Some benchmarking tests to localize differences between "drop sampling" and "filter sampling"

benchmark_drop.scala

import scala.util.Random

def benchmark_drop(n: Int = 1000, m: Int = 1000, k: Int = 10) {
  val rng = new Random()

  def dropIter[T](d: Iterator[T]) {
    var dd = d
    while (! dd.isEmpty) dd = dd.drop(rng.nextInt())
  }

  var data:Seq[Int] = (1 to m).toArray.toSeq
  println("\nArray")

  var start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    dropIter(data.iterator)
  }}
  println(s"  Time iterator: ${System.currentTimeMillis() - start}")  


  data = (1 to m).toList
  println("\nList")

  start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    dropIter(data.iterator)
  }}
  println(s"  Time iterator: ${System.currentTimeMillis() - start}")  
}


def benchmark_random_log(n: Int = 1000) {
  var start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    Math.random()
  }}
  println(s"random: ${System.currentTimeMillis() - start}")  
  
  start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    Math.log(Math.random())
  }}
  println(s"log random: ${System.currentTimeMillis() - start}")  

  val ln2 = Math.log(2.0)
  start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    Math.log(Math.random())/ln2
  }}
  println(s"log random div: ${System.currentTimeMillis() - start}")  

  start = System.currentTimeMillis()
  (1 to n).foreach { _ => {
    (Math.log(Math.random())/ln2).toInt
  }}
  println(s"log random div int: ${System.currentTimeMillis() - start}")  
}