Roman numerals in treetop grammar

Question

I want to parse an ordered list, which is something like:

I - Something
II - Something else...
IX - Something weird
XIII - etc

So far, my treetop grammar is:

rule text
    roman_numeral separator text newline
end

rule roman_numeral
    &. ('MMM' / 'MM' / 'M')? (('C' [DM]) / 
    ('D'? ('CCC' / 'CC' / 'C')?))? (('X' [LC]) / 
    ('L'? ('XXX' / 'XX' / 'X')?))? (('I' [VX]) / 
    ('V'? ('III' / 'II' / 'I')?))?
end

rule separator
    [\s] "-" [\s]
end

rule text
    (!"\n" .)*
end

rule newline
    ["\n"]
end

However, the corresponding parser is unable to parse the text. What is broken?

Answer 1

You accidentally overloaded text. Rename the first to line, and then add another rule for lines.

The quotes around newline also seem unnecessary.

Side tip - you can reuse the newline rule in your text rule to keep it DRY.

grammar Roman

  rule lines
    line*
  end

  rule line
    roman_numeral separator text newline
  end

  rule roman_numeral
    &. ('MMM' / 'MM' / 'M')? (('C' [DM]) /
    ('D'? ('CCC' / 'CC' / 'C')?))? (('X' [LC]) /
    ('L'? ('XXX' / 'XX' / 'X')?))? (('I' [VX]) /
    ('V'? ('III' / 'II' / 'I')?))?
  end

  rule separator
    [\s] "-" [\s]
  end

  rule text
    (!newline .)*
  end

  rule newline
    [\n]
  end

end

Update

You can simplify the grammar a little bit by removing the negative lookahead and the single character classes.

rule separator
  " - "
end

rule text
  [^\n]*
end

The resulting syntax graph becomes much simpler.