ScalaCheck生成BST [英] ScalaCheck generate BST

问题描述

我正在尝试使用ScalaCheck为BST创建Gen，但是当我调用.sample方法时，它会给我java.lang.NullPointerException.我哪里错了?

I'm trying to create a Gen for BST with ScalaCheck, but when I call .sample method it gives me java.lang.NullPointerException. Where do I wrong?

sealed trait Tree

case class Node(left: Tree, right: Tree, v: Int) extends Tree

case object Leaf extends Tree

import org.scalacheck._
import Gen._
import Arbitrary.arbitrary

case class GenerateBST() {

  def genValue(left: Tree, right: Tree): Gen[Int] = (left, right) match {
    case (Node(_, _, min), Node(_, _, max)) => arbitrary[Int].suchThat(x => x > min && x < max)
    case (Node(_, _, min), Leaf) => arbitrary[Int].suchThat(x => x > min)
    case (Leaf, Node(_, _, max)) => arbitrary[Int].suchThat(x => x < max)
    case (Leaf, Leaf) => arbitrary[Int]
  }


  val genNode = for {
    left <- genTree
    right <- genTree
    v <- genValue(left, right)
  } yield Node(left, right, v)

  def genTree: Gen[Tree] = oneOf(const(Leaf), genNode)
}

GenerateBST().genTree.sample

推荐答案

由于您为递归数据类型递归定义生成器的方式，您需要在顶部使用Gen.lzy:

Because of the way you're defining the generator recursively for a recursive data type, you need to use Gen.lzy at the top:

def genTree: Gen[Tree] = Gen.lzy(oneOf(const(Leaf), genNode))

作为无关的说明，在生成器定义中使用suchThat通常仅应作为最后的选择.这意味着sample经常会失败(使用固定版本的代码，大约会浪费三分之一的时间)，更重要的是，如果有一天您想创建导致Tree的任意函数，您将看到一个许多可怕的org.scalacheck.Gen$RetrievalError: couldn't generate value例外.

As an unrelated side note, using suchThat in your generator definitions should generally only be a last resort. It means that sample will often fail (about a third of the time with the fixed version of your code), and more importantly that if you someday want to create arbitrary functions resulting in Tree, you're going to see a lot of horrible org.scalacheck.Gen$RetrievalError: couldn't generate value exceptions.

在这种情况下，您可以通过使用Gen.chooseNum并轻松地替换左侧和右侧(如果顺序错误)来避免suchThat:

In this case you can avoid suchThat pretty easily by using Gen.chooseNum and swapping the left and right sides if they're in the wrong ordered:

sealed trait Tree
case class Node(left: Tree, right: Tree, v: Int) extends Tree
case object Leaf extends Tree

import org.scalacheck.{ Arbitrary, Gen }

object GenerateBST {
  def swapIfNeeded(l: Tree, r: Tree): (Tree, Tree) = (l, r) match {
    // If the two trees don't have space between them, we bump one and recheck:
    case (Node(_, _, x), n @ Node(_, _, y)) if math.abs(x - y) <= 1 =>
      swapIfNeeded(l, n.copy(v = y + 1))
    // If the values are in the wrong order, swap:
    case (Node(_, _, x), Node(_, _, y)) if x > y => (r, l)
    // Otherwise do nothing:
    case (_, _) => (l, r)
  }

  def genValue(left: Tree, right: Tree): Gen[Int] = (left, right) match {
    case (Node(_, _, min), Node(_, _, max)) => Gen.chooseNum(min + 1, max - 1)
    case (Node(_, _, min), Leaf) => Gen.chooseNum(min + 1, Int.MaxValue)
    case (Leaf, Node(_, _, max)) => Gen.chooseNum(Int.MinValue, max - 1)
    case (Leaf, Leaf) => Arbitrary.arbitrary[Int]
  }

  val genNode = for {
    l0 <- genTree
    r0 <- genTree
    (left, right) = swapIfNeeded(l0, r0)
    v <- genValue(left, right)
  } yield Node(left, right, v)

  def genTree: Gen[Tree] = Gen.lzy(Gen.oneOf(Gen.const(Leaf), genNode))
}

现在，您可以使用Arbitrary[Whatever => Tree]而不用担心发电机故障:

Now you can use Arbitrary[Whatever => Tree] without worrying about generator failures:

scala> implicit val arbTree: Arbitrary[Tree] = Arbitrary(GenerateBST.genTree)
arbTree: org.scalacheck.Arbitrary[Tree] = org.scalacheck.ArbitraryLowPriority$$anon$1@606abb0e

scala> val f = Arbitrary.arbitrary[Int => Tree].sample.get
f: Int => Tree = org.scalacheck.GenArities$$Lambda$7109/289518656@13eefeaf

scala> f(1)
res0: Tree = Leaf

scala> f(2)
res1: Tree = Node(Leaf,Leaf,-20313200)

scala> f(3)
res2: Tree = Leaf

scala> f(4)
res3: Tree = Node(Node(Leaf,Leaf,-850041807),Leaf,-1)

这篇关于ScalaCheck生成BST的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！