Functional listener with state

Question

Let's say in my pure Scala program i have a dependency to a Java service. This Java service accepts a listener to notify me when some data changes.

Let's say the data is a tuple(x, y) and the java service calls the listener whenever X or Y changes but i'm interested only when X.

For this my listener has to save the last value of X, and forward the update/call only when oldX != X, so in order to have this my impure scala listener implementation has to hold a var oldX

val listener = new JavaServiceListener() {

 var oldX;
 def updated(val x, val y): Unit = {
  if (oldX != x) {
     oldX = x
    //do stuff
  }
}

javaService.register(listener)

How would i go about to design a wrapper for this kind of thing in Scala without val or mutable collections ? I can't at the JavaServiceListener level since i'm bound by the method signature, so I need another layer above which the java listener forwards to somehow

Answer 1

I found the solution I like with Cats and Cats-Effect:

trait MyListener {
  def onChange(n: Int): Unit
}

class MyDistinctFunctionalListener(private val last: Ref[IO, Int], consumer: Int => Unit) extends MyListener {
  override def onChange(newValue: Int): Unit = {
    val program =
      last
        .getAndSet(newValue)
        .flatMap(oldValue => notify(newValue, oldValue))

    program.unsafeRunSync()
  }

  private def notify(newValue: Int, oldValue: Int): IO[Unit] = {
    if (oldValue != newValue) IO(consumer(newValue)) else IO.delay(println("found duplicate"))
  }
}

object MyDistinctFunctionalListener {
  def create(consumer: Int => Unit): IO[MyDistinctFunctionalListener] =
    Ref[IO].of(0).map(v => new MyDistinctFunctionalListener(v, consumer))
}

val printer: Int => Unit = println(_)

val functionalDistinctPrinterIO =  MyDistinctFunctionalListener.create(printer)

functionalDistinctPrinterIO.map(fl =>
  List(1, 1, 2, 2, 3, 3, 3, 4, 5, 5).foreach(fl.onChange)
).unsafeRunSync()

More stuff about handling shared state here https://github.com/systemfw/scala-italy-2018

it is debatable if this is worth it over the private var solution