Akka Streams - How to keep materialized value of an auxiliary Sink in a Graph

Question

I have a function returning a Flow whose logic involves passing some elements of the graph to an auxiliary Sink passed as a parameter. I want to retain the auxiliary Sink's materialized value so I'm able to act upon its value when the constructed stream is launched.

Here's a rough picture of the flow I'm building:

IN ~> (logic: In => Out) ~> Broadcast ~> AuxFilter ~> AuxSink
                                      ~> OutFilter ~> OUT

Sample code:

case class Incoming()
trait Element
case class Outcoming() extends Element
case class Persistent() extends Element

def flow[Mat](auxSink: Sink[Persistent, Mat]): Flow[Incoming, Outcoming, NotUsed] = {
  val isPersistent = Flow[Element].collect {
    case persistent: Persistent => persistent
  }

  val isRunning = Flow[Element].collect {
    case out: Outcoming => out
  }

  val magicFlow: Flow[Incoming, Element, NotUsed] = Flow[Incoming]
    .map(_ => if (Random.nextBoolean()) Outcoming() else Persistent())

  Flow.fromGraph {
    GraphDSL.create() { implicit b =>
      import GraphDSL.Implicits._

      val magic = b.add(magicFlow)
      val bcast = b.add(Broadcast[Element](2))
      val sink = b.add(isRunning)

                   bcast.out(0) ~> isPersistent ~> auxSink
      magic.out ~> bcast.in
                   bcast.out(1) ~> isRunning ~> sink.in

      FlowShape(magic.in, sink.out)
    }
  }
}

Is there a way to somehow pass the auxSink's Mat to the resulting Flow?

Thanks.

Answer 1

Answering my own question...

Found it! The source of Flow.alsoToMat pointed me to exactly the logic I needed - to access the materialized value of an auxiliary graph (in my case auxSink), one has to import its shape into the graph being constructed by passing it as a parameter to GraphDSL.create().

def flow[Mat](auxSink: Sink[Persistent, Mat]): Flow[Incoming, Outcoming, Mat] = {
  val isPersistent = ...
  val isRunning = ...
  val magicFlow = ...

  Flow.fromGraph {
    GraphDSL.create(auxSink) { implicit b => aux =>
      import GraphDSL.Implicits._

      val magic = b.add(magicFlow)
      val bcast = b.add(Broadcast[Element](2))
      val sink = b.add(isRunning)

      magic ~> bcast ~> isPersistent ~> aux
               bcast ~> isRunning ~> sink

      FlowShape(magic.in, sink.out)
    }
  }
}