Akka Streams的问题' GraphStage

时间:2017-03-25 10:24:51

标签: scala akka akka-stream

我需要编写一个GraphStage但遇到了一些问题。 我已经将代码浓缩到下面,希望你们能为我解释一下。

下面的示例代码不是我的真实用例,它只是在这里证明我的观点。希望这是我不了解akka流而不是它的局限性。

示例代码使用WrapFlowShape构建一个Graph,并基本上重定向" in"附图流中的图形和" out"图表的流程图。

import akka.actor.ActorSystem
import akka.stream._
import akka.stream.javadsl.RunnableGraph
import akka.stream.scaladsl.{Flow, GraphDSL, Sink, Source}
import akka.stream.stage.{GraphStage, GraphStageLogic, InHandler, OutHandler}

import scala.collection.immutable
import scala.io.StdIn

object WrapFlowSandbox extends App {
  case class WrapFlowShape[I, O](
      in: Inlet[I],
      out: Outlet[O],
      flowIn: Inlet[O],
      flowOut: Outlet[I]) extends Shape {
    val inlets: immutable.Seq[Inlet[_]] = in :: flowIn :: Nil
    val outlets: immutable.Seq[Outlet[_]] = out :: flowOut :: Nil
    def deepCopy = WrapFlowShape(in.carbonCopy, out.carbonCopy, flowIn.carbonCopy, flowOut.carbonCopy)
  }
  class WrapFlow[I, O] extends GraphStage[WrapFlowShape[I, O]] {
    val in: Inlet[I] = Inlet[I]("WrapFlow.in")
    val out: Outlet[O] = Outlet[O]("WrapFlow.out")
    val flowIn: Inlet[O] = Inlet[O](s"Select.flowIn")
    val flowOut: Outlet[I] = Outlet[I](s"Select.flowOut")
    val shape: WrapFlowShape[I, O] = WrapFlowShape(in, out, flowIn, flowOut)
    def createLogic(initialAttributes: Attributes): GraphStageLogic = new GraphStageLogic(shape) {
      var inElem: I = _
      setHandler(in, new InHandler {
        def onPush = {
          println("2 in.onPush")
          inElem = grab(in)
          pull(flowIn)
        }
      })
      setHandler(out, new OutHandler {
        def onPull = {
          println("1 out.onPull")
          pull(in)
        }
      })
      setHandler(flowIn, new InHandler {
        def onPush = {
          println("4 flowIn.onPush")
          val outElem = grab(flowIn)
          push(out, outElem)
        }
      })
      setHandler(flowOut, new OutHandler {
        def onPull = {
          println("3 flowOut.onPull")
          push(flowOut, inElem)
        }
      })
    }
  }
  implicit val system = ActorSystem()
  implicit val materializer = ActorMaterializer()
  val flow = Flow[Int].map(_ + 1)
  RunnableGraph.fromGraph(GraphDSL.create() { implicit b =>
    import GraphDSL.Implicits._
    val select = b.add(new WrapFlow[Int, Int])
    Source.single(1) ~> select.in
    select.out ~> Sink.foreach[Int](r => println(s"result = $r"))
    select.flowOut ~> flow ~> select.flowIn
    ClosedShape
  }).run(materializer)
  StdIn.readLine
  system.terminate
}

我希望看到的输出是:

1 out.onPull
2 in.onPush
3 flowOut.onPull
4 flowIn.onPush
result = 2

但实际输出只是前3行:

1 out.onPull
2 in.onPush
3 flowOut.onPull

InHandler.onPush()for" flowIn"永远不会被称为。

我知道以这种方式编写GraphStage是不合常规的,但我确实需要它。

让我感到困惑的是,我通过在步骤2(pull(flowIn))中拉动它来产生对附加流的需求, 而附加的流程反过来又产生了对流量的需求。在第3步。

但在步骤3中通过flowOut推送元素后,元素从未被推送,因此从未执行过步骤4。

为什么会这样?

如果附加流量感知到下游需求并在步骤3生成上游需求,为什么在步骤3推送的元素没有通过附加流?

1 个答案:

答案 0 :(得分:1)

不确定我是否遵循处理程序中的逻辑。我根据您对(V1, V2) => V内容的理解将其修改为以下内容:

GraphDSL.create()

执行它应该产生以下输出:

def createLogic(initialAttributes: Attributes): GraphStageLogic = new GraphStageLogic(shape) {
  var inElem: I = _
  setHandler(in, new InHandler {
    def onPush = {
      println("in.onPush")
      inElem = grab(in)
      push(flowOut, inElem)
    }
  })
  setHandler(out, new OutHandler {
    def onPull = {
      println("out.onPull")
      pull(flowIn)
    }
  })
  setHandler(flowIn, new InHandler {
    def onPush = {
      println("flowIn.onPush")
      val outElem = grab(flowIn)
      push(out, outElem)
    }
  })
  setHandler(flowOut, new OutHandler {
    def onPull = {
      println("flowOut.onPull")
      pull(in)
    }
  })
}

注意到out.onPull flowOut.onPull in.onPush flowIn.onPush result = 2 案例类(不是抽象类)中的方法copyFromPorts()未被覆盖。我相信你需要用以下内容覆盖它:

WrapFlowShape