Question

I am trying to write an interpreter for the programming language Icon. One of the steps in this process is writing a parser for Icon, which I’ve done in the following way:

import java.io.FileReader
import scala.util.parsing.combinator.syntactical._
import scala.util.parsing.combinator.RegexParsers
import scala.util.parsing.combinator.PackratParsers
import scala.util.parsing.combinator.JavaTokenParsers

abstract class expr
case class CstInt(val value : Int) extends expr
case class FromTo(val from : expr, val to : expr) extends expr
case class Write(val value : expr) extends expr
case class And(val e1 : expr, val e2 : expr) extends expr
case class Or(val e1 : expr, val e2 : expr) extends expr

object ExprParser extends JavaTokenParsers with PackratParsers{

lazy val exp : PackratParser[expr] = andexp | exp2

lazy val exp2 : PackratParser[expr] = fromTo | exp3

lazy val exp3 :PackratParser[expr] = orexp | exp4 

lazy val exp4 : PackratParser[expr] = integer | exp5

lazy val exp5 : PackratParser[expr] = write 

lazy val integer : PackratParser[expr] = wholeNumber ^^ { s => CstInt(s.toInt)}

lazy val  write : PackratParser[Write] =  "write" ~> "(" ~> exp <~ ")" ^^ {  e => Write(e)}

lazy val fromTo : PackratParser[FromTo] = ("(" ~> integer) ~ ("to" ~> integer <~ ")") ^^ { case from ~ to => FromTo(from, to)}

lazy val andexp : PackratParser[And] = exp ~ ("&" ~> exp) ^^ { case e1 ~ e2 => And(e1, e2)}

lazy val orexp : PackratParser[Or] = exp ~ ("|" ~> exp) ^^ { case e1 ~ e2 => Or(e1, e2)}

def parseInput(input: String) : expr =
    parseAll (exp, input) match {
        case Success(tree, _) => tree
        case e: NoSuccess => throw new IllegalArgumentException(e.toString())
    }

}

object Interpret {
def main(args : Array[String]) : Unit = {
    println(ExprParser.parseInput(args(0)))
    }
}

However, I’ve run into a few problems when I try to parse the following expression:

write((1 to 4) | 4)

I get this error:

java.lang.IllegalArgumentException: [9.17] failure: `)' expected but ` ' found

Whereas parsing

write((1 to 4) & 4)

works just fine. The first expression works fine if I move the orexp parser to an exp group above the fromto parser. However, this does not adhere to the rules given by Icon, and does not solve the underlying problem.

Does anyone have any ideas for solutions? According to the Scala docs, mixing packrat parsers and regular parsers should be ok.

Answer 1

Ok, I have read the paper on packrat parsers in Scala, and I’m afraid this grammar won’t work as is. The problem being that fromTo as exp inside write, and then write itself fails (and, having no other alternatives, the outer exp fails). It never goes back and say “well, let’s see if there’s another exp that is also valid”.

However, looking at this text, I don’t see fromTo having parenthesis as part of its grammar. If it were simply rewritten to remove those parenthesis from that level, it would work:

object ExprParser extends JavaTokenParsers with PackratParsers{
  lazy val exp : PackratParser[expr] = andexp | exp2
  lazy val exp2 : PackratParser[expr] = fromTo | exp3
  lazy val exp3 :PackratParser[expr] = orexp | exp4 
  lazy val exp4 : PackratParser[expr] = integer | exp5
  lazy val exp5 : PackratParser[expr] = write | exp6
  lazy val exp6 : PackratParser[expr] = "(" ~> exp <~ ")"
  lazy val integer : PackratParser[expr] = wholeNumber ^^ { s => CstInt(s.toInt)}
  lazy val  write : PackratParser[Write] =  "write" ~> "(" ~> exp <~ ")" ^^ {  e => Write(e)}
  lazy val fromTo : PackratParser[FromTo] = integer ~ ("to" ~> integer) ^^ { case from ~ to => FromTo(from, to)}
  lazy val andexp : PackratParser[And] = exp ~ ("&" ~> exp) ^^ { case e1 ~ e2 => And(e1, e2)}
  lazy val orexp : PackratParser[Or] = exp3 ~ ("|" ~> exp) ^^ { case e1 ~ e2 => Or(e1, e2)}
}