Scala：过滤强制评估整个Stream

private def placeMines( excludes: List[( Int,Int )] ) {
    def rndstream: Stream[( Int,Int )] = {
        def s: Stream[( Int,Int )] =
            ( Random.nextInt( ysize ),Random.nextInt(  xsize ) ) #:: s
        s
    }
    def posPermitted( pos: ( Int,Int ) ): Boolean = {
        !excludes.contains( pos ) &&
            fieldEmpty( pos._1,pos._2 )
    }
    val positions = rndstream.filter( posPermitted )
    positions.take( minecount ).foreach( x => grid( x._1 )( x._2 ) = MineField() )
}

为了找出发生了什么,我已经注释了具有副作用的最后一行(网格是二维数组)并用不同的替换了过滤谓词,包括x => false和x =>真正.有趣的是,它终结于真实的情况下,但始终以虚假的方式运行.插入一些printlns表明,在终止java进程之前,谓词被调用了数十万次.

我尝试使用以下代码重现该情况：

import scala.util.Random
import org.specs.SpecificationWithJUnit

class foobar extends SpecificationWithJUnit {
    val x = 0xDead
    val y = 0xBeef

    bar(x,y)

    private def bar(x: Int,y: Int) = foo(x)

    private def foo(x: Int) = {
        def s: Stream[( Int,Int )] = {
            def p: Stream[( Int,Int )] =
                ( Random.nextInt( x ),Random.nextInt( y ) ) #:: p
            p
        }
        val fiveodd = s.filter( x => x._1 % 2 == 1 )
        println( fiveodd.take( 5 ).toList )
    }
}

但是,该代码运行得很好.

搜索“scala流过滤器无尽”“scala流过滤器强制评估”和“scala流过滤器不终止”只提出了一些教程,展示了原则上与我的代码相同的流的用法.