ${rc} > 0
        Some keyword
        Another keyword
    END

This is my attempt at matching the conditional statement:

    conditional_statement: $ => seq(
      "IF",
      $._separator,
      $.argument,
      $._line_break,
      repeat1(choice($.statement, $._empty_line)),
      "END",
    ),

(For context, here is the full grammar)

The issue is, END is technically a completely valid statement in Robot Framework. Any statement in Robot is a keyword followed by 0 or more arguments. A keyword can be literally any string, except for the built-in constructs like IF, END, FOR etc.

This means that with the above code, the END is parsed as just another normal statement by the repeat1, and the parse output eventually ends with (MISSING "END")

The "exclude" rule would have been perfect in this situation, since I could just exclude built-in keywords like "IF" and "END" from being valid keyword names, but that PR was abandoned 😞

I have peppered my code with every combination of conflicts, prec.left and prec.dynamic I can imagine, but nothing has had any impact.

Is there any way I can get Tree-sitter to prioritize matching something after the repeat rule rather than keep repeating?

[amaanq]

tree-sitter will prioritize matching literals over a regex that could match END, however, if "END" is a valid choice in statement then that can be matched too

[Hubro]

Dang, turns out I just banged my head against this problem for too long. After a good night's sleep, it seemed obvious immediately. This grammar matches leading indentation as a token, and I forgot to match one before the "END" rule. Replacing "END" with seq($._indentation, "END") fixes the problem.