parsing - Scala: How to combine parser combinators from different objects

Question

Given a family of objects that implement parser combinators, how do I combine the parsers? Since Parsers.Parser is an inner class, and in Scala inner classes are bound to the outer object, the story becomes slightly complicated.

Here's an example that attempts to combine two parsers from different objects.

import scala.util.parsing.combinator._

class BinaryParser extends JavaTokenParsers {
  def anyrep: Parser[Any] = rep(any)
  def any: Parser[Any] = zero | one
  def zero: Parser[Any] = "0"
  def one: Parser[Any] = "1"
}

object LongChainParser extends BinaryParser {
  def parser1: Parser[Any] = zero~zero~one~one
}

object ShortChainParser extends BinaryParser {
  def parser2: Parser[Any] = zero~zero
}

object ExampleParser extends BinaryParser {
  def parser: Parser[Any] = (LongChainParser.parser1
    ||| ShortChainParser.parser2) ~ anyrep

  def main(args: Array[String]) {
    println(parseAll(parser, args(0) ))
  }
}

This results to the following error:

<console>:11: error: type mismatch;
 found   : ShortChainParser.Parser[Any]
 required: LongChainParser.Parser[?]
         def parser: Parser[Any] = (LongChainParser.parser1 
           ||| ShortChainParser.parser2) ~ anyrep

I've found the solution to this problem already, but since it was brought up recently on scala-user ML (Problem injecting one parser into another), it's probably worth putting it here too.

Answer 1

The quick answer is to use the traits instead of hosting the parsers in objects:

import scala.util.parsing.combinator._

trait BinaryParser extends JavaTokenParsers {
  def anyrep: Parser[Any] = rep(any)
  def any: Parser[Any] = zero | one
  def zero: Parser[Any] = "0"
  def one: Parser[Any] = "1"
}

trait LongChainParser extends BinaryParser {
  def parser1: Parser[Any] = zero~zero~one~one
}

trait ShortChainParser extends BinaryParser {
  def parser2: Parser[Any] = zero~zero
}

object ExampleParser extends LongChainParser with ShortChainParser  {
  def parser: Parser[Any] = (parser1 ||| parser2) ~ anyrep

  def main(args: Array[String]) {
    println(parseAll(parser, args(0) ))
  }
}

Because the combinator operators like ~ and | are written against the inner class, escalating the parser references to class-level by saying BinaryParser#Parser[_] doesn't do you any good. Using traits solves all that inner-class issues since both Parser[Any] from LongChainParser and ShortChainParser now refer to the inner class of ExampleParser object.