A little performance test for two approaches to resolving duplicates between two lists

fitler-first-vs-distinct-after.scala

// -- Measuring ---

def time[A](label: String, count: Int = 100)(block: => A): Long = {
  // the first test run on the console is compiling it
  block

  // Take the average of <count> runs
  (1 to count).toSeq.map { i =>
    val t0 = System.nanoTime()
    block
    val t1 = System.nanoTime()
    t1 - t0
  }.sum / count
}

case class SubData(id: Int)
case class Data(value: String, sub: SubData)

// -- Setup ---

val sub1 = SubData(1)
val sub2 = SubData(2)
val sub3 = SubData(3)
val sub4 = SubData(4)
val sub5 = SubData(5)

val dataSet1 = Seq(
  Data("foo", sub1),
  Data("bar", sub2),
  Data("baz", sub3)
)

val dataSet2 = Seq(
  Data("lorem", sub4),
  Data("foo", sub1),
  Data("ipsum", sub5)
)

// -- Test1 ---

def filterFirst = time("filter first") {
  val idsFromSet1 = dataSet1.map(_.sub.id)
  val dataSet2Prime = dataSet2.filter { d =>
    d.sub.id != 0 &&
    !idsFromSet1.contains(d.sub.id)
  }

  dataSet1 ++ dataSet2
}


// -- Test2 ---

def distinctAfter = time("distinct after") {
  val dataSet2Prime = dataSet2.filter(_.sub.id != 0)
  (dataSet1 ++ dataSet2Prime).distinct
}

Results for my machine:

filterFirst = 5650
distinctAfter = 19830