为什么这组解析器组合器会溢出堆栈？

Question

TL;DR

我的 EBNF 语法规范的解析器组合器溢出了堆栈。为什么？我该如何解决？

背景

我正在尝试通过 scala 库中的组合器为 EBNF syntax 定义解析器。实际上，代码构建了语法的 AST，但我已经删除了这些位并内联了一个实用方法以生成 MVCE（如下）。

问题

所写的代码在运行时出现堆栈溢出（也在下方）。我无法理解的是它似乎在解析的跳过空白部分溢出了。我该如何解决这个错误？如果无法解析 EBNF 语法，那将是非常不幸的——我打算为此开发一些工具。

MVCE

package org.benknoble.ebnf

import scala.util.parsing.combinator._

class EbnfParserSimple extends RegexParsers {

  def nonterminal: Parser[String] = """<[^>]+>""".r

  def goesTo: Parser[String] = """::="""

  def epsilon: Parser[String] = "ε"

  def terminal: Parser[String] = "\"[^\"]+\"".r

  def sequence: Parser[String] =
    exp ~ exp ^^ { case left ~ right => left + right }

  def alternation: Parser[String] =
    exp ~ "|" ~ exp ^^ { case left ~ _ ~ right => left + "|" + right }

  def opt: Parser[String] =
    "[" ~ exp ~ "]" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def repetition: Parser[String] =
    "{" ~ exp ~ "}" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def group: Parser[String] =
    "(" ~ exp ~ ")" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def exp: Parser[String] =
    (epsilon
      | terminal
      | nonterminal
      | sequence
      | alternation
      | opt
      | repetition
      | group)

  def rule: Parser[String] = nonterminal ~ goesTo ~ exp ~ ";" ^^ {
    case nt ~ delim ~ e ~ semi => nt + delim + e + semi
  }

  def join[A](sep: String, list: Seq[A]): String = list match {
    case h :: t => h.toString() + t.foldLeft("")(_.toString() + sep + _.toString())
    case Nil => ""
  }

  def root: Parser[String] = phrase(
    rep(rule) ^^ {
      case rules =>
        val joined = join(" ;\n", rules)
        if (joined.isEmpty)
          joined
        else
          joined + " ;"
    }
  )

}

object Main extends App {
  val parser = new EbnfParserSimple()
  val grammar = """<A> ::= ["a"|ε]"c" ; """
// <B> ::= <A>"b" ;
// <C> ::= {<B>}"$" ;
// <D> ::= "a""b""d" ;
// <E> ::= ("a"|"b")"c" ;
// """
  val rule = parser.root
  println(parser.parse(rule, grammar))
}

错误跟踪

可以找到完整日志 as a Gist。

[info] Loading project definition from /Users/Knoble/loner/project
[info] Loading settings for project root from build.sbt ...
[info] Set current project to loner (in build file:/Users/Knoble/loner/)
[info] Set current project to ebnf (in build file:/Users/Knoble/loner/)
[info] Running org.benknoble.ebnf.Main 
[error] (run-main-0) java.lang.WhosebugError
[error] java.lang.WhosebugError
[error]     at java.util.regex.Pattern.compile(Pattern.java:1673)
[error]     at java.util.regex.Pattern.<init>(Pattern.java:1351)
[error]     at java.util.regex.Pattern.compile(Pattern.java:1028)
[error]     at scala.util.matching.Regex.<init>(Regex.scala:226)
[error]     at scala.collection.immutable.StringLike.r(StringLike.scala:284)
[error]     at scala.collection.immutable.StringLike.r$(StringLike.scala:284)
[error]     at scala.collection.immutable.StringOps.r(StringOps.scala:33)
[error]     at scala.collection.immutable.StringLike.r(StringLike.scala:273)
[error]     at scala.collection.immutable.StringLike.r$(StringLike.scala:273)
[error]     at scala.collection.immutable.StringOps.r(StringOps.scala:33)
[error]     at org.benknoble.ebnf.EbnfParserSimple.terminal(EbnfParser_strings.scala:13)
[error]     at org.benknoble.ebnf.EbnfParserSimple.$anonfun$exp(EbnfParser_strings.scala:32)
[error]     at scala.util.parsing.combinator.Parsers$Parser.p$lzycompute(Parsers.scala:253)
[error]     at scala.util.parsing.combinator.Parsers$Parser.p(Parsers.scala:253)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$Failure.append(Parsers.scala:202)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[...]
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$append(Parsers.scala:254)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error]     at scala.util.parsing.combinator.Parsers$Parser.$anonfun$flatMap(Parsers.scala:239)
[error]     at scala.util.parsing.combinator.Parsers$$anon.apply(Parsers.scala:222)
[error] Nonzero exit code: 1
[error] (Compile / run) Nonzero exit code: 1
[error] Total time: 1 s, completed 12 mars 2019 21:11:15

Answer 1

我最终能够通过删除左递归来解决我的问题。下面找到工作代码。

我不得不仔细考虑这些转换：特别是 alternation.+ ^^ { _.reduce(_ + _) } 和 sequence.+ ^^ { _.reduce(_ + _) }——将它们转换回 AST 生成器可能并不简单（因为它们的构造函数只需要左和正确的）。重复也让我有点困扰，但在不提取辅助函数的情况下，这是唯一要做的事情。

package org.benknoble.ebnf

import scala.util.parsing.combinator._

class EbnfParserSimple extends RegexParsers {

  def epsilon: Parser[String] = "ε"

  def terminal: Parser[String] = "\"[^\"]+\"".r

  def nonterminal: Parser[String] = """<[^>]+>""".r

  def opt: Parser[String] =
    "[" ~ exp ~ "]" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def repetition: Parser[String] =
    "{" ~ exp ~ "}" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def group: Parser[String] =
    "(" ~ exp ~ ")" ^^ { case lb ~ e ~ rb => lb + e + rb }

  def alternation: Parser[String] =
    chainl1(epsilon
      | terminal
      | nonterminal
      | opt
      | repetition
      | group,
      "|" ^^^ { (lb: String, rb: String) => lb + "|" + rb })
  //   exp ~ "|" ~ exp ^^ { case left ~ _ ~ right => left + "|" + right }

  def sequence: Parser[String] =
    alternation.+ ^^ { _.reduce(_ + _) }
    // alternation ~ alternation ^^ { case lb ~ rb => lb + rb }
  //   exp ~ exp ^^ { case left ~ right => left + right }

  def exp: Parser[String] =
    sequence.+ ^^ { _.reduce(_ + _) }

  def goesTo: Parser[String] = """::="""

  def rule: Parser[String] = nonterminal ~ goesTo ~ exp ~ ";" ^^ {
    case nt ~ delim ~ e ~ semi => nt + delim + e + semi
  }

  def join[A](sep: String, list: Seq[A]): String = list match {
    case h :: t => h.toString() + t.foldLeft("")(_.toString() + sep + _.toString())
    case Nil => ""
  }

  def root: Parser[String] = phrase(
    rep(rule) ^^ {
      case rules =>
        val joined = join(" ;\n", rules)
        if (joined.isEmpty)
          joined
        else
          joined + " ;"
    }
  )

}

object Main extends App {
  val parser = new EbnfParserSimple()
  val grammar = """<A> ::= ["a"|ε]"c" ; """
// <B> ::= <A>"b" ;
// <C> ::= {<B>}"$" ;
// <D> ::= "a""b""d" ;
// <E> ::= ("a"|"b")"c" ;
// """
  val rule = parser.root
  println(parser.parse(rule, grammar))
}

为什么这组解析器组合器会溢出堆栈？

Why is this set of parser combinators overflowing the stack?

stack-overflow

scala

ebnf

parser-combinators

TL;DR

背景

问题

MVCE

错误跟踪