Parsec parsing multiline Ruby expressions

Learn
#1

Hi,

I’m trying to extend herbalizer, which converts HAML templates to ERB format using the Parsec library.

Right now, ruby expression are parsed as single line only: herbalizer/src/Main.hs at master · danchoi/herbalizer · GitHub

rubyExp = do
  line <- ((:) <$> char '=' >> spaces >> manyTill anyChar newline <* spaces)
  return (RubyExp line)

Indeed, in HAML: “A line of Ruby code can be stretched over multiple lines as long as each line but the last ends with a comma” - File: REFERENCE — Haml Documentation

This is my first attempt at extending that parser expression:

rubyExp = do
  char '='
  spaces
  commaLines <- option [] (try $ many commaLine)
  line <- manyTill anyChar newline
  spaces
  return (RubyExp $ concat commaLines ++ line)
  where 
    commaNewline = string ",\n"
    commaLine = (++) <$> (manyTill anyChar (try (lookAhead commaNewline))) <*> commaNewline

which only matches a multi-line ruby expression string up until the first newline, even though it ends with a comma. The commaLines parser, i.e. the many parser, doesn’t seem to be matching greedily, which is what I was expecting. Where is my thinking error here?

I arrived at a somewhat working solution using this parser expression:

rubyExp = do
  char '='
  spaces
  commaLines <- option [] (try $ many commaLine)
  line <- manyTill anyChar (try (lookAhead(noneOf "," >> newline)))
  lastChar <- noneOf "," <* newline
  spaces
  return (RubyExp $ concat commaLines ++ reverse(lastChar : (reverse line)))
  where 
    commaNewline = string ",\n"
    commaLine = (++) <$> (manyTill anyChar (try (lookAhead commaNewline))) <*> commaNewline

I’m sure this can be expressed more elegantly than this. I’m still very new to Haskell, any help is greatly appreciated!

#2

I do have a few suggestions:

  • Switch from Parsec to Megaparsec, Attoparsec, or another maintained parser library. A PEG parser library like Text.Grampa.PEG.Backtrack might be a better fit for this use case.
  • Simplify option [] (try $ many commaLine) to many (try commaLine).
  • Since commaLines should consume all the lines ending with comma, you shouldn’t worry about the last line ending with comma. You can replace (try (lookAhead(noneOf "," >> newline))) with just newLine and drop the lastChar and reverse wrangling.
  • You can use the sepBy1 combinator to make the parser a bit more obvious.
1 Like