Accessing the underlying ActorRef of an akka stream Source created by Source.actorRef

Question

I'm trying to use the Source.actorRef method to create an akka.stream.scaladsl.Source object. Something of the form

import akka.stream.OverflowStrategy.fail
import akka.stream.scaladsl.Source

case class Weather(zip : String, temp : Double, raining : Boolean)

val weatherSource = Source.actorRef[Weather](Int.MaxValue, fail)

val sunnySource = weatherSource.filter(!_.raining)
...

My question is: how do I send data to my ActorRef based Source object?

I assumed sending messages to the Source was something of the form

//does not compile
weatherSource ! Weather("90210", 72.0, false)
weatherSource ! Weather("02139", 32.0, true)

But weatherSource doesn't have a ! operator or tell method.

The documentation isn't too descriptive on how to use Source.actorRef, it just says you can...

Thank you in advance for your review and response.

Answer 1

You need a Flow:

  import akka.stream.OverflowStrategy.fail
  import akka.stream.scaladsl.Source
  import akka.stream.scaladsl.{Sink, Flow}

  case class Weather(zip : String, temp : Double, raining : Boolean)

  val weatherSource = Source.actorRef[Weather](Int.MaxValue, fail)

  val sunnySource = weatherSource.filter(!_.raining)

  val ref = Flow[Weather]
    .to(Sink.ignore)
    .runWith(sunnySource)

  ref ! Weather("02139", 32.0, true)

Remember this is all experimental and may change!

Answer 2

As @Noah points out the experimental nature of akka-streams, his answer might not work with the 1.0 release. I had to follow the example given by this example:

implicit val materializer = ActorMaterializer()
val (actorRef: ActorRef, publisher: Publisher[TweetInfo]) = Source.actorRef[TweetInfo](1000, OverflowStrategy.fail).toMat(Sink.publisher)(Keep.both).run()
actorRef ! TweetInfo(...)
val source: Source[TweetInfo, Unit] = Source[TweetInfo](publisher)

Answer 3

Instance of ActorRef, like all 'materialized values', will become accessible only once whole stream is materialized, or, in other words, when RunnableGraph is being run.

// RunnableGraph[ActorRef] means that you get ActorRef when you run the graph
val rg1: RunnableGraph[ActorRef] = sunnySource.to(Sink.foreach(println))

// You get ActorRef instance as a materialized value
val actorRef1: ActorRef = rg1.run()

// Or even more correct way: to materialize both ActorRef and future to completion 
// of the stream, so that we know when we are done:

// RunnableGraph[(ActorRef, Future[Done])] means that you get tuple
// (ActorRef, Future[Done]) when you run the graph
val rg2: RunnableGraph[(ActorRef, Future[Done])] =
  sunnySource.toMat(Sink.foreach(println))(Keep.both)

// You get both ActorRef and Future[Done] instances as materialized values
val (actorRef2, future) = rg2.run()

actorRef2 ! Weather("90210", 72.0, false)
actorRef2 ! Weather("02139", 32.0, true)
actorRef2 ! akka.actor.Status.Success("Done!") // Complete the stream
future onComplete { /* ... */ }