Matching a name in preference to a literal

[ndw]

That title probably doesn't help much, but I'm not quite sure how best to categorize the problem. Here's a tiny grammar:

module.exports = grammar({
  name: 'examples',

  extras: ($) => [
    /\s/
  ],

  rules: {
    source_file: ($) => seq(optional($.prolog), $.rule),

    rule: ($) => seq($.name, /[:=]/, $.name, '.'),

    prolog: ($) => $.version,

    version: ($) => seq("file", "version", $.string, '.'),
    
    name: ($) => /[a-zA-Z]+/,

    string: ($) => seq('"', /[^\"]*/, '"'),
  }
});

And here's a small corpus:

===============
a rule
===============

S: a .

---

(source_file
  (rule
    (name)
    (name)))

===============
a rule with a prolog
===============

file version "7" .
file: a .

---

(source_file
  (prolog
    (version
      (string)))
  (rule
    (name)
    (name)))

===============
a rule named file
===============

file: a .

---

(source_file
  (rule
    (name)
    (name)))

The first two tests work fine, the last test fails because (I gather) file has been matched against the prolog version. I would have thought that the : following file in the rule would be enough to work out the correct match.

Can this be fixed? I've poked about in various ways with precedence without succes.

[ahlinc]

Your case can be fixed simply by defining the name rule as:

name: $ => choice(/[a-zA-Z]+/, "file"),

[ndw]

Thank you. Not one of the several things I tried!