Flow.scala

/*
 * Copyright (C) 2014-2019 Lightbend Inc. <https://www.lightbend.com>
 */

package akka.stream.scaladsl

import akka.event.LoggingAdapter
import akka.stream._
import akka.Done
import akka.stream.impl.{
  fusing,
  LinearTraversalBuilder,
  ProcessorModule,
  SetupFlowStage,
  SubFlowImpl,
  Throttle,
  Timers,
  TraversalBuilder
}
import akka.stream.impl.fusing._
import akka.stream.stage._
import akka.util.{ ConstantFun, Timeout }
import org.reactivestreams.{ Processor, Publisher, Subscriber, Subscription }

import scala.annotation.unchecked.uncheckedVariance
import scala.collection.immutable
import scala.concurrent.Future
import scala.concurrent.duration.FiniteDuration
import scala.language.higherKinds
import akka.stream.impl.fusing.FlattenMerge
import akka.NotUsed
import akka.actor.ActorRef
import akka.annotation.DoNotInherit

import scala.annotation.implicitNotFound
import scala.reflect.ClassTag

/**
 * A `Flow` is a set of stream processing steps that has one open input and one open output.
 */
final class Flow[-In, +Out, +Mat](
    override val traversalBuilder: LinearTraversalBuilder,
    override val shape: FlowShape[In, Out])
    extends FlowOpsMat[Out, Mat]
    with Graph[FlowShape[In, Out], Mat] {

  // TODO: debug string
  override def toString: String = s"Flow($shape)"

  override type Repr[+O] = Flow[In @uncheckedVariance, O, Mat @uncheckedVariance]
  override type ReprMat[+O, +M] = Flow[In @uncheckedVariance, O, M]

  override type Closed = Sink[In @uncheckedVariance, Mat @uncheckedVariance]
  override type ClosedMat[+M] = Sink[In @uncheckedVariance, M]

  private[stream] def isIdentity: Boolean = this.traversalBuilder eq Flow.identityTraversalBuilder

  override def via[T, Mat2](flow: Graph[FlowShape[Out, T], Mat2]): Repr[T] = viaMat(flow)(Keep.left)

  override def viaMat[T, Mat2, Mat3](flow: Graph[FlowShape[Out, T], Mat2])(
      combine: (Mat, Mat2) => Mat3): Flow[In, T, Mat3] = {
    if (this.isIdentity) {
      // optimization by returning flow if possible since we know Mat2 == Mat3 from flow
      if (combine == Keep.right) Flow.fromGraph(flow).asInstanceOf[Flow[In, T, Mat3]]
      else {
        // Keep.none is optimized and we know left means Mat3 == NotUsed
        val useCombine =
          if (combine == Keep.left) Keep.none
          else combine
        new Flow(LinearTraversalBuilder.empty().append(flow.traversalBuilder, flow.shape, useCombine), flow.shape)
          .asInstanceOf[Flow[In, T, Mat3]]
      }
    } else if (flow.traversalBuilder eq Flow.identityTraversalBuilder) {
      // optimization by returning this if possible since we know Mat2 == Mat from this
      if (combine == Keep.left) this.asInstanceOf[Flow[In, T, Mat3]]
      else {
        // Keep.none is somewhat optimized and we know Mat == NotUsed
        val useCombine =
          if (combine == Keep.right) Keep.none
          else combine
        new Flow(
          traversalBuilder.append(LinearTraversalBuilder.empty(), shape, useCombine),
          FlowShape[In, T](shape.in, flow.shape.out))
      }
    } else {
      new Flow(
        traversalBuilder.append(flow.traversalBuilder, flow.shape, combine),
        FlowShape[In, T](shape.in, flow.shape.out))
    }
  }

  /**
   * Connect this [[Flow]] to a [[Sink]], concatenating the processing steps of both.
   * {{{
   *     +------------------------------+
   *     | Resulting Sink[In, Mat]      |
   *     |                              |
   *     |  +------+          +------+  |
   *     |  |      |          |      |  |
   * In ~~> | flow | ~~Out~~> | sink |  |
   *     |  |   Mat|          |     M|  |
   *     |  +------+          +------+  |
   *     +------------------------------+
   * }}}
   * The materialized value of the combined [[Sink]] will be the materialized
   * value of the current flow (ignoring the given Sink’s value), use
   * [[Flow#toMat[Mat2* toMat]] if a different strategy is needed.
   *
   * See also [[toMat]] when access to materialized values of the parameter is needed.
   */
  def to[Mat2](sink: Graph[SinkShape[Out], Mat2]): Sink[In, Mat] = toMat(sink)(Keep.left)

  /**
   * Connect this [[Flow]] to a [[Sink]], concatenating the processing steps of both.
   * {{{
   *     +----------------------------+
   *     | Resulting Sink[In, M2]     |
   *     |                            |
   *     |  +------+        +------+  |
   *     |  |      |        |      |  |
   * In ~~> | flow | ~Out~> | sink |  |
   *     |  |   Mat|        |     M|  |
   *     |  +------+        +------+  |
   *     +----------------------------+
   * }}}
   * The `combine` function is used to compose the materialized values of this flow and that
   * Sink into the materialized value of the resulting Sink.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def toMat[Mat2, Mat3](sink: Graph[SinkShape[Out], Mat2])(combine: (Mat, Mat2) => Mat3): Sink[In, Mat3] = {
    if (isIdentity) {
      new Sink(LinearTraversalBuilder.fromBuilder(sink.traversalBuilder, sink.shape, combine), SinkShape(sink.shape.in))
        .asInstanceOf[Sink[In, Mat3]]
    } else {
      new Sink(traversalBuilder.append(sink.traversalBuilder, sink.shape, combine), SinkShape(shape.in))
    }
  }

  /**
   * Transform the materialized value of this Flow, leaving all other properties as they were.
   */
  override def mapMaterializedValue[Mat2](f: Mat => Mat2): ReprMat[Out, Mat2] =
    new Flow(traversalBuilder.transformMat(f), shape)

  /**
   * Join this [[Flow]] to another [[Flow]], by cross connecting the inputs and outputs, creating a [[RunnableGraph]].
   * {{{
   * +------+        +-------+
   * |      | ~Out~> |       |
   * | this |        | other |
   * |      | <~In~  |       |
   * +------+        +-------+
   * }}}
   * The materialized value of the combined [[Flow]] will be the materialized
   * value of the current flow (ignoring the other Flow’s value), use
   * [[Flow#joinMat[Mat2* joinMat]] if a different strategy is needed.
   */
  def join[Mat2](flow: Graph[FlowShape[Out, In], Mat2]): RunnableGraph[Mat] = joinMat(flow)(Keep.left)

  /**
   * Join this [[Flow]] to another [[Flow]], by cross connecting the inputs and outputs, creating a [[RunnableGraph]]
   * {{{
   * +------+        +-------+
   * |      | ~Out~> |       |
   * | this |        | other |
   * |      | <~In~  |       |
   * +------+        +-------+
   * }}}
   * The `combine` function is used to compose the materialized values of this flow and that
   * Flow into the materialized value of the resulting Flow.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def joinMat[Mat2, Mat3](flow: Graph[FlowShape[Out, In], Mat2])(combine: (Mat, Mat2) => Mat3): RunnableGraph[Mat3] = {
    val resultBuilder =
      traversalBuilder.append(flow.traversalBuilder, flow.shape, combine).wire(flow.shape.out, shape.in)

    RunnableGraph(resultBuilder)
  }

  /**
   * Join this [[Flow]] to a [[BidiFlow]] to close off the “top” of the protocol stack:
   * {{{
   * +---------------------------+
   * | Resulting Flow            |
   * |                           |
   * | +------+        +------+  |
   * | |      | ~Out~> |      | ~~> O1
   * | | flow |        | bidi |  |
   * | |      | <~In~  |      | <~~ I2
   * | +------+        +------+  |
   * +---------------------------+
   * }}}
   * The materialized value of the combined [[Flow]] will be the materialized
   * value of the current flow (ignoring the [[BidiFlow]]’s value), use
   * [[Flow#joinMat[I2* joinMat]] if a different strategy is needed.
   */
  def join[I2, O1, Mat2](bidi: Graph[BidiShape[Out, O1, I2, In], Mat2]): Flow[I2, O1, Mat] = joinMat(bidi)(Keep.left)

  /**
   * Join this [[Flow]] to a [[BidiFlow]] to close off the “top” of the protocol stack:
   * {{{
   * +---------------------------+
   * | Resulting Flow            |
   * |                           |
   * | +------+        +------+  |
   * | |      | ~Out~> |      | ~~> O1
   * | | flow |        | bidi |  |
   * | |      | <~In~  |      | <~~ I2
   * | +------+        +------+  |
   * +---------------------------+
   * }}}
   * The `combine` function is used to compose the materialized values of this flow and that
   * [[BidiFlow]] into the materialized value of the resulting [[Flow]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def joinMat[I2, O1, Mat2, M](bidi: Graph[BidiShape[Out, O1, I2, In], Mat2])(
      combine: (Mat, Mat2) => M): Flow[I2, O1, M] = {
    val newBidiShape = bidi.shape.deepCopy()
    val newFlowShape = shape.deepCopy()

    val resultBuilder =
      TraversalBuilder
        .empty()
        .add(traversalBuilder, newFlowShape, Keep.right)
        .add(bidi.traversalBuilder, newBidiShape, combine)
        .wire(newFlowShape.out, newBidiShape.in1)
        .wire(newBidiShape.out2, newFlowShape.in)

    val newShape = FlowShape(newBidiShape.in2, newBidiShape.out1)

    new Flow(LinearTraversalBuilder.fromBuilder(resultBuilder, newShape, Keep.right), newShape)
  }

  /**
   * Replace the attributes of this [[Flow]] with the given ones. If this Flow is a composite
   * of multiple graphs, new attributes on the composite will be less specific than attributes
   * set directly on the individual graphs of the composite.
   *
   * Note that this operation has no effect on an empty Flow (because the attributes apply
   * only to the contained processing operators).
   */
  override def withAttributes(attr: Attributes): Repr[Out] =
    new Flow(traversalBuilder.setAttributes(attr), shape)

  /**
   * Add the given attributes to this [[Flow]]. If the specific attribute was already present
   * on this graph this means the added attribute will be more specific than the existing one.
   * If this Flow is a composite of multiple graphs, new attributes on the composite will be
   * less specific than attributes set directly on the individual graphs of the composite.
   */
  override def addAttributes(attr: Attributes): Repr[Out] = withAttributes(traversalBuilder.attributes and attr)

  /**
   * Add a ``name`` attribute to this Flow.
   */
  override def named(name: String): Repr[Out] = addAttributes(Attributes.name(name))

  /**
   * Put an asynchronous boundary around this `Flow`
   */
  override def async: Repr[Out] = super.async.asInstanceOf[Repr[Out]]

  /**
   * Put an asynchronous boundary around this `Flow`
   *
   * @param dispatcher Run the graph on this dispatcher
   */
  override def async(dispatcher: String): Repr[Out] =
    super.async(dispatcher).asInstanceOf[Repr[Out]]

  /**
   * Put an asynchronous boundary around this `Flow`
   *
   * @param dispatcher      Run the graph on this dispatcher
   * @param inputBufferSize Set the input buffer to this size for the graph
   */
  override def async(dispatcher: String, inputBufferSize: Int): Repr[Out] =
    super.async(dispatcher, inputBufferSize).asInstanceOf[Repr[Out]]

  /**
   * Connect the `Source` to this `Flow` and then connect it to the `Sink` and run it. The returned tuple contains
   * the materialized values of the `Source` and `Sink`, e.g. the `Subscriber` of a of a [[Source#subscriber]] and
   * and `Publisher` of a [[Sink#publisher]].
   *
   * Note that the `ActorSystem` can be used as the implicit `materializer` parameter to use the
   * [[akka.stream.SystemMaterializer]] for running the stream.
   */
  def runWith[Mat1, Mat2](source: Graph[SourceShape[In], Mat1], sink: Graph[SinkShape[Out], Mat2])(
      implicit materializer: Materializer): (Mat1, Mat2) =
    Source.fromGraph(source).via(this).toMat(sink)(Keep.both).run()

  /**
   * Converts this Flow to a [[RunnableGraph]] that materializes to a Reactive Streams [[org.reactivestreams.Processor]]
   * which implements the operations encapsulated by this Flow. Every materialization results in a new Processor
   * instance, i.e. the returned [[RunnableGraph]] is reusable.
   *
   * @return A [[RunnableGraph]] that materializes to a Processor when run() is called on it.
   */
  def toProcessor: RunnableGraph[Processor[In @uncheckedVariance, Out @uncheckedVariance]] =
    Source
      .asSubscriber[In]
      .via(this)
      .toMat(Sink.asPublisher[Out](false))(Keep.both[Subscriber[In], Publisher[Out]])
      .mapMaterializedValue {
        case (sub, pub) =>
          new Processor[In, Out] {
            override def onError(t: Throwable): Unit = sub.onError(t)
            override def onSubscribe(s: Subscription): Unit = sub.onSubscribe(s)
            override def onComplete(): Unit = sub.onComplete()
            override def onNext(t: In): Unit = sub.onNext(t)
            override def subscribe(s: Subscriber[_ >: Out]): Unit = pub.subscribe(s)
          }
      }

  /**
   * Turns a Flow into a FlowWithContext which manages a context per element along a stream.
   *
   * @param collapseContext turn each incoming pair of element and context value into an element of this Flow
   * @param extractContext turn each outgoing element of this Flow into an outgoing context value
   */
  def asFlowWithContext[U, CtxU, CtxOut](collapseContext: (U, CtxU) => In)(
      extractContext: Out => CtxOut): FlowWithContext[U, CtxU, Out, CtxOut, Mat] =
    new FlowWithContext(
      Flow[(U, CtxU)]
        .map {
          case (e, ctx) =>
            collapseContext(e, ctx)
        }
        .viaMat(this)(Keep.right)
        .map(e => (e, extractContext(e))))

  /** Converts this Scala DSL element to it's Java DSL counterpart. */
  def asJava[JIn <: In]: javadsl.Flow[JIn, Out @uncheckedVariance, Mat @uncheckedVariance] =
    new javadsl.Flow(this)
}

object Flow {
  private[stream] val identityTraversalBuilder =
    LinearTraversalBuilder.fromBuilder(GraphStages.identity.traversalBuilder, GraphStages.identity.shape, Keep.right)

  private[this] val identity: Flow[Any, Any, NotUsed] =
    new Flow[Any, Any, NotUsed](identityTraversalBuilder, GraphStages.identity.shape)

  /**
   * Creates a Flow from a Reactive Streams [[org.reactivestreams.Processor]]
   */
  def fromProcessor[I, O](processorFactory: () => Processor[I, O]): Flow[I, O, NotUsed] = {
    fromProcessorMat(() => (processorFactory(), NotUsed))
  }

  /**
   * Creates a Flow from a Reactive Streams [[org.reactivestreams.Processor]] and returns a materialized value.
   */
  def fromProcessorMat[I, O, M](processorFactory: () => (Processor[I, O], M)): Flow[I, O, M] =
    fromGraph(ProcessorModule(processorFactory))

  /**
   * Returns a `Flow` which outputs all its inputs.
   */
  def apply[T]: Flow[T, T, NotUsed] = identity.asInstanceOf[Flow[T, T, NotUsed]]

  /**
   * Creates a [Flow] which will use the given function to transform its inputs to outputs. It is equivalent
   * to `Flow[T].map(f)`
   */
  def fromFunction[A, B](f: A => B): Flow[A, B, NotUsed] = apply[A].map(f)

  /**
   * A graph with the shape of a flow logically is a flow, this method makes
   * it so also in type.
   */
  def fromGraph[I, O, M](g: Graph[FlowShape[I, O], M]): Flow[I, O, M] =
    g match {
      case f: Flow[I, O, M]                                       => f
      case f: javadsl.Flow[I, O, M]                               => f.asScala
      case g: GraphStageWithMaterializedValue[FlowShape[I, O], M] =>
        // move these from the operator itself to make the returned source
        // behave as it is the operator with regards to attributes
        val attrs = g.traversalBuilder.attributes
        val noAttrStage = g.withAttributes(Attributes.none)
        new Flow(
          LinearTraversalBuilder.fromBuilder(noAttrStage.traversalBuilder, noAttrStage.shape, Keep.right),
          noAttrStage.shape).withAttributes(attrs)

      case _ => new Flow(LinearTraversalBuilder.fromBuilder(g.traversalBuilder, g.shape, Keep.right), g.shape)
    }

  /**
   * Defers the creation of a [[Flow]] until materialization. The `factory` function
   * exposes [[Materializer]] which is going to be used during materialization and
   * [[Attributes]] of the [[Flow]] returned by this method.
   */
  def fromMaterializer[T, U, M](factory: (Materializer, Attributes) => Flow[T, U, M]): Flow[T, U, Future[M]] =
    Flow.fromGraph(new SetupFlowStage(factory))

  /**
   * Defers the creation of a [[Flow]] until materialization. The `factory` function
   * exposes [[ActorMaterializer]] which is going to be used during materialization and
   * [[Attributes]] of the [[Flow]] returned by this method.
   */
  @deprecated("Use 'fromMaterializer' instead", "2.6.0")
  def setup[T, U, M](factory: (ActorMaterializer, Attributes) => Flow[T, U, M]): Flow[T, U, Future[M]] =
    Flow.fromGraph(new SetupFlowStage((materializer, attributes) =>
      factory(ActorMaterializerHelper.downcast(materializer), attributes)))

  /**
   * Creates a `Flow` from a `Sink` and a `Source` where the Flow's input
   * will be sent to the Sink and the Flow's output will come from the Source.
   *
   * The resulting flow can be visualized as:
   * {{{
   *     +----------------------------------------------+
   *     | Resulting Flow[I, O, NotUsed]                |
   *     |                                              |
   *     |  +---------+                  +-----------+  |
   *     |  |         |                  |           |  |
   * I  ~~> | Sink[I] | [no-connection!] | Source[O] | ~~> O
   *     |  |         |                  |           |  |
   *     |  +---------+                  +-----------+  |
   *     +----------------------------------------------+
   * }}}
   *
   * The completion of the Sink and Source sides of a Flow constructed using
   * this method are independent. So if the Sink receives a completion signal,
   * the Source side will remain unaware of that. If you are looking to couple
   * the termination signals of the two sides use `Flow.fromSinkAndSourceCoupled` instead.
   *
   * See also [[fromSinkAndSourceMat]] when access to materialized values of the parameters is needed.
   */
  def fromSinkAndSource[I, O](sink: Graph[SinkShape[I], _], source: Graph[SourceShape[O], _]): Flow[I, O, NotUsed] =
    fromSinkAndSourceMat(sink, source)(Keep.none)

  /**
   * Creates a `Flow` from a `Sink` and a `Source` where the Flow's input
   * will be sent to the Sink and the Flow's output will come from the Source.
   *
   * The resulting flow can be visualized as:
   * {{{
   *     +-------------------------------------------------------+
   *     | Resulting Flow[I, O, M]                              |
   *     |                                                      |
   *     |  +-------------+                  +---------------+  |
   *     |  |             |                  |               |  |
   * I  ~~> | Sink[I, M1] | [no-connection!] | Source[O, M2] | ~~> O
   *     |  |             |                  |               |  |
   *     |  +-------------+                  +---------------+  |
   *     +------------------------------------------------------+
   * }}}
   *
   * The completion of the Sink and Source sides of a Flow constructed using
   * this method are independent. So if the Sink receives a completion signal,
   * the Source side will remain unaware of that. If you are looking to couple
   * the termination signals of the two sides use `Flow.fromSinkAndSourceCoupledMat` instead.
   *
   * The `combine` function is used to compose the materialized values of the `sink` and `source`
   * into the materialized value of the resulting [[Flow]].
   */
  def fromSinkAndSourceMat[I, O, M1, M2, M](sink: Graph[SinkShape[I], M1], source: Graph[SourceShape[O], M2])(
      combine: (M1, M2) => M): Flow[I, O, M] =
    fromGraph(GraphDSL.create(sink, source)(combine) { _ => (in, out) =>
      FlowShape(in.in, out.out)
    })

  /**
   * Allows coupling termination (cancellation, completion, erroring) of Sinks and Sources while creating a Flow from them.
   * Similar to [[Flow.fromSinkAndSource]] however couples the termination of these two operators.
   *
   * The resulting flow can be visualized as:
   * {{{
   *     +---------------------------------------------+
   *     | Resulting Flow[I, O, NotUsed]               |
   *     |                                             |
   *     |  +---------+                 +-----------+  |
   *     |  |         |                 |           |  |
   * I  ~~> | Sink[I] | ~~~(coupled)~~~ | Source[O] | ~~> O
   *     |  |         |                 |           |  |
   *     |  +---------+                 +-----------+  |
   *     +---------------------------------------------+
   * }}}
   *
   * E.g. if the emitted [[Flow]] gets a cancellation, the [[Source]] of course is cancelled,
   * however the Sink will also be completed. The table below illustrates the effects in detail:
   *
   * <table>
   *   <tr>
   *     <th>Returned Flow</th>
   *     <th>Sink (<code>in</code>)</th>
   *     <th>Source (<code>out</code>)</th>
   *   </tr>
   *   <tr>
   *     <td><i>cause:</i> upstream (sink-side) receives completion</td>
   *     <td><i>effect:</i> receives completion</td>
   *     <td><i>effect:</i> receives cancel</td>
   *   </tr>
   *   <tr>
   *     <td><i>cause:</i> upstream (sink-side) receives error</td>
   *     <td><i>effect:</i> receives error</td>
   *     <td><i>effect:</i> receives cancel</td>
   *   </tr>
   *   <tr>
   *     <td><i>cause:</i> downstream (source-side) receives cancel</td>
   *     <td><i>effect:</i> completes</td>
   *     <td><i>effect:</i> receives cancel</td>
   *   </tr>
   *   <tr>
   *     <td><i>effect:</i> cancels upstream, completes downstream</td>
   *     <td><i>effect:</i> completes</td>
   *     <td><i>cause:</i> signals complete</td>
   *   </tr>
   *   <tr>
   *     <td><i>effect:</i> cancels upstream, errors downstream</td>
   *     <td><i>effect:</i> receives error</td>
   *     <td><i>cause:</i> signals error or throws</td>
   *   </tr>
   *   <tr>
   *     <td><i>effect:</i> cancels upstream, completes downstream</td>
   *     <td><i>cause:</i> cancels</td>
   *     <td><i>effect:</i> receives cancel</td>
   *   </tr>
   * </table>
   *
   * See also [[fromSinkAndSourceCoupledMat]] when access to materialized values of the parameters is needed.
   */
  def fromSinkAndSourceCoupled[I, O](
      sink: Graph[SinkShape[I], _],
      source: Graph[SourceShape[O], _]): Flow[I, O, NotUsed] =
    fromSinkAndSourceCoupledMat(sink, source)(Keep.none)

  /**
   * Allows coupling termination (cancellation, completion, erroring) of Sinks and Sources while creating a Flow from them.
   * Similar to [[Flow.fromSinkAndSource]] however couples the termination of these two operators.
   *
   * The resulting flow can be visualized as:
   * {{{
   *     +-----------------------------------------------------+
   *     | Resulting Flow[I, O, M]                             |
   *     |                                                     |
   *     |  +-------------+                 +---------------+  |
   *     |  |             |                 |               |  |
   * I  ~~> | Sink[I, M1] | ~~~(coupled)~~~ | Source[O, M2] | ~~> O
   *     |  |             |                 |               |  |
   *     |  +-------------+                 +---------------+  |
   *     +-----------------------------------------------------+
   * }}}
   *
   * E.g. if the emitted [[Flow]] gets a cancellation, the [[Source]] of course is cancelled,
   * however the Sink will also be completed. The table on [[Flow.fromSinkAndSourceCoupled]]
   * illustrates the effects in detail.
   *
   * The `combine` function is used to compose the materialized values of the `sink` and `source`
   * into the materialized value of the resulting [[Flow]].
   */
  def fromSinkAndSourceCoupledMat[I, O, M1, M2, M](sink: Graph[SinkShape[I], M1], source: Graph[SourceShape[O], M2])(
      combine: (M1, M2) => M): Flow[I, O, M] =
    // format: OFF
    Flow.fromGraph(GraphDSL.create(sink, source)(combine) { implicit b => (i, o) =>
      import GraphDSL.Implicits._
      val bidi = b.add(new CoupledTerminationBidi[I, O])
      /* bidi.in1 ~> */ bidi.out1 ~> i; o ~> bidi.in2 /* ~> bidi.out2 */
      FlowShape(bidi.in1, bidi.out2)
    })
  // format: ON

  /**
   * Creates a real `Flow` upon receiving the first element. Internal `Flow` will not be created
   * if there are no elements, because of completion, cancellation, or error.
   *
   * The materialized value of the `Flow` is the value that is created by the `fallback` function.
   *
   * '''Emits when''' the internal flow is successfully created and it emits
   *
   * '''Backpressures when''' the internal flow is successfully created and it backpressures
   *
   * '''Completes when''' upstream completes and all elements have been emitted from the internal flow
   *
   * '''Cancels when''' downstream cancels
   */
  @deprecated(
    "Use 'Flow.futureFlow' in combination with prefixAndTail(1) instead, see `futureFlow` operator docs for details",
    "2.6.0")
  def lazyInit[I, O, M](flowFactory: I => Future[Flow[I, O, M]], fallback: () => M): Flow[I, O, M] =
    Flow.fromGraph(new LazyFlow[I, O, M](flowFactory)).mapMaterializedValue(_ => fallback())

  /**
   * Creates a real `Flow` upon receiving the first element. Internal `Flow` will not be created
   * if there are no elements, because of completion, cancellation, or error.
   *
   * The materialized value of the `Flow` is a `Future[Option[M]]` that is completed with `Some(mat)` when the internal
   * flow gets materialized or with `None` when there where no elements. If the flow materialization (including
   * the call of the `flowFactory`) fails then the future is completed with a failure.
   *
   * '''Emits when''' the internal flow is successfully created and it emits
   *
   * '''Backpressures when''' the internal flow is successfully created and it backpressures
   *
   * '''Completes when''' upstream completes and all elements have been emitted from the internal flow
   *
   * '''Cancels when''' downstream cancels
   */
  @deprecated("Use 'Flow.lazyFutureFlow' instead", "2.6.0")
  def lazyInitAsync[I, O, M](flowFactory: () => Future[Flow[I, O, M]]): Flow[I, O, Future[Option[M]]] =
    Flow.fromGraph(new LazyFlow[I, O, M](_ => flowFactory())).mapMaterializedValue { v =>
      implicit val ec = akka.dispatch.ExecutionContexts.sameThreadExecutionContext
      v.map[Option[M]](Some.apply _).recover { case _: NeverMaterializedException => None }
    }

  /**
   * Turn a `Future[Flow]` into a flow that will consume the values of the source when the future completes successfully.
   * If the `Future` is completed with a failure the stream is failed.
   *
   * The materialized future value is completed with the materialized value of the future flow or failed with a
   * [[NeverMaterializedException]] if upstream fails or downstream cancels before the future has completed.
   */
  def futureFlow[I, O, M](flow: Future[Flow[I, O, M]]): Flow[I, O, Future[M]] =
    lazyFutureFlow(() => flow)

  /**
   * Defers invoking the `create` function to create a future flow until there is downstream demand and passing
   * that downstream demand upstream triggers the first element.
   *
   * The materialized future value is completed with the materialized value of the created flow when that has successfully
   * been materialized.
   *
   * If the `create` function throws or returns a future that fails the stream is failed, in this case the materialized
   * future value is failed with a [[NeverMaterializedException]].
   *
   * Note that asynchronous boundaries (and other operators) in the stream may do pre-fetching which counter acts
   * the laziness and can trigger the factory earlier than expected.
   *
   * '''Emits when''' the internal flow is successfully created and it emits
   *
   * '''Backpressures when''' the internal flow is successfully created and it backpressures or downstream backpressures
   *
   * '''Completes when''' upstream completes and all elements have been emitted from the internal flow
   *
   * '''Cancels when''' downstream cancels
   */
  def lazyFlow[I, O, M](create: () => Flow[I, O, M]): Flow[I, O, Future[M]] =
    lazyFutureFlow(() => Future.successful(create()))

  /**
   * Defers invoking the `create` function to create a future flow until there downstream demand has caused upstream
   * to send a first element.
   *
   * The materialized future value is completed with the materialized value of the created flow when that has successfully
   * been materialized.
   *
   * If the `create` function throws or returns a future that fails the stream is failed, in this case the materialized
   * future value is failed with a [[NeverMaterializedException]].
   *
   * Note that asynchronous boundaries (and other operators) in the stream may do pre-fetching which counter acts
   * the laziness and can trigger the factory earlier than expected.
   *
   * '''Emits when''' the internal flow is successfully created and it emits
   *
   * '''Backpressures when''' the internal flow is successfully created and it backpressures or downstream backpressures
   *
   * '''Completes when''' upstream completes and all elements have been emitted from the internal flow
   *
   * '''Cancels when''' downstream cancels
   */
  def lazyFutureFlow[I, O, M](create: () => Future[Flow[I, O, M]]): Flow[I, O, Future[M]] =
    Flow.fromGraph(new LazyFlow(_ => create()))

}

object RunnableGraph {

  /**
   * A graph with a closed shape is logically a runnable graph, this method makes
   * it so also in type.
   */
  def fromGraph[Mat](g: Graph[ClosedShape, Mat]): RunnableGraph[Mat] =
    g match {
      case r: RunnableGraph[Mat] => r
      case other                 => RunnableGraph(other.traversalBuilder)
    }
}

/**
 * Flow with attached input and output, can be executed.
 */
final case class RunnableGraph[+Mat](override val traversalBuilder: TraversalBuilder) extends Graph[ClosedShape, Mat] {
  override def shape = ClosedShape

  /**
   * Transform only the materialized value of this RunnableGraph, leaving all other properties as they were.
   */
  def mapMaterializedValue[Mat2](f: Mat => Mat2): RunnableGraph[Mat2] =
    copy(traversalBuilder.transformMat(f.asInstanceOf[Any => Any]))

  /**
   * Run this flow and return the materialized instance from the flow.
   *
   * Note that the `ActorSystem` can be used as the implicit `materializer` parameter to use the
   * [[akka.stream.SystemMaterializer]] for running the stream.
   */
  def run()(implicit materializer: Materializer): Mat = materializer.materialize(this)

  override def addAttributes(attr: Attributes): RunnableGraph[Mat] =
    withAttributes(traversalBuilder.attributes and attr)

  override def withAttributes(attr: Attributes): RunnableGraph[Mat] =
    new RunnableGraph(traversalBuilder.setAttributes(attr))

  override def named(name: String): RunnableGraph[Mat] =
    addAttributes(Attributes.name(name))

  /**
   * Note that an async boundary around a runnable graph does not make sense
   */
  override def async: RunnableGraph[Mat] =
    super.async.asInstanceOf[RunnableGraph[Mat]]

  /**
   * Note that an async boundary around a runnable graph does not make sense
   */
  override def async(dispatcher: String): RunnableGraph[Mat] =
    super.async(dispatcher).asInstanceOf[RunnableGraph[Mat]]

  /**
   * Note that an async boundary around a runnable graph does not make sense
   */
  override def async(dispatcher: String, inputBufferSize: Int): RunnableGraph[Mat] =
    super.async(dispatcher, inputBufferSize).asInstanceOf[RunnableGraph[Mat]]

  /** Converts this Scala DSL element to it's Java DSL counterpart. */
  def asJava: javadsl.RunnableGraph[Mat] = javadsl.RunnableGraph.fromGraph(this)
}

/**
 * Scala API: Operations offered by Sources and Flows with a free output side: the DSL flows left-to-right only.
 *
 * INTERNAL API: this trait will be changed in binary-incompatible ways for classes that are derived from it!
 * Do not implement this interface outside the Akka code base!
 *
 * Binary compatibility is only maintained for callers of this trait’s interface.
 */
@DoNotInherit
trait FlowOps[+Out, +Mat] {
  import akka.stream.impl.Stages._
  import GraphDSL.Implicits._

  type Repr[+O] <: FlowOps[O, Mat] {
    type Repr[+OO] = FlowOps.this.Repr[OO]
    type Closed = FlowOps.this.Closed
  }

  // result of closing a Source is RunnableGraph, closing a Flow is Sink
  type Closed

  /**
   * Transform this [[Flow]] by appending the given processing steps.
   * {{{
   *     +---------------------------------+
   *     | Resulting Flow[In, T, Mat]  |
   *     |                                 |
   *     |  +------+             +------+  |
   *     |  |      |             |      |  |
   * In ~~> | this |  ~~Out~~>   | flow | ~~> T
   *     |  |   Mat|             |     M|  |
   *     |  +------+             +------+  |
   *     +---------------------------------+
   * }}}
   * The materialized value of the combined [[Flow]] will be the materialized
   * value of the current flow (ignoring the other Flow’s value), use
   * [[Flow#viaMat viaMat]] if a different strategy is needed.
   *
   * See also [[FlowOpsMat.viaMat]] when access to materialized values of the parameter is needed.
   */
  def via[T, Mat2](flow: Graph[FlowShape[Out, T], Mat2]): Repr[T]

  /**
   * Recover allows to send last element on failure and gracefully complete the stream
   * Since the underlying failure signal onError arrives out-of-band, it might jump over existing elements.
   * This operator can recover the failure signal, but not the skipped elements, which will be dropped.
   *
   * Throwing an exception inside `recover` _will_ be logged on ERROR level automatically.
   *
   * '''Emits when''' element is available from the upstream or upstream is failed and pf returns an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or upstream failed with exception pf can handle
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def recover[T >: Out](pf: PartialFunction[Throwable, T]): Repr[T] = via(Recover(pf))

  /**
   * RecoverWith allows to switch to alternative Source on flow failure. It will stay in effect after
   * a failure has been recovered so that each time there is a failure it is fed into the `pf` and a new
   * Source may be materialized.
   *
   * Since the underlying failure signal onError arrives out-of-band, it might jump over existing elements.
   * This operator can recover the failure signal, but not the skipped elements, which will be dropped.
   *
   * Throwing an exception inside `recoverWith` _will_ be logged on ERROR level automatically.
   *
   * '''Emits when''' element is available from the upstream or upstream is failed and element is available
   * from alternative Source
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or upstream failed with exception pf can handle
   *
   * '''Cancels when''' downstream cancels
   *
   */
  @deprecated("Use recoverWithRetries instead.", "2.4.4")
  def recoverWith[T >: Out](pf: PartialFunction[Throwable, Graph[SourceShape[T], NotUsed]]): Repr[T] =
    via(new RecoverWith(-1, pf))

  /**
   * RecoverWithRetries allows to switch to alternative Source on flow failure. It will stay in effect after
   * a failure has been recovered up to `attempts` number of times so that each time there is a failure
   * it is fed into the `pf` and a new Source may be materialized. Note that if you pass in 0, this won't
   * attempt to recover at all.
   *
   * A negative `attempts` number is interpreted as "infinite", which results in the exact same behavior as `recoverWith`.
   *
   * Since the underlying failure signal onError arrives out-of-band, it might jump over existing elements.
   * This operator can recover the failure signal, but not the skipped elements, which will be dropped.
   *
   * Throwing an exception inside `recoverWithRetries` _will_ be logged on ERROR level automatically.
   *
   * '''Emits when''' element is available from the upstream or upstream is failed and element is available
   * from alternative Source
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or upstream failed with exception pf can handle
   *
   * '''Cancels when''' downstream cancels
   *
   * @param attempts Maximum number of retries or -1 to retry indefinitely
   * @param pf Receives the failure cause and returns the new Source to be materialized if any
   *
   */
  def recoverWithRetries[T >: Out](
      attempts: Int,
      pf: PartialFunction[Throwable, Graph[SourceShape[T], NotUsed]]): Repr[T] =
    via(new RecoverWith(attempts, pf))

  /**
   * While similar to [[recover]] this operator can be used to transform an error signal to a different one *without* logging
   * it as an error in the process. So in that sense it is NOT exactly equivalent to `recover(t => throw t2)` since recover
   * would log the `t2` error.
   *
   * Since the underlying failure signal onError arrives out-of-band, it might jump over existing elements.
   * This operator can recover the failure signal, but not the skipped elements, which will be dropped.
   *
   * Similarly to [[recover]] throwing an exception inside `mapError` _will_ be logged.
   *
   * '''Emits when''' element is available from the upstream or upstream is failed and pf returns an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or upstream failed with exception pf can handle
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def mapError(pf: PartialFunction[Throwable, Throwable]): Repr[Out] = via(MapError(pf))

  /**
   * Transform this stream by applying the given function to each of the elements
   * as they pass through this processing step.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the mapping function returns an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def map[T](f: Out => T): Repr[T] = via(Map(f))

  /**
   * This is a simplified version of `wireTap(Sink)` that takes only a simple function.
   * Elements will be passed into this "side channel" function, and any of its results will be ignored.
   *
   * If the wire-tap operation is slow (it backpressures), elements that would've been sent to it will be dropped instead.
   * It is similar to [[#alsoTo]] which does backpressure instead of dropping elements.
   *
   * This operation is useful for inspecting the passed through element, usually by means of side-effecting
   * operations (such as `println`, or emitting metrics), for each element without having to modify it.
   *
   * For logging signals (elements, completion, error) consider using the [[log]] operator instead,
   * along with appropriate `ActorAttributes.logLevels`.
   *
   * '''Emits when''' upstream emits an element; the same element will be passed to the attached function,
   *                  as well as to the downstream operator
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def wireTap(f: Out => Unit): Repr[Out] =
    wireTap(Sink.foreach(f)).named("wireTap")

  /**
   * Transform each input element into an `Iterable` of output elements that is
   * then flattened into the output stream.
   *
   * The returned `Iterable` MUST NOT contain `null` values,
   * as they are illegal as stream elements - according to the Reactive Streams specification.
   *
   * '''Emits when''' the mapping function returns an element or there are still remaining elements
   * from the previously calculated collection
   *
   * '''Backpressures when''' downstream backpressures or there are still remaining elements from the
   * previously calculated collection
   *
   * '''Completes when''' upstream completes and all remaining elements have been emitted
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def mapConcat[T](f: Out => immutable.Iterable[T]): Repr[T] = statefulMapConcat(() => f)

  /**
   * Transform each input element into an `Iterable` of output elements that is
   * then flattened into the output stream. The transformation is meant to be stateful,
   * which is enabled by creating the transformation function anew for every materialization —
   * the returned function will typically close over mutable objects to store state between
   * invocations. For the stateless variant see [[FlowOps.mapConcat]].
   *
   * The returned `Iterable` MUST NOT contain `null` values,
   * as they are illegal as stream elements - according to the Reactive Streams specification.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the mapping function returns an element or there are still remaining elements
   * from the previously calculated collection
   *
   * '''Backpressures when''' downstream backpressures or there are still remaining elements from the
   * previously calculated collection
   *
   * '''Completes when''' upstream completes and all remaining elements has been emitted
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.mapConcat]]
   */
  def statefulMapConcat[T](f: () => Out => immutable.Iterable[T]): Repr[T] =
    via(new StatefulMapConcat(f))

  /**
   * Transform this stream by applying the given function to each of the elements
   * as they pass through this processing step. The function returns a `Future` and the
   * value of that future will be emitted downstream. The number of Futures
   * that shall run in parallel is given as the first argument to ``mapAsync``.
   * These Futures may complete in any order, but the elements that
   * are emitted downstream are in the same order as received from upstream.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Stop]]
   * the stream will be completed with failure.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Resume]] or
   * [[akka.stream.Supervision.Restart]] the element is dropped and the stream continues.
   *
   * The function `f` is always invoked on the elements in the order they arrive.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the Future returned by the provided function finishes for the next element in sequence
   *
   * '''Backpressures when''' the number of futures reaches the configured parallelism and the downstream
   * backpressures or the first future is not completed
   *
   * '''Completes when''' upstream completes and all futures have been completed and all elements have been emitted
   *
   * '''Cancels when''' downstream cancels
   *
   * @see [[#mapAsyncUnordered]]
   */
  def mapAsync[T](parallelism: Int)(f: Out => Future[T]): Repr[T] =
    if (parallelism == 1) mapAsyncUnordered[T](parallelism = 1)(f) // optimization for parallelism 1
    else via(MapAsync(parallelism, f))

  /**
   * Transform this stream by applying the given function to each of the elements
   * as they pass through this processing step. The function returns a `Future` and the
   * value of that future will be emitted downstream. The number of Futures
   * that shall run in parallel is given as the first argument to ``mapAsyncUnordered``.
   * Each processed element will be emitted downstream as soon as it is ready, i.e. it is possible
   * that the elements are not emitted downstream in the same order as received from upstream.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Stop]]
   * the stream will be completed with failure.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Resume]] or
   * [[akka.stream.Supervision.Restart]] the element is dropped and the stream continues.
   *
   * The function `f` is always invoked on the elements in the order they arrive (even though the result of the futures
   * returned by `f` might be emitted in a different order).
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' any of the Futures returned by the provided function complete
   *
   * '''Backpressures when''' the number of futures reaches the configured parallelism and the downstream backpressures
   *
   * '''Completes when''' upstream completes and all futures have been completed and all elements have been emitted
   *
   * '''Cancels when''' downstream cancels
   *
   * @see [[#mapAsync]]
   */
  def mapAsyncUnordered[T](parallelism: Int)(f: Out => Future[T]): Repr[T] = via(MapAsyncUnordered(parallelism, f))

  /**
   * Use the `ask` pattern to send a request-reply message to the target `ref` actor.
   * If any of the asks times out it will fail the stream with a [[akka.pattern.AskTimeoutException]].
   *
   * Do not forget to include the expected response type in the method call, like so:
   *
   * {{{
   * flow.ask[ExpectedReply](ref)
   * }}}
   *
   * otherwise `Nothing` will be assumed, which is most likely not what you want.
   *
   * Defaults to parallelism of 2 messages in flight, since while one ask message may be being worked on, the second one
   * still be in the mailbox, so defaulting to sending the second one a bit earlier than when first ask has replied maintains
   * a slightly healthier throughput.
   *
   * Similar to the plain ask pattern, the target actor is allowed to reply with `akka.util.Status`.
   * An `akka.util.Status#Failure` will cause the operator to fail with the cause carried in the `Failure` message.
   *
   * The operator fails with an [[akka.stream.WatchedActorTerminatedException]] if the target actor is terminated.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the futures (in submission order) created by the ask pattern internally are completed
   *
   * '''Backpressures when''' the number of futures reaches the configured parallelism and the downstream backpressures
   *
   * '''Completes when''' upstream completes and all futures have been completed and all elements have been emitted
   *
   * '''Fails when''' the passed in actor terminates, or a timeout is exceeded in any of the asks performed
   *
   * '''Cancels when''' downstream cancels
   */
  @implicitNotFound("Missing an implicit akka.util.Timeout for the ask() operator")
  def ask[S](ref: ActorRef)(implicit timeout: Timeout, tag: ClassTag[S]): Repr[S] =
    ask(2)(ref)(timeout, tag)

  /**
   * Use the `ask` pattern to send a request-reply message to the target `ref` actor.
   * If any of the asks times out it will fail the stream with a [[akka.pattern.AskTimeoutException]].
   *
   * Do not forget to include the expected response type in the method call, like so:
   *
   * {{{
   * flow.ask[ExpectedReply](parallelism = 4)(ref)
   * }}}
   *
   * otherwise `Nothing` will be assumed, which is most likely not what you want.
   *
   * Parallelism limits the number of how many asks can be "in flight" at the same time.
   * Please note that the elements emitted by this operator are in-order with regards to the asks being issued
   * (i.e. same behaviour as mapAsync).
   *
   * The operator fails with an [[akka.stream.WatchedActorTerminatedException]] if the target actor is terminated,
   * or with an [[java.util.concurrent.TimeoutException]] in case the ask exceeds the timeout passed in.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the futures (in submission order) created by the ask pattern internally are completed
   *
   * '''Backpressures when''' the number of futures reaches the configured parallelism and the downstream backpressures
   *
   * '''Completes when''' upstream completes and all futures have been completed and all elements have been emitted
   *
   * '''Fails when''' the passed in actor terminates, or a timeout is exceeded in any of the asks performed
   *
   * '''Cancels when''' downstream cancels
   */
  @implicitNotFound("Missing an implicit akka.util.Timeout for the ask() operator")
  def ask[S](parallelism: Int)(ref: ActorRef)(implicit timeout: Timeout, tag: ClassTag[S]): Repr[S] = {
    val askFlow = Flow[Out]
      .watch(ref)
      .mapAsync(parallelism) { el =>
        akka.pattern.ask(ref).?(el)(timeout)
      }
      .map {
        case e: S => e
        case o =>
          throw new ClassCastException(
            s"'Flow.ask' failed: expected response of type [${tag.runtimeClass}], got [${o.getClass}]")
      }
      .mapError {
        // the purpose of this recovery is to change the name of the stage in that exception
        // we do so in order to help users find which stage caused the failure -- "the ask stage"
        case ex: WatchedActorTerminatedException =>
          throw new WatchedActorTerminatedException("ask()", ex.ref)
      }
      .named("ask")

    via(askFlow)
  }

  /**
   * The operator fails with an [[akka.stream.WatchedActorTerminatedException]] if the target actor is terminated.
   *
   * '''Emits when''' upstream emits
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Fails when''' the watched actor terminates
   *
   * '''Cancels when''' downstream cancels
   */
  def watch(ref: ActorRef): Repr[Out] =
    via(Watch(ref))

  /**
   * Only pass on those elements that satisfy the given predicate.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the given predicate returns true for the element
   *
   * '''Backpressures when''' the given predicate returns true for the element and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def filter(p: Out => Boolean): Repr[Out] = via(Filter(p))

  /**
   * Only pass on those elements that NOT satisfy the given predicate.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the given predicate returns false for the element
   *
   * '''Backpressures when''' the given predicate returns false for the element and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def filterNot(p: Out => Boolean): Repr[Out] =
    via(Flow[Out].filter(!p(_)).withAttributes(DefaultAttributes.filterNot))

  /**
   * Terminate processing (and cancel the upstream publisher) after predicate
   * returns false for the first time,
   * Due to input buffering some elements may have been requested from upstream publishers
   * that will then not be processed downstream of this step.
   *
   * The stream will be completed without producing any elements if predicate is false for
   * the first stream element.
   *
   * '''Emits when''' the predicate is true
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' predicate returned false (or 1 after predicate returns false if `inclusive` or upstream completes
   *
   * '''Cancels when''' predicate returned false or downstream cancels
   *
   * See also [[FlowOps.limit]], [[FlowOps.limitWeighted]]
   */
  def takeWhile(p: Out => Boolean): Repr[Out] = takeWhile(p, false)

  /**
   * Terminate processing (and cancel the upstream publisher) after predicate
   * returns false for the first time, including the first failed element iff inclusive is true
   * Due to input buffering some elements may have been requested from upstream publishers
   * that will then not be processed downstream of this step.
   *
   * The stream will be completed without producing any elements if predicate is false for
   * the first stream element.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the predicate is true
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' predicate returned false (or 1 after predicate returns false if `inclusive` or upstream completes
   *
   * '''Cancels when''' predicate returned false or downstream cancels
   *
   * See also [[FlowOps.limit]], [[FlowOps.limitWeighted]]
   */
  def takeWhile(p: Out => Boolean, inclusive: Boolean): Repr[Out] = via(TakeWhile(p, inclusive))

  /**
   * Discard elements at the beginning of the stream while predicate is true.
   * All elements will be taken after predicate returns false first time.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' predicate returned false and for all following stream elements
   *
   * '''Backpressures when''' predicate returned false and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def dropWhile(p: Out => Boolean): Repr[Out] = via(DropWhile(p))

  /**
   * Transform this stream by applying the given partial function to each of the elements
   * on which the function is defined as they pass through this processing step.
   * Non-matching elements are filtered out.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the provided partial function is defined for the element
   *
   * '''Backpressures when''' the partial function is defined for the element and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def collect[T](pf: PartialFunction[Out, T]): Repr[T] = via(Collect(pf))

  /**
   * Transform this stream by testing the type of each of the elements
   * on which the element is an instance of the provided type as they pass through this processing step.
   *
   * Non-matching elements are filtered out.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the element is an instance of the provided type
   *
   * '''Backpressures when''' the element is an instance of the provided type and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def collectType[T](implicit tag: ClassTag[T]): Repr[T] =
    collect { case c if tag.runtimeClass.isInstance(c) => c.asInstanceOf[T] }

  /**
   * Chunk up this stream into groups of the given size, with the last group
   * possibly smaller than requested due to end-of-stream.
   *
   * `n` must be positive, otherwise IllegalArgumentException is thrown.
   *
   * '''Emits when''' the specified number of elements have been accumulated or upstream completed
   *
   * '''Backpressures when''' a group has been assembled and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def grouped(n: Int): Repr[immutable.Seq[Out]] = via(Grouped(n))

  /**
   * Ensure stream boundedness by limiting the number of elements from upstream.
   * If the number of incoming elements exceeds max, it will signal
   * upstream failure `StreamLimitException` downstream.
   *
   * Due to input buffering some elements may have been
   * requested from upstream publishers that will then not be processed downstream
   * of this step.
   *
   * '''Emits when''' upstream emits and the number of emitted elements has not reached max
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes and the number of emitted elements has not reached max
   *
   * '''Errors when''' the total number of incoming element exceeds max
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.take]], [[FlowOps.takeWithin]], [[FlowOps.takeWhile]]
   */
  def limit(max: Long): Repr[Out] = limitWeighted(max)(_ => 1)

  /**
   * Ensure stream boundedness by evaluating the cost of incoming elements
   * using a cost function. Exactly how many elements will be allowed to travel downstream depends on the
   * evaluated cost of each element. If the accumulated cost exceeds max, it will signal
   * upstream failure `StreamLimitException` downstream.
   *
   * Due to input buffering some elements may have been
   * requested from upstream publishers that will then not be processed downstream
   * of this step.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' upstream emits and the accumulated cost has not reached max
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes and the number of emitted elements has not reached max
   *
   * '''Errors when''' when the accumulated cost exceeds max
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.take]], [[FlowOps.takeWithin]], [[FlowOps.takeWhile]]
   */
  def limitWeighted[T](max: Long)(costFn: Out => Long): Repr[Out] = via(LimitWeighted(max, costFn))

  /**
   * Apply a sliding window over the stream and return the windows as groups of elements, with the last group
   * possibly smaller than requested due to end-of-stream.
   *
   * `n` must be positive, otherwise IllegalArgumentException is thrown.
   * `step` must be positive, otherwise IllegalArgumentException is thrown.
   *
   * '''Emits when''' enough elements have been collected within the window or upstream completed
   *
   * '''Backpressures when''' a window has been assembled and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def sliding(n: Int, step: Int = 1): Repr[immutable.Seq[Out]] = via(Sliding(n, step))

  /**
   * Similar to `fold` but is not a terminal operation,
   * emits its current value which starts at `zero` and then
   * applies the current and next value to the given function `f`,
   * emitting the next current value.
   *
   * If the function `f` throws an exception and the supervision decision is
   * [[akka.stream.Supervision.Restart]] current value starts at `zero` again
   * the stream will continue.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * Note that the `zero` value must be immutable.
   *
   * '''Emits when''' the function scanning the element returns a new element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.scanAsync]]
   */
  def scan[T](zero: T)(f: (T, Out) => T): Repr[T] = via(Scan(zero, f))

  /**
   * Similar to `scan` but with a asynchronous function,
   * emits its current value which starts at `zero` and then
   * applies the current and next value to the given function `f`,
   * emitting a `Future` that resolves to the next current value.
   *
   * If the function `f` throws an exception and the supervision decision is
   * [[akka.stream.Supervision.Restart]] current value starts at `zero` again
   * the stream will continue.
   *
   * If the function `f` throws an exception and the supervision decision is
   * [[akka.stream.Supervision.Resume]] current value starts at the previous
   * current value, or zero when it doesn't have one, and the stream will continue.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * Note that the `zero` value must be immutable.
   *
   * '''Emits when''' the future returned by f` completes
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes and the last future returned by `f` completes
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.scan]]
   */
  def scanAsync[T](zero: T)(f: (T, Out) => Future[T]): Repr[T] = via(ScanAsync(zero, f))

  /**
   * Similar to `scan` but only emits its result when the upstream completes,
   * after which it also completes. Applies the given function towards its current and next value,
   * yielding the next current value.
   *
   * If the function `f` throws an exception and the supervision decision is
   * [[akka.stream.Supervision.Restart]] current value starts at `zero` again
   * the stream will continue.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * Note that the `zero` value must be immutable.
   *
   * '''Emits when''' upstream completes
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.scan]]
   */
  def fold[T](zero: T)(f: (T, Out) => T): Repr[T] = via(Fold(zero, f))

  /**
   * Similar to `fold` but with an asynchronous function.
   * Applies the given function towards its current and next value,
   * yielding the next current value.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * If the function `f` returns a failure and the supervision decision is
   * [[akka.stream.Supervision.Restart]] current value starts at `zero` again
   * the stream will continue.
   *
   * Note that the `zero` value must be immutable.
   *
   * '''Emits when''' upstream completes
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.fold]]
   */
  def foldAsync[T](zero: T)(f: (T, Out) => Future[T]): Repr[T] = via(new FoldAsync(zero, f))

  /**
   * Similar to `fold` but uses first element as zero element.
   * Applies the given function towards its current and next value,
   * yielding the next current value.
   *
   * If the stream is empty (i.e. completes before signalling any elements),
   * the reduce operator will fail its downstream with a [[NoSuchElementException]],
   * which is semantically in-line with that Scala's standard library collections
   * do in such situations.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' upstream completes
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.fold]]
   */
  def reduce[T >: Out](f: (T, T) => T): Repr[T] = via(new Reduce[T](f))

  /**
   * Intersperses stream with provided element, similar to how [[scala.collection.immutable.List.mkString]]
   * injects a separator between a List's elements.
   *
   * Additionally can inject start and end marker elements to stream.
   *
   * Examples:
   *
   * {{{
   * val nums = Source(List(1,2,3)).map(_.toString)
   * nums.intersperse(",")            //   1 , 2 , 3
   * nums.intersperse("[", ",", "]")  // [ 1 , 2 , 3 ]
   * }}}
   *
   * In case you want to only prepend or only append an element (yet still use the `intercept` feature
   * to inject a separator between elements, you may want to use the following pattern instead of the 3-argument
   * version of intersperse (See [[Source.concat]] for semantics details):
   *
   * {{{
   * Source.single(">> ") ++ Source(List("1", "2", "3")).intersperse(",")
   * Source(List("1", "2", "3")).intersperse(",") ++ Source.single("END")
   * }}}
   *
   * '''Emits when''' upstream emits (or before with the `start` element if provided)
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def intersperse[T >: Out](start: T, inject: T, end: T): Repr[T] =
    via(Intersperse(Some(start), inject, Some(end)))

  /**
   * Intersperses stream with provided element, similar to how [[scala.collection.immutable.List.mkString]]
   * injects a separator between a List's elements.
   *
   * Additionally can inject start and end marker elements to stream.
   *
   * Examples:
   *
   * {{{
   * val nums = Source(List(1,2,3)).map(_.toString)
   * nums.intersperse(",")            //   1 , 2 , 3
   * nums.intersperse("[", ",", "]")  // [ 1 , 2 , 3 ]
   * }}}
   *
   * '''Emits when''' upstream emits (or before with the `start` element if provided)
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def intersperse[T >: Out](inject: T): Repr[T] =
    via(Intersperse(None, inject, None))

  /**
   * Chunk up this stream into groups of elements received within a time window,
   * or limited by the given number of elements, whatever happens first.
   * Empty groups will not be emitted if no elements are received from upstream.
   * The last group before end-of-stream will contain the buffered elements
   * since the previously emitted group.
   *
   * `n` must be positive, and `d` must be greater than 0 seconds, otherwise
   * IllegalArgumentException is thrown.
   *
   * '''Emits when''' the configured time elapses since the last group has been emitted or `n` elements is buffered
   *
   * '''Backpressures when''' downstream backpressures, and there are `n+1` buffered elements
   *
   * '''Completes when''' upstream completes (emits last group)
   *
   * '''Cancels when''' downstream completes
   */
  def groupedWithin(n: Int, d: FiniteDuration): Repr[immutable.Seq[Out]] =
    via(new GroupedWeightedWithin[Out](n, ConstantFun.oneLong, d).withAttributes(DefaultAttributes.groupedWithin))

  /**
   * Chunk up this stream into groups of elements received within a time window,
   * or limited by the weight of the elements, whatever happens first.
   * Empty groups will not be emitted if no elements are received from upstream.
   * The last group before end-of-stream will contain the buffered elements
   * since the previously emitted group.
   *
   * `maxWeight` must be positive, and `d` must be greater than 0 seconds, otherwise
   * IllegalArgumentException is thrown.
   *
   * '''Emits when''' the configured time elapses since the last group has been emitted or weight limit reached
   *
   * '''Backpressures when''' downstream backpressures, and buffered group (+ pending element) weighs more than `maxWeight`
   *
   * '''Completes when''' upstream completes (emits last group)
   *
   * '''Cancels when''' downstream completes
   */
  def groupedWeightedWithin(maxWeight: Long, d: FiniteDuration)(costFn: Out => Long): Repr[immutable.Seq[Out]] =
    via(new GroupedWeightedWithin[Out](maxWeight, costFn, d))

  /**
   * Shifts elements emission in time by a specified amount. It allows to store elements
   * in internal buffer while waiting for next element to be emitted. Depending on the defined
   * [[akka.stream.DelayOverflowStrategy]] it might drop elements or backpressure the upstream if
   * there is no space available in the buffer.
   *
   * Delay precision is 10ms to avoid unnecessary timer scheduling cycles
   *
   * Internal buffer has default capacity 16. You can set buffer size by calling `addAttributes(inputBuffer)`
   *
   * '''Emits when''' there is a pending element in the buffer and configured time for this element elapsed
   *  * EmitEarly - strategy do not wait to emit element if buffer is full
   *
   * '''Backpressures when''' depending on OverflowStrategy
   *  * Backpressure - backpressures when buffer is full
   *  * DropHead, DropTail, DropBuffer - never backpressures
   *  * Fail - fails the stream if buffer gets full
   *
   * '''Completes when''' upstream completes and buffered elements have been drained
   *
   * '''Cancels when''' downstream cancels
   *
   * @param of time to shift all messages
   * @param strategy Strategy that is used when incoming elements cannot fit inside the buffer
   */
  def delay(of: FiniteDuration, strategy: DelayOverflowStrategy = DelayOverflowStrategy.dropTail): Repr[Out] =
    via(new Delay[Out](of, strategy))

  /**
   * Discard the given number of elements at the beginning of the stream.
   * No elements will be dropped if `n` is zero or negative.
   *
   * '''Emits when''' the specified number of elements has been dropped already
   *
   * '''Backpressures when''' the specified number of elements has been dropped and downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def drop(n: Long): Repr[Out] =
    via(Drop[Out](n))

  /**
   * Discard the elements received within the given duration at beginning of the stream.
   *
   * '''Emits when''' the specified time elapsed and a new upstream element arrives
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def dropWithin(d: FiniteDuration): Repr[Out] =
    via(new DropWithin[Out](d))

  /**
   * Terminate processing (and cancel the upstream publisher) after the given
   * number of elements. Due to input buffering some elements may have been
   * requested from upstream publishers that will then not be processed downstream
   * of this step.
   *
   * The stream will be completed without producing any elements if `n` is zero
   * or negative.
   *
   * '''Emits when''' the specified number of elements to take has not yet been reached
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' the defined number of elements has been taken or upstream completes
   *
   * '''Cancels when''' the defined number of elements has been taken or downstream cancels
   *
   * See also [[FlowOps.limit]], [[FlowOps.limitWeighted]]
   */
  def take(n: Long): Repr[Out] =
    via(Take[Out](n))

  /**
   * Terminate processing (and cancel the upstream publisher) after the given
   * duration. Due to input buffering some elements may have been
   * requested from upstream publishers that will then not be processed downstream
   * of this step.
   *
   * Note that this can be combined with [[#take]] to limit the number of elements
   * within the duration.
   *
   * '''Emits when''' an upstream element arrives
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or timer fires
   *
   * '''Cancels when''' downstream cancels or timer fires
   */
  def takeWithin(d: FiniteDuration): Repr[Out] = via(new TakeWithin[Out](d))

  /**
   * Allows a faster upstream to progress independently of a slower subscriber by conflating elements into a summary
   * until the subscriber is ready to accept them. For example a conflate step might average incoming numbers if the
   * upstream publisher is faster.
   *
   * This version of conflate allows to derive a seed from the first element and change the aggregated type to be
   * different than the input type. See [[FlowOps.conflate]] for a simpler version that does not change types.
   *
   * This element only rolls up elements if the upstream is faster, but if the downstream is faster it will not
   * duplicate elements.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' downstream stops backpressuring and there is a conflated element available
   *
   * '''Backpressures when''' never
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * @param seed Provides the first state for a conflated value using the first unconsumed element as a start
   * @param aggregate Takes the currently aggregated value and the current pending element to produce a new aggregate
   *
   * See also [[FlowOps.conflate]], [[FlowOps.limit]], [[FlowOps.limitWeighted]] [[FlowOps.batch]] [[FlowOps.batchWeighted]]
   */
  def conflateWithSeed[S](seed: Out => S)(aggregate: (S, Out) => S): Repr[S] =
    via(Batch(1L, ConstantFun.zeroLong, seed, aggregate).withAttributes(DefaultAttributes.conflate))

  /**
   * Allows a faster upstream to progress independently of a slower subscriber by conflating elements into a summary
   * until the subscriber is ready to accept them. For example a conflate step might average incoming numbers if the
   * upstream publisher is faster.
   *
   * This version of conflate does not change the output type of the stream. See [[FlowOps.conflateWithSeed]] for a
   * more flexible version that can take a seed function and transform elements while rolling up.
   *
   * This element only rolls up elements if the upstream is faster, but if the downstream is faster it will not
   * duplicate elements.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' downstream stops backpressuring and there is a conflated element available
   *
   * '''Backpressures when''' never
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * @param aggregate Takes the currently aggregated value and the current pending element to produce a new aggregate
   *
   * See also [[FlowOps.conflate]], [[FlowOps.limit]], [[FlowOps.limitWeighted]] [[FlowOps.batch]] [[FlowOps.batchWeighted]]
   */
  def conflate[O2 >: Out](aggregate: (O2, O2) => O2): Repr[O2] =
    conflateWithSeed[O2](ConstantFun.scalaIdentityFunction)(aggregate)

  /**
   * Allows a faster upstream to progress independently of a slower subscriber by aggregating elements into batches
   * until the subscriber is ready to accept them. For example a batch step might store received elements in
   * an array up to the allowed max limit if the upstream publisher is faster.
   *
   * This only rolls up elements if the upstream is faster, but if the downstream is faster it will not
   * duplicate elements.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' downstream stops backpressuring and there is an aggregated element available
   *
   * '''Backpressures when''' there are `max` batched elements and 1 pending element and downstream backpressures
   *
   * '''Completes when''' upstream completes and there is no batched/pending element waiting
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.conflateWithSeed]], [[FlowOps.batchWeighted]]
   *
   * @param max maximum number of elements to batch before backpressuring upstream (must be positive non-zero)
   * @param seed Provides the first state for a batched value using the first unconsumed element as a start
   * @param aggregate Takes the currently batched value and the current pending element to produce a new aggregate
   */
  def batch[S](max: Long, seed: Out => S)(aggregate: (S, Out) => S): Repr[S] =
    via(Batch(max, ConstantFun.oneLong, seed, aggregate).withAttributes(DefaultAttributes.batch))

  /**
   * Allows a faster upstream to progress independently of a slower subscriber by aggregating elements into batches
   * until the subscriber is ready to accept them. For example a batch step might concatenate `ByteString`
   * elements up to the allowed max limit if the upstream publisher is faster.
   *
   * This element only rolls up elements if the upstream is faster, but if the downstream is faster it will not
   * duplicate elements.
   *
   * Batching will apply for all elements, even if a single element cost is greater than the total allowed limit.
   * In this case, previous batched elements will be emitted, then the "heavy" element will be emitted (after
   * being applied with the `seed` function) without batching further elements with it, and then the rest of the
   * incoming elements are batched.
   *
   * '''Emits when''' downstream stops backpressuring and there is a batched element available
   *
   * '''Backpressures when''' there are `max` weighted batched elements + 1 pending element and downstream backpressures
   *
   * '''Completes when''' upstream completes and there is no batched/pending element waiting
   *
   * '''Cancels when''' downstream cancels
   *
   * See also [[FlowOps.conflateWithSeed]], [[FlowOps.batch]]
   *
   * @param max maximum weight of elements to batch before backpressuring upstream (must be positive non-zero)
   * @param costFn a function to compute a single element weight
   * @param seed Provides the first state for a batched value using the first unconsumed element as a start
   * @param aggregate Takes the currently batched value and the current pending element to produce a new batch
   */
  def batchWeighted[S](max: Long, costFn: Out => Long, seed: Out => S)(aggregate: (S, Out) => S): Repr[S] =
    via(Batch(max, costFn, seed, aggregate).withAttributes(DefaultAttributes.batchWeighted))

  /**
   * Allows a faster downstream to progress independently of a slower upstream by extrapolating elements from an older
   * element until new element comes from the upstream. For example an expand step might repeat the last element for
   * the subscriber until it receives an update from upstream.
   *
   * This element will never "drop" upstream elements as all elements go through at least one extrapolation step.
   * This means that if the upstream is actually faster than the upstream it will be backpressured by the downstream
   * subscriber.
   *
   * Expand does not support [[akka.stream.Supervision.Restart]] and [[akka.stream.Supervision.Resume]].
   * Exceptions from the `seed` function will complete the stream with failure.
   *
   * '''Emits when''' downstream stops backpressuring
   *
   * '''Backpressures when''' downstream backpressures or iterator runs empty
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   * @param expander       Takes the current extrapolation state to produce an output element and the next extrapolation
   *                       state.
   * @see [[#extrapolate]] for a version that always preserves the original element and allows for an initial "startup"
   *                       element.
   */
  def expand[U](expander: Out => Iterator[U]): Repr[U] = via(new Expand(expander))

  /**
   * Allows a faster downstream to progress independent of a slower upstream.
   *
   * This is achieved by introducing "extrapolated" elements - based on those from upstream - whenever downstream
   * signals demand.
   *
   * Extrapolate does not support [[akka.stream.Supervision.Restart]] and [[akka.stream.Supervision.Resume]].
   * Exceptions from the `extrapolate` function will complete the stream with failure.
   *
   * '''Emits when''' downstream stops backpressuring, AND EITHER upstream emits OR initial element is present OR
   * `extrapolate` is non-empty and applicable
   *
   * '''Backpressures when''' downstream backpressures or current `extrapolate` runs empty
   *
   * '''Completes when''' upstream completes and current `extrapolate` runs empty
   *
   * '''Cancels when''' downstream cancels
   *
   * @param extrapolator takes the current upstream element and provides a sequence of "extrapolated" elements based
   *                    on the original, to be emitted in case downstream signals demand.
   * @param initial the initial element to be emitted, in case upstream is able to stall the entire stream.
   * @see [[#expand]]    for a version that can overwrite the original element.
   */
  def extrapolate[U >: Out](extrapolator: U => Iterator[U], initial: Option[U] = None): Repr[U] = {
    val expandArg = (u: U) => Iterator.single(u) ++ extrapolator(u)

    val expandStep = new Expand[U, U](expandArg)

    initial.map(e => prepend(Source.single(e)).via(expandStep)).getOrElse(via(expandStep))
  }

  /**
   * Adds a fixed size buffer in the flow that allows to store elements from a faster upstream until it becomes full.
   * Depending on the defined [[akka.stream.OverflowStrategy]] it might drop elements or backpressure the upstream if
   * there is no space available
   *
   * '''Emits when''' downstream stops backpressuring and there is a pending element in the buffer
   *
   * '''Backpressures when''' downstream backpressures or depending on OverflowStrategy:
   *  <ul>
   *    <li>Backpressure - backpressures when buffer is full</li>
   *    <li>DropHead, DropTail, DropBuffer - never backpressures</li>
   *    <li>Fail - fails the stream if buffer gets full</li>
   *  </ul>
   *
   * '''Completes when''' upstream completes and buffered elements have been drained
   *
   * '''Cancels when''' downstream cancels
   *
   * @param size The size of the buffer in element count
   * @param overflowStrategy Strategy that is used when incoming elements cannot fit inside the buffer
   */
  def buffer(size: Int, overflowStrategy: OverflowStrategy): Repr[Out] = via(fusing.Buffer(size, overflowStrategy))

  /**
   * Takes up to `n` elements from the stream (less than `n` only if the upstream completes before emitting `n` elements)
   * and returns a pair containing a strict sequence of the taken element
   * and a stream representing the remaining elements. If ''n'' is zero or negative, then this will return a pair
   * of an empty collection and a stream containing the whole upstream unchanged.
   *
   * In case of an upstream error, depending on the current state
   *  - the master stream signals the error if less than `n` elements has been seen, and therefore the substream
   *    has not yet been emitted
   *  - the tail substream signals the error after the prefix and tail has been emitted by the main stream
   *    (at that point the main stream has already completed)
   *
   * '''Emits when''' the configured number of prefix elements are available. Emits this prefix, and the rest
   * as a substream
   *
   * '''Backpressures when''' downstream backpressures or substream backpressures
   *
   * '''Completes when''' prefix elements have been consumed and substream has been consumed
   *
   * '''Cancels when''' downstream cancels or substream cancels
   */
  def prefixAndTail[U >: Out](n: Int): Repr[(immutable.Seq[Out], Source[U, NotUsed])] =
    via(new PrefixAndTail[Out](n))

  /**
   * This operation demultiplexes the incoming stream into separate output
   * streams, one for each element key. The key is computed for each element
   * using the given function. When a new key is encountered for the first time
   * a new substream is opened and subsequently fed with all elements belonging to
   * that key.
   *
   * WARNING: If `allowClosedSubstreamRecreation` is set to `false` (default behavior) the operator
   * keeps track of all keys of streams that have already been closed. If you expect an infinite
   * number of keys this can cause memory issues. Elements belonging to those keys are drained
   * directly and not send to the substream.
   *
   * Note: If `allowClosedSubstreamRecreation` is set to `true` substream completion and incoming
   * elements are subject to race-conditions. If elements arrive for a stream that is in the process
   * of closing these elements might get lost.
   *
   * The object returned from this method is not a normal [[Source]] or [[Flow]],
   * it is a [[SubFlow]]. This means that after this operator all transformations
   * are applied to all encountered substreams in the same fashion. Substream mode
   * is exited either by closing the substream (i.e. connecting it to a [[Sink]])
   * or by merging the substreams back together; see the `to` and `mergeBack` methods
   * on [[SubFlow]] for more information.
   *
   * It is important to note that the substreams also propagate back-pressure as
   * any other stream, which means that blocking one substream will block the `groupBy`
   * operator itself—and thereby all substreams—once all internal or
   * explicit buffers are filled.
   *
   * If the group by function `f` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Stop]] the stream and substreams will be completed
   * with failure.
   *
   * If the group by function `f` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Resume]] or [[akka.stream.Supervision.Restart]]
   * the element is dropped and the stream and substreams continue.
   *
   * Function `f`  MUST NOT return `null`. This will throw exception and trigger supervision decision mechanism.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' an element for which the grouping function returns a group that has not yet been created.
   * Emits the new group
   *
   * '''Backpressures when''' there is an element pending for a group whose substream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels and all substreams cancel
   *
   * @param maxSubstreams configures the maximum number of substreams (keys)
   *        that are supported; if more distinct keys are encountered then the stream fails
   * @param f computes the key for each element
   * @param allowClosedSubstreamRecreation enables recreation of already closed substreams if elements with their
   *        corresponding keys arrive after completion
   */
  def groupBy[K](
      maxSubstreams: Int,
      f: Out => K,
      allowClosedSubstreamRecreation: Boolean): SubFlow[Out, Mat, Repr, Closed] = {
    val merge = new SubFlowImpl.MergeBack[Out, Repr] {
      override def apply[T](flow: Flow[Out, T, NotUsed], breadth: Int): Repr[T] =
        via(new GroupBy(maxSubstreams, f, allowClosedSubstreamRecreation))
          .map(_.via(flow))
          .via(new FlattenMerge(breadth))
    }
    val finish: (Sink[Out, NotUsed]) => Closed = s =>
      via(new GroupBy(maxSubstreams, f, allowClosedSubstreamRecreation))
        .to(Sink.foreach(_.runWith(s)(GraphInterpreter.currentInterpreter.materializer)))
    new SubFlowImpl(Flow[Out], merge, finish)
  }

  /**
   * This operation demultiplexes the incoming stream into separate output
   * streams, one for each element key. The key is computed for each element
   * using the given function. When a new key is encountered for the first time
   * a new substream is opened and subsequently fed with all elements belonging to
   * that key.
   *
   * WARNING: The operator keeps track of all keys of streams that have already been closed.
   * If you expect an infinite number of keys this can cause memory issues. Elements belonging
   * to those keys are drained directly and not send to the substream.
   *
   * @see [[#groupBy]]
   */
  def groupBy[K](maxSubstreams: Int, f: Out => K): SubFlow[Out, Mat, Repr, Closed] = groupBy(maxSubstreams, f, false)

  /**
   * This operation applies the given predicate to all incoming elements and
   * emits them to a stream of output streams, always beginning a new one with
   * the current element if the given predicate returns true for it. This means
   * that for the following series of predicate values, three substreams will
   * be produced with lengths 1, 2, and 3:
   *
   * {{{
   * false,             // element goes into first substream
   * true, false,       // elements go into second substream
   * true, false, false // elements go into third substream
   * }}}
   *
   * In case the *first* element of the stream matches the predicate, the first
   * substream emitted by splitWhen will start from that element. For example:
   *
   * {{{
   * true, false, false // first substream starts from the split-by element
   * true, false        // subsequent substreams operate the same way
   * }}}
   *
   * The object returned from this method is not a normal [[Source]] or [[Flow]],
   * it is a [[SubFlow]]. This means that after this operator all transformations
   * are applied to all encountered substreams in the same fashion. Substream mode
   * is exited either by closing the substream (i.e. connecting it to a [[Sink]])
   * or by merging the substreams back together; see the `to` and `mergeBack` methods
   * on [[SubFlow]] for more information.
   *
   * It is important to note that the substreams also propagate back-pressure as
   * any other stream, which means that blocking one substream will block the `splitWhen`
   * operator itself—and thereby all substreams—once all internal or
   * explicit buffers are filled.
   *
   * If the split predicate `p` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Stop]] the stream and substreams will be completed
   * with failure.
   *
   * If the split predicate `p` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Resume]] or [[akka.stream.Supervision.Restart]]
   * the element is dropped and the stream and substreams continue.
   *
   * '''Emits when''' an element for which the provided predicate is true, opening and emitting
   * a new substream for subsequent element
   *
   * '''Backpressures when''' there is an element pending for the next substream, but the previous
   * is not fully consumed yet, or the substream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels and substreams cancel on `SubstreamCancelStrategy.drain`, downstream
   * cancels or any substream cancels on `SubstreamCancelStrategy.propagate`
   *
   * See also [[FlowOps.splitAfter]].
   */
  def splitWhen(substreamCancelStrategy: SubstreamCancelStrategy)(
      p: Out => Boolean): SubFlow[Out, Mat, Repr, Closed] = {
    val merge = new SubFlowImpl.MergeBack[Out, Repr] {
      override def apply[T](flow: Flow[Out, T, NotUsed], breadth: Int): Repr[T] =
        via(Split.when(p, substreamCancelStrategy)).map(_.via(flow)).via(new FlattenMerge(breadth))
    }

    val finish: (Sink[Out, NotUsed]) => Closed = s =>
      via(Split.when(p, substreamCancelStrategy))
        .to(Sink.foreach(_.runWith(s)(GraphInterpreter.currentInterpreter.materializer)))

    new SubFlowImpl(Flow[Out], merge, finish)
  }

  /**
   * This operation applies the given predicate to all incoming elements and
   * emits them to a stream of output streams, always beginning a new one with
   * the current element if the given predicate returns true for it.
   *
   * @see [[#splitWhen]]
   */
  def splitWhen(p: Out => Boolean): SubFlow[Out, Mat, Repr, Closed] =
    splitWhen(SubstreamCancelStrategy.drain)(p)

  /**
   * This operation applies the given predicate to all incoming elements and
   * emits them to a stream of output streams. It *ends* the current substream when the
   * predicate is true. This means that for the following series of predicate values,
   * three substreams will be produced with lengths 2, 2, and 3:
   *
   * {{{
   * false, true,        // elements go into first substream
   * false, true,        // elements go into second substream
   * false, false, true  // elements go into third substream
   * }}}
   *
   * The object returned from this method is not a normal [[Source]] or [[Flow]],
   * it is a [[SubFlow]]. This means that after this operator all transformations
   * are applied to all encountered substreams in the same fashion. Substream mode
   * is exited either by closing the substream (i.e. connecting it to a [[Sink]])
   * or by merging the substreams back together; see the `to` and `mergeBack` methods
   * on [[SubFlow]] for more information.
   *
   * It is important to note that the substreams also propagate back-pressure as
   * any other stream, which means that blocking one substream will block the `splitAfter`
   * operator itself—and thereby all substreams—once all internal or
   * explicit buffers are filled.
   *
   * If the split predicate `p` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Stop]] the stream and substreams will be completed
   * with failure.
   *
   * If the split predicate `p` throws an exception and the supervision decision
   * is [[akka.stream.Supervision.Resume]] or [[akka.stream.Supervision.Restart]]
   * the element is dropped and the stream and substreams continue.
   *
   * '''Emits when''' an element passes through. When the provided predicate is true it emits the element
   * and opens a new substream for subsequent element
   *
   * '''Backpressures when''' there is an element pending for the next substream, but the previous
   * is not fully consumed yet, or the substream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels and substreams cancel on `SubstreamCancelStrategy.drain`, downstream
   * cancels or any substream cancels on `SubstreamCancelStrategy.propagate`
   *
   * See also [[FlowOps.splitWhen]].
   */
  def splitAfter(substreamCancelStrategy: SubstreamCancelStrategy)(
      p: Out => Boolean): SubFlow[Out, Mat, Repr, Closed] = {
    val merge = new SubFlowImpl.MergeBack[Out, Repr] {
      override def apply[T](flow: Flow[Out, T, NotUsed], breadth: Int): Repr[T] =
        via(Split.after(p, substreamCancelStrategy)).map(_.via(flow)).via(new FlattenMerge(breadth))
    }
    val finish: (Sink[Out, NotUsed]) => Closed = s =>
      via(Split.after(p, substreamCancelStrategy))
        .to(Sink.foreach(_.runWith(s)(GraphInterpreter.currentInterpreter.materializer)))
    new SubFlowImpl(Flow[Out], merge, finish)
  }

  /**
   * This operation applies the given predicate to all incoming elements and
   * emits them to a stream of output streams. It *ends* the current substream when the
   * predicate is true.
   *
   * @see [[#splitAfter]]
   */
  def splitAfter(p: Out => Boolean): SubFlow[Out, Mat, Repr, Closed] =
    splitAfter(SubstreamCancelStrategy.drain)(p)

  /**
   * Transform each input element into a `Source` of output elements that is
   * then flattened into the output stream by concatenation,
   * fully consuming one Source after the other.
   *
   * '''Emits when''' a currently consumed substream has an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes and all consumed substreams complete
   *
   * '''Cancels when''' downstream cancels
   */
  def flatMapConcat[T, M](f: Out => Graph[SourceShape[T], M]): Repr[T] = map(f).via(new FlattenMerge[T, M](1))

  /**
   * Transform each input element into a `Source` of output elements that is
   * then flattened into the output stream by merging, where at most `breadth`
   * substreams are being consumed at any given time.
   *
   * '''Emits when''' a currently consumed substream has an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes and all consumed substreams complete
   *
   * '''Cancels when''' downstream cancels
   */
  def flatMapMerge[T, M](breadth: Int, f: Out => Graph[SourceShape[T], M]): Repr[T] =
    map(f).via(new FlattenMerge[T, M](breadth))

  /**
   * If the first element has not passed through this operator before the provided timeout, the stream is failed
   * with a [[scala.concurrent.TimeoutException]].
   *
   * '''Emits when''' upstream emits an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or fails if timeout elapses before first element arrives
   *
   * '''Cancels when''' downstream cancels
   */
  def initialTimeout(timeout: FiniteDuration): Repr[Out] = via(new Timers.Initial[Out](timeout))

  /**
   * If the completion of the stream does not happen until the provided timeout, the stream is failed
   * with a [[scala.concurrent.TimeoutException]].
   *
   * '''Emits when''' upstream emits an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or fails if timeout elapses before upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def completionTimeout(timeout: FiniteDuration): Repr[Out] = via(new Timers.Completion[Out](timeout))

  /**
   * If the time between two processed elements exceeds the provided timeout, the stream is failed
   * with a [[scala.concurrent.TimeoutException]]. The timeout is checked periodically,
   * so the resolution of the check is one period (equals to timeout value).
   *
   * '''Emits when''' upstream emits an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or fails if timeout elapses between two emitted elements
   *
   * '''Cancels when''' downstream cancels
   */
  def idleTimeout(timeout: FiniteDuration): Repr[Out] = via(new Timers.Idle[Out](timeout))

  /**
   * If the time between the emission of an element and the following downstream demand exceeds the provided timeout,
   * the stream is failed with a [[scala.concurrent.TimeoutException]]. The timeout is checked periodically,
   * so the resolution of the check is one period (equals to timeout value).
   *
   * '''Emits when''' upstream emits an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes or fails if timeout elapses between element emission and downstream demand.
   *
   * '''Cancels when''' downstream cancels
   */
  def backpressureTimeout(timeout: FiniteDuration): Repr[Out] = via(new Timers.BackpressureTimeout[Out](timeout))

  /**
   * Injects additional elements if upstream does not emit for a configured amount of time. In other words, this
   * operator attempts to maintains a base rate of emitted elements towards the downstream.
   *
   * If the downstream backpressures then no element is injected until downstream demand arrives. Injected elements
   * do not accumulate during this period.
   *
   * Upstream elements are always preferred over injected elements.
   *
   * '''Emits when''' upstream emits an element or if the upstream was idle for the configured period
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def keepAlive[U >: Out](maxIdle: FiniteDuration, injectedElem: () => U): Repr[U] =
    via(new Timers.IdleInject[Out, U](maxIdle, injectedElem))

  /**
   * Sends elements downstream with speed limited to `elements/per`. In other words, this operator set the maximum rate
   * for emitting messages. This operator works for streams where all elements have the same cost or length.
   *
   * Throttle implements the token bucket model. There is a bucket with a given token capacity (burst size).
   * Tokens drops into the bucket at a given rate and can be `spared` for later use up to bucket capacity
   * to allow some burstiness. Whenever stream wants to send an element, it takes as many
   * tokens from the bucket as element costs. If there isn't any, throttle waits until the
   * bucket accumulates enough tokens. Elements that costs more than the allowed burst will be delayed proportionally
   * to their cost minus available tokens, meeting the target rate. Bucket is full when stream just materialized and
   * started.
   *
   * The burst size is calculated based on the given rate (`cost/per`) as 0.1 * rate, for example:
   * - rate < 20/second => burst size 1
   * - rate 20/second => burst size 2
   * - rate 100/second => burst size 10
   * - rate 200/second => burst size 20
   *
   * The throttle `mode` is [[akka.stream.ThrottleMode.Shaping]], which makes pauses before emitting messages to
   * meet throttle rate.
   *
   * '''Emits when''' upstream emits an element and configured time per each element elapsed
   *
   * '''Backpressures when''' downstream backpressures or the incoming rate is higher than the speed limit
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def throttle(elements: Int, per: FiniteDuration): Repr[Out] =
    throttle(elements, per, maximumBurst = Throttle.AutomaticMaximumBurst, ConstantFun.oneInt, ThrottleMode.Shaping)

  /**
   * Sends elements downstream with speed limited to `elements/per`. In other words, this operator set the maximum rate
   * for emitting messages. This operator works for streams where all elements have the same cost or length.
   *
   * Throttle implements the token bucket model. There is a bucket with a given token capacity (burst size or maximumBurst).
   * Tokens drops into the bucket at a given rate and can be `spared` for later use up to bucket capacity
   * to allow some burstiness. Whenever stream wants to send an element, it takes as many
   * tokens from the bucket as element costs. If there isn't any, throttle waits until the
   * bucket accumulates enough tokens. Elements that costs more than the allowed burst will be delayed proportionally
   * to their cost minus available tokens, meeting the target rate. Bucket is full when stream just materialized and started.
   *
   * Parameter `mode` manages behavior when upstream is faster than throttle rate:
   *  - [[akka.stream.ThrottleMode.Shaping]] makes pauses before emitting messages to meet throttle rate
   *  - [[akka.stream.ThrottleMode.Enforcing]] fails with exception when upstream is faster than throttle rate. Enforcing
   *  cannot emit elements that cost more than the maximumBurst
   *
   * It is recommended to use non-zero burst sizes as they improve both performance and throttling precision by allowing
   * the implementation to avoid using the scheduler when input rates fall below the enforced limit and to reduce
   * most of the inaccuracy caused by the scheduler resolution (which is in the range of milliseconds).
   *
   *  WARNING: Be aware that throttle is using scheduler to slow down the stream. This scheduler has minimal time of triggering
   *  next push. Consequently it will slow down the stream as it has minimal pause for emitting. This can happen in
   *  case burst is 0 and speed is higher than 30 events per second. You need to increase the `maximumBurst`  if
   *  elements arrive with small interval (30 milliseconds or less). Use the overloaded `throttle` method without
   *  `maximumBurst` parameter to automatically calculate the `maximumBurst` based on the given rate (`cost/per`).
   *  In other words the throttler always enforces the rate limit when `maximumBurst` parameter is given, but in
   *  certain cases (mostly due to limited scheduler resolution) it enforces a tighter bound than what was prescribed.
   *
   * '''Emits when''' upstream emits an element and configured time per each element elapsed
   *
   * '''Backpressures when''' downstream backpressures or the incoming rate is higher than the speed limit
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def throttle(elements: Int, per: FiniteDuration, maximumBurst: Int, mode: ThrottleMode): Repr[Out] =
    throttle(elements, per, maximumBurst, ConstantFun.oneInt, mode)

  /**
   * Sends elements downstream with speed limited to `cost/per`. Cost is
   * calculating for each element individually by calling `calculateCost` function.
   * This operator works for streams when elements have different cost(length).
   * Streams of `ByteString` for example.
   *
   * Throttle implements the token bucket model. There is a bucket with a given token capacity (burst size).
   * Tokens drops into the bucket at a given rate and can be `spared` for later use up to bucket capacity
   * to allow some burstiness. Whenever stream wants to send an element, it takes as many
   * tokens from the bucket as element costs. If there isn't any, throttle waits until the
   * bucket accumulates enough tokens. Elements that costs more than the allowed burst will be delayed proportionally
   * to their cost minus available tokens, meeting the target rate. Bucket is full when stream just materialized and
   * started.
   *
   * The burst size is calculated based on the given rate (`cost/per`) as 0.1 * rate, for example:
   * - rate < 20/second => burst size 1
   * - rate 20/second => burst size 2
   * - rate 100/second => burst size 10
   * - rate 200/second => burst size 20
   *
   * The throttle `mode` is [[akka.stream.ThrottleMode.Shaping]], which makes pauses before emitting messages to
   * meet throttle rate.
   *
   * '''Emits when''' upstream emits an element and configured time per each element elapsed
   *
   * '''Backpressures when''' downstream backpressures or the incoming rate is higher than the speed limit
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def throttle(cost: Int, per: FiniteDuration, costCalculation: (Out) => Int): Repr[Out] =
    via(new Throttle(cost, per, Throttle.AutomaticMaximumBurst, costCalculation, ThrottleMode.Shaping))

  /**
   * Sends elements downstream with speed limited to `cost/per`. Cost is
   * calculating for each element individually by calling `calculateCost` function.
   * This operator works for streams when elements have different cost(length).
   * Streams of `ByteString` for example.
   *
   * Throttle implements the token bucket model. There is a bucket with a given token capacity (burst size or maximumBurst).
   * Tokens drops into the bucket at a given rate and can be `spared` for later use up to bucket capacity
   * to allow some burstiness. Whenever stream wants to send an element, it takes as many
   * tokens from the bucket as element costs. If there isn't any, throttle waits until the
   * bucket accumulates enough tokens. Elements that costs more than the allowed burst will be delayed proportionally
   * to their cost minus available tokens, meeting the target rate. Bucket is full when stream just materialized and started.
   *
   * Parameter `mode` manages behavior when upstream is faster than throttle rate:
   *  - [[akka.stream.ThrottleMode.Shaping]] makes pauses before emitting messages to meet throttle rate
   *  - [[akka.stream.ThrottleMode.Enforcing]] fails with exception when upstream is faster than throttle rate. Enforcing
   *  cannot emit elements that cost more than the maximumBurst
   *
   * It is recommended to use non-zero burst sizes as they improve both performance and throttling precision by allowing
   * the implementation to avoid using the scheduler when input rates fall below the enforced limit and to reduce
   * most of the inaccuracy caused by the scheduler resolution (which is in the range of milliseconds).
   *
   *  WARNING: Be aware that throttle is using scheduler to slow down the stream. This scheduler has minimal time of triggering
   *  next push. Consequently it will slow down the stream as it has minimal pause for emitting. This can happen in
   *  case burst is 0 and speed is higher than 30 events per second. You need to increase the `maximumBurst`  if
   *  elements arrive with small interval (30 milliseconds or less). Use the overloaded `throttle` method without
   *  `maximumBurst` parameter to automatically calculate the `maximumBurst` based on the given rate (`cost/per`).
   *  In other words the throttler always enforces the rate limit when `maximumBurst` parameter is given, but in
   *  certain cases (mostly due to limited scheduler resolution) it enforces a tighter bound than what was prescribed.
   *
   * '''Emits when''' upstream emits an element and configured time per each element elapsed
   *
   * '''Backpressures when''' downstream backpressures or the incoming rate is higher than the speed limit
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   *
   */
  def throttle(
      cost: Int,
      per: FiniteDuration,
      maximumBurst: Int,
      costCalculation: (Out) => Int,
      mode: ThrottleMode): Repr[Out] =
    via(new Throttle(cost, per, maximumBurst, costCalculation, mode))

  /**
   * This is a simplified version of throttle that spreads events evenly across the given time interval. throttleEven using
   * best effort approach to meet throttle rate.
   *
   * Use this operator when you need just slow down a stream without worrying about exact amount
   * of time between events.
   *
   * If you want to be sure that no time interval has no more than specified number of events you need to use
   * [[throttle]] with maximumBurst attribute.
   * @see [[throttle]]
   */
  @Deprecated
  @deprecated("Use throttle without `maximumBurst` parameter instead.", "2.5.12")
  def throttleEven(elements: Int, per: FiniteDuration, mode: ThrottleMode): Repr[Out] =
    throttle(elements, per, Throttle.AutomaticMaximumBurst, ConstantFun.oneInt, mode)

  /**
   * This is a simplified version of throttle that spreads events evenly across the given time interval.
   *
   * Use this operator when you need just slow down a stream without worrying about exact amount
   * of time between events.
   *
   * If you want to be sure that no time interval has no more than specified number of events you need to use
   * [[throttle]] with maximumBurst attribute.
   * @see [[throttle]]
   */
  @Deprecated
  @deprecated("Use throttle without `maximumBurst` parameter instead.", "2.5.12")
  def throttleEven(cost: Int, per: FiniteDuration, costCalculation: (Out) => Int, mode: ThrottleMode): Repr[Out] =
    throttle(cost, per, Throttle.AutomaticMaximumBurst, costCalculation, mode)

  /**
   * Detaches upstream demand from downstream demand without detaching the
   * stream rates; in other words acts like a buffer of size 1.
   *
   * '''Emits when''' upstream emits an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def detach: Repr[Out] = via(GraphStages.detacher)

  /**
   * Delays the initial element by the specified duration.
   *
   * '''Emits when''' upstream emits an element if the initial delay is already elapsed
   *
   * '''Backpressures when''' downstream backpressures or initial delay is not yet elapsed
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def initialDelay(delay: FiniteDuration): Repr[Out] = via(new Timers.DelayInitial[Out](delay))

  /**
   * Logs elements flowing through the stream as well as completion and erroring.
   *
   * By default element and completion signals are logged on debug level, and errors are logged on Error level.
   * This can be adjusted according to your needs by providing a custom [[Attributes.LogLevels]] attribute on the given Flow:
   *
   * Uses implicit [[LoggingAdapter]] if available, otherwise uses an internally created one,
   * which uses `akka.stream.Log` as it's source (use this class to configure slf4j loggers).
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * '''Emits when''' the mapping function returns an element
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def log(name: String, extract: Out => Any = ConstantFun.scalaIdentityFunction)(
      implicit log: LoggingAdapter = null): Repr[Out] =
    via(Log(name, extract.asInstanceOf[Any => Any], Option(log)))

  /**
   * Combine the elements of current flow and the given [[Source]] into a stream of tuples.
   *
   * '''Emits when''' all of the inputs have an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' any upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zip[U](that: Graph[SourceShape[U], _]): Repr[(Out, U)] = via(zipGraph(that))

  /**
   * Combine the elements of current flow and the given [[Source]] into a stream of tuples.
   *
   * '''Emits when''' at first emits when both inputs emit, and then as long as any input emits (coupled to the default value of the completed input).
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' all upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zipAll[U, A >: Out](that: Graph[SourceShape[U], _], thisElem: A, thatElem: U): Repr[(A, U)] = {
    via(zipAllFlow(that, thisElem, thatElem))
  }

  protected def zipAllFlow[U, A >: Out, Mat2](
      that: Graph[SourceShape[U], Mat2],
      thisElem: A,
      thatElem: U): Flow[Out @uncheckedVariance, (A, U), Mat2] = {
    case object passedEnd
    val passedEndSrc = Source.repeat(passedEnd)
    val left: Flow[Out, Any, NotUsed] = Flow[A].concat(passedEndSrc)
    val right: Source[Any, Mat2] = Source.fromGraph(that).concat(passedEndSrc)
    val zipFlow: Flow[Out, (A, U), Mat2] = left
      .zipMat(right)(Keep.right)
      .takeWhile {
        case (`passedEnd`, `passedEnd`) => false
        case _                          => true
      }
      .map {
        case (`passedEnd`, r: U @unchecked) => (thisElem, r)
        case (l: A @unchecked, `passedEnd`) => (l, thatElem)
        case t: (A, U) @unchecked           => t
      }
    zipFlow
  }

  protected def zipGraph[U, M](that: Graph[SourceShape[U], M]): Graph[FlowShape[Out @uncheckedVariance, (Out, U)], M] =
    GraphDSL.create(that) { implicit b => r =>
      val zip = b.add(Zip[Out, U]())
      r ~> zip.in1
      FlowShape(zip.in0, zip.out)
    }

  /**
   * Combine the elements of 2 streams into a stream of tuples, picking always the latest element of each.
   *
   * A `ZipLatest` has a `left` and a `right` input port and one `out` port.
   *
   * No element is emitted until at least one element from each Source becomes available.
   *
   * '''Emits when''' all of the inputs have at least an element available, and then each time an element becomes
   *  available on either of the inputs
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' any upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zipLatest[U](that: Graph[SourceShape[U], _]): Repr[(Out, U)] = via(zipLatestGraph(that))

  protected def zipLatestGraph[U, M](
      that: Graph[SourceShape[U], M]): Graph[FlowShape[Out @uncheckedVariance, (Out, U)], M] =
    GraphDSL.create(that) { implicit b => r =>
      val zip = b.add(ZipLatest[Out, U]())
      r ~> zip.in1
      FlowShape(zip.in0, zip.out)
    }

  /**
   * Put together the elements of current flow and the given [[Source]]
   * into a stream of combined elements using a combiner function.
   *
   * '''Emits when''' all of the inputs have an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' any upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zipWith[Out2, Out3](that: Graph[SourceShape[Out2], _])(combine: (Out, Out2) => Out3): Repr[Out3] =
    via(zipWithGraph(that)(combine))

  protected def zipWithGraph[Out2, Out3, M](that: Graph[SourceShape[Out2], M])(
      combine: (Out, Out2) => Out3): Graph[FlowShape[Out @uncheckedVariance, Out3], M] =
    GraphDSL.create(that) { implicit b => r =>
      val zip = b.add(ZipWith[Out, Out2, Out3](combine))
      r ~> zip.in1
      FlowShape(zip.in0, zip.out)
    }

  /**
   * Combine the elements of multiple streams into a stream of combined elements using a combiner function,
   * picking always the latest of the elements of each source.
   *
   * No element is emitted until at least one element from each Source becomes available. Whenever a new
   * element appears, the zipping function is invoked with a tuple containing the new element
   * and the other last seen elements.
   *
   *   '''Emits when''' all of the inputs have at least an element available, and then each time an element becomes
   *   available on either of the inputs
   *
   *   '''Backpressures when''' downstream backpressures
   *
   *   '''Completes when''' any of the upstreams completes
   *
   *   '''Cancels when''' downstream cancels
   */
  def zipLatestWith[Out2, Out3](that: Graph[SourceShape[Out2], _])(combine: (Out, Out2) => Out3): Repr[Out3] =
    via(zipLatestWithGraph(that)(combine))

  protected def zipLatestWithGraph[Out2, Out3, M](that: Graph[SourceShape[Out2], M])(
      combine: (Out, Out2) => Out3): Graph[FlowShape[Out @uncheckedVariance, Out3], M] =
    GraphDSL.create(that) { implicit b => r =>
      val zip = b.add(ZipLatestWith[Out, Out2, Out3](combine))
      r ~> zip.in1
      FlowShape(zip.in0, zip.out)
    }

  /**
   * Combine the elements of current flow into a stream of tuples consisting
   * of all elements paired with their index. Indices start at 0.
   *
   * '''Emits when''' upstream emits an element and is paired with their index
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zipWithIndex: Repr[(Out, Long)] = {
    statefulMapConcat[(Out, Long)] { () =>
      var index: Long = 0L
      elem => {
        val zipped = (elem, index)
        index += 1
        immutable.Iterable[(Out, Long)](zipped)
      }
    }
  }

  /**
   * Interleave is a deterministic merge of the given [[Source]] with elements of this [[Flow]].
   * It first emits `segmentSize` number of elements from this flow to downstream, then - same amount for `that`
   * source, then repeat process.
   *
   * Example:
   * {{{
   * Source(List(1, 2, 3)).interleave(List(4, 5, 6, 7), 2) // 1, 2, 4, 5, 3, 6, 7
   * }}}
   *
   * After one of upstreams is complete then all the rest elements will be emitted from the second one
   *
   * If it gets error from one of upstreams - stream completes with failure.
   *
   * '''Emits when''' element is available from the currently consumed upstream
   *
   * '''Backpressures when''' downstream backpressures. Signal to current
   * upstream, switch to next upstream when received `segmentSize` elements
   *
   * '''Completes when''' the [[Flow]] and given [[Source]] completes
   *
   * '''Cancels when''' downstream cancels
   */
  def interleave[U >: Out](that: Graph[SourceShape[U], _], segmentSize: Int): Repr[U] =
    interleave(that, segmentSize, eagerClose = false)

  /**
   * Interleave is a deterministic merge of the given [[Source]] with elements of this [[Flow]].
   * It first emits `segmentSize` number of elements from this flow to downstream, then - same amount for `that`
   * source, then repeat process.
   *
   * If eagerClose is false and one of the upstreams complete the elements from the other upstream will continue passing
   * through the interleave operator. If eagerClose is true and one of the upstream complete interleave will cancel the
   * other upstream and complete itself.
   *
   * If it gets error from one of upstreams - stream completes with failure.
   *
   * '''Emits when''' element is available from the currently consumed upstream
   *
   * '''Backpressures when''' downstream backpressures. Signal to current
   * upstream, switch to next upstream when received `segmentSize` elements
   *
   * '''Completes when''' the [[Flow]] and given [[Source]] completes
   *
   * '''Cancels when''' downstream cancels
   */
  def interleave[U >: Out](that: Graph[SourceShape[U], _], segmentSize: Int, eagerClose: Boolean): Repr[U] =
    via(interleaveGraph(that, segmentSize, eagerClose))

  protected def interleaveGraph[U >: Out, M](
      that: Graph[SourceShape[U], M],
      segmentSize: Int,
      eagerClose: Boolean = false): Graph[FlowShape[Out @uncheckedVariance, U], M] =
    GraphDSL.create(that) { implicit b => r =>
      val interleave = b.add(Interleave[U](2, segmentSize, eagerClose))
      r ~> interleave.in(1)
      FlowShape(interleave.in(0), interleave.out)
    }

  /**
   * Merge the given [[Source]] to this [[Flow]], taking elements as they arrive from input streams,
   * picking randomly when several elements ready.
   *
   * '''Emits when''' one of the inputs has an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' all upstreams complete (eagerComplete=false) or one upstream completes (eagerComplete=true), default value is `false`
   *
   * '''Cancels when''' downstream cancels
   */
  def merge[U >: Out, M](that: Graph[SourceShape[U], M], eagerComplete: Boolean = false): Repr[U] =
    via(mergeGraph(that, eagerComplete))

  protected def mergeGraph[U >: Out, M](
      that: Graph[SourceShape[U], M],
      eagerComplete: Boolean): Graph[FlowShape[Out @uncheckedVariance, U], M] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(Merge[U](2, eagerComplete))
      r ~> merge.in(1)
      FlowShape(merge.in(0), merge.out)
    }

  /**
   * MergeLatest joins elements from N input streams into stream of lists of size N.
   * i-th element in list is the latest emitted element from i-th input stream.
   * MergeLatest emits list for each element emitted from some input stream,
   * but only after each input stream emitted at least one element.
   *
   * '''Emits when''' an element is available from some input and each input emits at least one element from stream start
   *
   * '''Completes when''' all upstreams complete (eagerClose=false) or one upstream completes (eagerClose=true)
   */
  def mergeLatest[U >: Out, M](that: Graph[SourceShape[U], M], eagerComplete: Boolean = false): Repr[immutable.Seq[U]] =
    via(mergeLatestGraph(that, eagerComplete))

  protected def mergeLatestGraph[U >: Out, M](
      that: Graph[SourceShape[U], M],
      eagerComplete: Boolean): Graph[FlowShape[Out @uncheckedVariance, immutable.Seq[U]], M] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(MergeLatest[U](2, eagerComplete))
      r ~> merge.in(1)
      FlowShape(merge.in(0), merge.out)
    }

  /**
   * Merge two sources. Prefer one source if both sources have elements ready.
   *
   * '''emits''' when one of the inputs has an element available. If multiple have elements available, prefer the 'right' one when 'preferred' is 'true', or the 'left' one when 'preferred' is 'false'.
   *
   * '''backpressures''' when downstream backpressures
   *
   * '''completes''' when all upstreams complete (This behavior is changeable to completing when any upstream completes by setting `eagerComplete=true`.)
   */
  def mergePreferred[U >: Out, M](
      that: Graph[SourceShape[U], M],
      priority: Boolean,
      eagerComplete: Boolean = false): Repr[U] =
    via(mergePreferredGraph(that, priority, eagerComplete))

  protected def mergePreferredGraph[U >: Out, M](
      that: Graph[SourceShape[U], M],
      priority: Boolean,
      eagerComplete: Boolean): Graph[FlowShape[Out @uncheckedVariance, U], M] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(MergePreferred[U](1, eagerComplete))
      r ~> merge.in(if (priority) 0 else 1)
      FlowShape(merge.in(if (priority) 1 else 0), merge.out)
    }

  /**
   * Merge two sources. Prefer the sources depending on the 'priority' parameters.
   *
   * '''emits''' when one of the inputs has an element available, preferring inputs based on the 'priority' parameters if both have elements available
   *
   * '''backpressures''' when downstream backpressures
   *
   * '''completes''' when both upstreams complete (This behavior is changeable to completing when any upstream completes by setting `eagerComplete=true`.)
   */
  def mergePrioritized[U >: Out, M](
      that: Graph[SourceShape[U], M],
      leftPriority: Int,
      rightPriority: Int,
      eagerComplete: Boolean = false): Repr[U] =
    via(mergePrioritizedGraph(that, leftPriority, rightPriority, eagerComplete))

  protected def mergePrioritizedGraph[U >: Out, M](
      that: Graph[SourceShape[U], M],
      leftPriority: Int,
      rightPriority: Int,
      eagerComplete: Boolean): Graph[FlowShape[Out @uncheckedVariance, U], M] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(MergePrioritized[U](Seq(leftPriority, rightPriority), eagerComplete))
      r ~> merge.in(1)
      FlowShape(merge.in(0), merge.out)
    }

  /**
   * Merge the given [[Source]] to this [[Flow]], taking elements as they arrive from input streams,
   * picking always the smallest of the available elements (waiting for one element from each side
   * to be available). This means that possible contiguity of the input streams is not exploited to avoid
   * waiting for elements, this merge will block when one of the inputs does not have more elements (and
   * does not complete).
   *
   * '''Emits when''' all of the inputs have an element available
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' all upstreams complete
   *
   * '''Cancels when''' downstream cancels
   */
  def mergeSorted[U >: Out, M](that: Graph[SourceShape[U], M])(implicit ord: Ordering[U]): Repr[U] =
    via(mergeSortedGraph(that))

  protected def mergeSortedGraph[U >: Out, M](that: Graph[SourceShape[U], M])(
      implicit ord: Ordering[U]): Graph[FlowShape[Out @uncheckedVariance, U], M] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(new MergeSorted[U])
      r ~> merge.in1
      FlowShape(merge.in0, merge.out)
    }

  /**
   * Concatenate the given [[Source]] to this [[Flow]], meaning that once this
   * Flow’s input is exhausted and all result elements have been generated,
   * the Source’s elements will be produced.
   *
   * Note that the [[Source]] is materialized together with this Flow and just kept
   * from producing elements by asserting back-pressure until its time comes.
   *
   * If this [[Flow]] gets upstream error - no elements from the given [[Source]] will be pulled.
   *
   * '''Emits when''' element is available from current stream or from the given [[Source]] when current is completed
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' given [[Source]] completes
   *
   * '''Cancels when''' downstream cancels
   */
  def concat[U >: Out, Mat2](that: Graph[SourceShape[U], Mat2]): Repr[U] =
    via(concatGraph(that))

  protected def concatGraph[U >: Out, Mat2](
      that: Graph[SourceShape[U], Mat2]): Graph[FlowShape[Out @uncheckedVariance, U], Mat2] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(Concat[U]())
      r ~> merge.in(1)
      FlowShape(merge.in(0), merge.out)
    }

  /**
   * Prepend the given [[Source]] to this [[Flow]], meaning that before elements
   * are generated from this Flow, the Source's elements will be produced until it
   * is exhausted, at which point Flow elements will start being produced.
   *
   * Note that this Flow will be materialized together with the [[Source]] and just kept
   * from producing elements by asserting back-pressure until its time comes.
   *
   * If the given [[Source]] gets upstream error - no elements from this [[Flow]] will be pulled.
   *
   * '''Emits when''' element is available from the given [[Source]] or from current stream when the [[Source]] is completed
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' this [[Flow]] completes
   *
   * '''Cancels when''' downstream cancels
   */
  def prepend[U >: Out, Mat2](that: Graph[SourceShape[U], Mat2]): Repr[U] =
    via(prependGraph(that))

  protected def prependGraph[U >: Out, Mat2](
      that: Graph[SourceShape[U], Mat2]): Graph[FlowShape[Out @uncheckedVariance, U], Mat2] =
    GraphDSL.create(that) { implicit b => r =>
      val merge = b.add(Concat[U]())
      r ~> merge.in(0)
      FlowShape(merge.in(1), merge.out)
    }

  /**
   * Provides a secondary source that will be consumed if this stream completes without any
   * elements passing by. As soon as the first element comes through this stream, the alternative
   * will be cancelled.
   *
   * Note that this Flow will be materialized together with the [[Source]] and just kept
   * from producing elements by asserting back-pressure until its time comes or it gets
   * cancelled.
   *
   * On errors the operator is failed regardless of source of the error.
   *
   * '''Emits when''' element is available from first stream or first stream closed without emitting any elements and an element
   *                  is available from the second stream
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' the primary stream completes after emitting at least one element, when the primary stream completes
   *                      without emitting and the secondary stream already has completed or when the secondary stream completes
   *
   * '''Cancels when''' downstream cancels and additionally the alternative is cancelled as soon as an element passes
   *                    by from this stream.
   */
  def orElse[U >: Out, Mat2](secondary: Graph[SourceShape[U], Mat2]): Repr[U] =
    via(orElseGraph(secondary))

  protected def orElseGraph[U >: Out, Mat2](
      secondary: Graph[SourceShape[U], Mat2]): Graph[FlowShape[Out @uncheckedVariance, U], Mat2] =
    GraphDSL.create(secondary) { implicit b => secondary =>
      val orElse = b.add(OrElse[U]())

      secondary ~> orElse.in(1)

      FlowShape(orElse.in(0), orElse.out)
    }

  /**
   * Concatenates this [[Flow]] with the given [[Source]] so the first element
   * emitted by that source is emitted after the last element of this
   * flow.
   *
   * This is a shorthand for [[concat]]
   */
  def ++[U >: Out, M](that: Graph[SourceShape[U], M]): Repr[U] = concat(that)

  /**
   * Connect this [[Flow]] to a [[Sink]], concatenating the processing steps of both.
   * {{{
   *     +------------------------------+
   *     | Resulting Sink[In, Mat]      |
   *     |                              |
   *     |  +------+          +------+  |
   *     |  |      |          |      |  |
   * In ~~> | flow | ~~Out~~> | sink |  |
   *     |  |   Mat|          |     M|  |
   *     |  +------+          +------+  |
   *     +------------------------------+
   * }}}
   *
   * The materialized value of the combined [[Sink]] will be the materialized
   * value of the current flow (ignoring the given Sink’s value), use
   * [[Flow#toMat[Mat2* toMat]] if a different strategy is needed.
   *
   * See also [[FlowOpsMat.toMat]] when access to materialized values of the parameter is needed.
   */
  def to[Mat2](sink: Graph[SinkShape[Out], Mat2]): Closed

  /**
   * Attaches the given [[Sink]] to this [[Flow]], meaning that elements that pass
   * through will also be sent to the [[Sink]].
   *
   * It is similar to [[#wireTap]] but will backpressure instead of dropping elements when the given [[Sink]] is not ready.
   *
   * '''Emits when''' element is available and demand exists both from the Sink and the downstream.
   *
   * '''Backpressures when''' downstream or Sink backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream or Sink cancels
   */
  def alsoTo(that: Graph[SinkShape[Out], _]): Repr[Out] = via(alsoToGraph(that))

  protected def alsoToGraph[M](that: Graph[SinkShape[Out], M]): Graph[FlowShape[Out @uncheckedVariance, Out], M] =
    GraphDSL.create(that) { implicit b => r =>
      import GraphDSL.Implicits._
      val bcast = b.add(Broadcast[Out](2, eagerCancel = true))
      bcast.out(1) ~> r
      FlowShape(bcast.in, bcast.out(0))
    }

  /**
   * Attaches the given [[Sink]] to this [[Flow]], meaning that elements will be sent to the [[Sink]]
   * instead of being passed through if the predicate `when` returns `true`.
   *
   * '''Emits when''' emits when an element is available from the input and the chosen output has demand
   *
   * '''Backpressures when''' the currently chosen output back-pressures
   *
   * '''Completes when''' upstream completes and no output is pending
   *
   * '''Cancels when''' any of the downstreams cancel
   */
  def divertTo(that: Graph[SinkShape[Out], _], when: Out => Boolean): Repr[Out] = via(divertToGraph(that, when))

  protected def divertToGraph[M](
      that: Graph[SinkShape[Out], M],
      when: Out => Boolean): Graph[FlowShape[Out @uncheckedVariance, Out], M] =
    GraphDSL.create(that) { implicit b => r =>
      import GraphDSL.Implicits._
      val partition = b.add(new Partition[Out](2, out => if (when(out)) 1 else 0, true))
      partition.out(1) ~> r
      FlowShape(partition.in, partition.out(0))
    }

  /**
   * Attaches the given [[Sink]] to this [[Flow]] as a wire tap, meaning that elements that pass
   * through will also be sent to the wire-tap Sink, without the latter affecting the mainline flow.
   * If the wire-tap Sink backpressures, elements that would've been sent to it will be dropped instead.
   *
   * It is similar to [[#alsoTo]] which does backpressure instead of dropping elements.
   *
   * '''Emits when''' element is available and demand exists from the downstream; the element will
   * also be sent to the wire-tap Sink if there is demand.
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def wireTap(that: Graph[SinkShape[Out], _]): Repr[Out] = via(wireTapGraph(that))

  protected def wireTapGraph[M](that: Graph[SinkShape[Out], M]): Graph[FlowShape[Out @uncheckedVariance, Out], M] =
    GraphDSL.create(that) { implicit b => r =>
      import GraphDSL.Implicits._
      val bcast = b.add(WireTap[Out]())
      bcast.out1 ~> r
      FlowShape(bcast.in, bcast.out0)
    }

  def withAttributes(attr: Attributes): Repr[Out]

  def addAttributes(attr: Attributes): Repr[Out]

  def named(name: String): Repr[Out]

  /**
   * Put an asynchronous boundary around this `Flow`.
   *
   * If this is a `SubFlow` (created e.g. by `groupBy`), this creates an
   * asynchronous boundary around each materialized sub-flow, not the
   * super-flow. That way, the super-flow will communicate with sub-flows
   * asynchronously.
   */
  def async: Repr[Out]
}

/**
 * INTERNAL API: this trait will be changed in binary-incompatible ways for classes that are derived from it!
 * Do not implement this interface outside the Akka code base!
 *
 * Binary compatibility is only maintained for callers of this trait’s interface.
 */
trait FlowOpsMat[+Out, +Mat] extends FlowOps[Out, Mat] {

  type Repr[+O] <: ReprMat[O, Mat] {
    type Repr[+OO] = FlowOpsMat.this.Repr[OO]
    type ReprMat[+OO, +MM] = FlowOpsMat.this.ReprMat[OO, MM]
    type Closed = FlowOpsMat.this.Closed
    type ClosedMat[+M] = FlowOpsMat.this.ClosedMat[M]
  }
  type ReprMat[+O, +M] <: FlowOpsMat[O, M] {
    type Repr[+OO] = FlowOpsMat.this.ReprMat[OO, M @uncheckedVariance]
    type ReprMat[+OO, +MM] = FlowOpsMat.this.ReprMat[OO, MM]
    type Closed = FlowOpsMat.this.ClosedMat[M @uncheckedVariance]
    type ClosedMat[+MM] = FlowOpsMat.this.ClosedMat[MM]
  }
  type ClosedMat[+M] <: Graph[_, M]

  /**
   * Transform this [[Flow]] by appending the given processing steps.
   * {{{
   *     +---------------------------------+
   *     | Resulting Flow[In, T, M2]       |
   *     |                                 |
   *     |  +------+            +------+   |
   *     |  |      |            |      |   |
   * In ~~> | this |  ~~Out~~>  | flow |  ~~> T
   *     |  |   Mat|            |     M|   |
   *     |  +------+            +------+   |
   *     +---------------------------------+
   * }}}
   * The `combine` function is used to compose the materialized values of this flow and that
   * flow into the materialized value of the resulting Flow.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def viaMat[T, Mat2, Mat3](flow: Graph[FlowShape[Out, T], Mat2])(combine: (Mat, Mat2) => Mat3): ReprMat[T, Mat3]

  /**
   * Connect this [[Flow]] to a [[Sink]], concatenating the processing steps of both.
   * {{{
   *     +----------------------------+
   *     | Resulting Sink             |
   *     |                            |
   *     |  +------+        +------+  |
   *     |  |      |        |      |  |
   * In ~~> | flow | ~Out~> | sink |  |
   *     |  |      |        |      |  |
   *     |  +------+        +------+  |
   *     +----------------------------+
   * }}}
   * The `combine` function is used to compose the materialized values of this flow and that
   * Sink into the materialized value of the resulting Sink.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def toMat[Mat2, Mat3](sink: Graph[SinkShape[Out], Mat2])(combine: (Mat, Mat2) => Mat3): ClosedMat[Mat3]

  /**
   * Combine the elements of current flow and the given [[Source]] into a stream of tuples.
   *
   * @see [[#zip]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def zipMat[U, Mat2, Mat3](that: Graph[SourceShape[U], Mat2])(matF: (Mat, Mat2) => Mat3): ReprMat[(Out, U), Mat3] =
    viaMat(zipGraph(that))(matF)

  /**
   * Combine the elements of current flow and the given [[Source]] into a stream of tuples.
   *
   * @see [[#zipAll]]
   *
   * '''Emits when''' at first emits when both inputs emit, and then as long as any input emits (coupled to the default value of the completed input).
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' all upstream completes
   *
   * '''Cancels when''' downstream cancels
   */
  def zipAllMat[U, Mat2, Mat3, A >: Out](that: Graph[SourceShape[U], Mat2], thisElem: A, thatElem: U)(
      matF: (Mat, Mat2) => Mat3): ReprMat[(A, U), Mat3] = {
    viaMat(zipAllFlow(that, thisElem, thatElem))(matF)
  }

  /**
   * Put together the elements of current flow and the given [[Source]]
   * into a stream of combined elements using a combiner function.
   *
   * @see [[#zipWith]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def zipWithMat[Out2, Out3, Mat2, Mat3](that: Graph[SourceShape[Out2], Mat2])(combine: (Out, Out2) => Out3)(
      matF: (Mat, Mat2) => Mat3): ReprMat[Out3, Mat3] =
    viaMat(zipWithGraph(that)(combine))(matF)

  /**
   * Combine the elements of current flow and the given [[Source]] into a stream of tuples,
   * picking always the latest of the elements of each source.
   *
   * @see [[#zipLatest]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def zipLatestMat[U, Mat2, Mat3](that: Graph[SourceShape[U], Mat2])(
      matF: (Mat, Mat2) => Mat3): ReprMat[(Out, U), Mat3] =
    viaMat(zipLatestGraph(that))(matF)

  /**
   * Put together the elements of current flow and the given [[Source]]
   * into a stream of combined elements using a combiner function, picking always the latest of the elements of each source.
   *
   * @see [[#zipLatestWith]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def zipLatestWithMat[Out2, Out3, Mat2, Mat3](that: Graph[SourceShape[Out2], Mat2])(combine: (Out, Out2) => Out3)(
      matF: (Mat, Mat2) => Mat3): ReprMat[Out3, Mat3] =
    viaMat(zipLatestWithGraph(that)(combine))(matF)

  /**
   * Merge the given [[Source]] to this [[Flow]], taking elements as they arrive from input streams,
   * picking randomly when several elements ready.
   *
   * @see [[#merge]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def mergeMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2], eagerComplete: Boolean = false)(
      matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(mergeGraph(that, eagerComplete))(matF)

  /**
   * Interleave is a deterministic merge of the given [[Source]] with elements of this [[Flow]].
   * It first emits `segmentSize` number of elements from this flow to downstream, then - same amount for `that` source,
   * then repeat process.
   *
   * After one of upstreams is complete then all the rest elements will be emitted from the second one
   *
   * If it gets error from one of upstreams - stream completes with failure.
   *
   * @see [[#interleave]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def interleaveMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2], request: Int)(
      matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    interleaveMat(that, request, eagerClose = false)(matF)

  /**
   * Interleave is a deterministic merge of the given [[Source]] with elements of this [[Flow]].
   * It first emits `segmentSize` number of elements from this flow to downstream, then - same amount for `that` source,
   * then repeat process.
   *
   * If eagerClose is false and one of the upstreams complete the elements from the other upstream will continue passing
   * through the interleave operator. If eagerClose is true and one of the upstream complete interleave will cancel the
   * other upstream and complete itself.
   *
   * If it gets error from one of upstreams - stream completes with failure.
   *
   * @see [[#interleave]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def interleaveMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2], request: Int, eagerClose: Boolean)(
      matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(interleaveGraph(that, request, eagerClose))(matF)

  /**
   * MergeLatest joins elements from N input streams into stream of lists of size N.
   * i-th element in list is the latest emitted element from i-th input stream.
   * MergeLatest emits list for each element emitted from some input stream,
   * but only after each input stream emitted at least one element.
   *
   * @see [[#mergeLatest]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def mergeLatestMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2], eagerClose: Boolean)(
      matF: (Mat, Mat2) => Mat3): ReprMat[immutable.Seq[U], Mat3] =
    viaMat(mergeLatestGraph(that, eagerClose))(matF)

  /**
   * Merge two sources. Prefer one source if both sources have elements ready.
   *
   * @see [[#mergePreferred]]
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def mergePreferredMat[U >: Out, Mat2, Mat3](
      that: Graph[SourceShape[U], Mat2],
      preferred: Boolean,
      eagerClose: Boolean)(matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(mergePreferredGraph(that, preferred, eagerClose))(matF)

  /**
   * Merge two sources. Prefer the sources depending on the 'priority' parameters.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def mergePrioritizedMat[U >: Out, Mat2, Mat3](
      that: Graph[SourceShape[U], Mat2],
      leftPriority: Int,
      rightPriority: Int,
      eagerClose: Boolean)(matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(mergePrioritizedGraph(that, leftPriority, rightPriority, eagerClose))(matF)

  /**
   * Merge the given [[Source]] to this [[Flow]], taking elements as they arrive from input streams,
   * picking always the smallest of the available elements (waiting for one element from each side
   * to be available). This means that possible contiguity of the input streams is not exploited to avoid
   * waiting for elements, this merge will block when one of the inputs does not have more elements (and
   * does not complete).
   *
   * @see [[#mergeSorted]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def mergeSortedMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2])(matF: (Mat, Mat2) => Mat3)(
      implicit ord: Ordering[U]): ReprMat[U, Mat3] =
    viaMat(mergeSortedGraph(that))(matF)

  /**
   * Concatenate the given [[Source]] to this [[Flow]], meaning that once this
   * Flow’s input is exhausted and all result elements have been generated,
   * the Source’s elements will be produced.
   *
   * Note that the [[Source]] is materialized together with this Flow and just kept
   * from producing elements by asserting back-pressure until its time comes.
   *
   * If this [[Flow]] gets upstream error - no elements from the given [[Source]] will be pulled.
   *
   * @see [[#concat]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def concatMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2])(matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(concatGraph(that))(matF)

  /**
   * Prepend the given [[Source]] to this [[Flow]], meaning that before elements
   * are generated from this Flow, the Source's elements will be produced until it
   * is exhausted, at which point Flow elements will start being produced.
   *
   * Note that this Flow will be materialized together with the [[Source]] and just kept
   * from producing elements by asserting back-pressure until its time comes.
   *
   * If the given [[Source]] gets upstream error - no elements from this [[Flow]] will be pulled.
   *
   * @see [[#prepend]].
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def prependMat[U >: Out, Mat2, Mat3](that: Graph[SourceShape[U], Mat2])(matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(prependGraph(that))(matF)

  /**
   * Provides a secondary source that will be consumed if this stream completes without any
   * elements passing by. As soon as the first element comes through this stream, the alternative
   * will be cancelled.
   *
   * Note that this Flow will be materialized together with the [[Source]] and just kept
   * from producing elements by asserting back-pressure until its time comes or it gets
   * cancelled.
   *
   * On errors the operator is failed regardless of source of the error.
   *
   * '''Emits when''' element is available from first stream or first stream closed without emitting any elements and an element
   *                  is available from the second stream
   *
   * '''Backpressures when''' downstream backpressures
   *
   * '''Completes when''' the primary stream completes after emitting at least one element, when the primary stream completes
   *                      without emitting and the secondary stream already has completed or when the secondary stream completes
   *
   * '''Cancels when''' downstream cancels and additionally the alternative is cancelled as soon as an element passes
   *                    by from this stream.
   */
  def orElseMat[U >: Out, Mat2, Mat3](secondary: Graph[SourceShape[U], Mat2])(
      matF: (Mat, Mat2) => Mat3): ReprMat[U, Mat3] =
    viaMat(orElseGraph(secondary))(matF)

  /**
   * Attaches the given [[Sink]] to this [[Flow]], meaning that elements that pass
   * through will also be sent to the [[Sink]].
   *
   * @see [[#alsoTo]]
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def alsoToMat[Mat2, Mat3](that: Graph[SinkShape[Out], Mat2])(matF: (Mat, Mat2) => Mat3): ReprMat[Out, Mat3] =
    viaMat(alsoToGraph(that))(matF)

  /**
   * Attaches the given [[Sink]] to this [[Flow]], meaning that elements will be sent to the [[Sink]]
   * instead of being passed through if the predicate `when` returns `true`.
   *
   * @see [[#divertTo]]
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def divertToMat[Mat2, Mat3](that: Graph[SinkShape[Out], Mat2], when: Out => Boolean)(
      matF: (Mat, Mat2) => Mat3): ReprMat[Out, Mat3] =
    viaMat(divertToGraph(that, when))(matF)

  /**
   * Attaches the given [[Sink]] to this [[Flow]] as a wire tap, meaning that elements that pass
   * through will also be sent to the wire-tap Sink, without the latter affecting the mainline flow.
   * If the wire-tap Sink backpressures, elements that would've been sent to it will be dropped instead.
   *
   * It is similar to [[#alsoToMat]] which does backpressure instead of dropping elements.
   *
   * @see [[#wireTap]]
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def wireTapMat[Mat2, Mat3](that: Graph[SinkShape[Out], Mat2])(matF: (Mat, Mat2) => Mat3): ReprMat[Out, Mat3] =
    viaMat(wireTapGraph(that))(matF)

  /**
   * Materializes to `Future[Done]` that completes on getting termination message.
   * The Future completes with success when received complete message from upstream or cancel
   * from downstream. It fails with the propagated error when received error message from
   * upstream or downstream.
   *
   * It is recommended to use the internally optimized `Keep.left` and `Keep.right` combiners
   * where appropriate instead of manually writing functions that pass through one of the values.
   */
  def watchTermination[Mat2]()(matF: (Mat, Future[Done]) => Mat2): ReprMat[Out, Mat2] =
    viaMat(GraphStages.terminationWatcher)(matF)

  /**
   * Transform the materialized value of this graph, leaving all other properties as they were.
   */
  def mapMaterializedValue[Mat2](f: Mat => Mat2): ReprMat[Out, Mat2]

  /**
   * Materializes to `FlowMonitor[Out]` that allows monitoring of the current flow. All events are propagated
   * by the monitor unchanged. Note that the monitor inserts a memory barrier every time it processes an
   * event, and may therefor affect performance.
   *
   * The `combine` function is used to combine the `FlowMonitor` with this flow's materialized value.
   */
  @Deprecated
  @deprecated("Use monitor() or monitorMat(combine) instead", "2.5.17")
  def monitor[Mat2]()(combine: (Mat, FlowMonitor[Out]) => Mat2): ReprMat[Out, Mat2] =
    viaMat(GraphStages.monitor)(combine)

  /**
   * Materializes to `FlowMonitor[Out]` that allows monitoring of the current flow. All events are propagated
   * by the monitor unchanged. Note that the monitor inserts a memory barrier every time it processes an
   * event, and may therefor affect performance.
   *
   * The `combine` function is used to combine the `FlowMonitor` with this flow's materialized value.
   */
  def monitorMat[Mat2](combine: (Mat, FlowMonitor[Out]) => Mat2): ReprMat[Out, Mat2] =
    viaMat(GraphStages.monitor)(combine)

  /**
   * Materializes to `(Mat, FlowMonitor[Out])`, which is unlike most other operators (!),
   * in which usually the default materialized value keeping semantics is to keep the left value
   * (by passing `Keep.left()` to a `*Mat` version of a method). This operator is an exception from
   * that rule and keeps both values since dropping its sole purpose is to introduce that materialized value.
   *
   * The `FlowMonitor[Out]` allows monitoring of the current flow. All events are propagated
   * by the monitor unchanged. Note that the monitor inserts a memory barrier every time it processes an
   * event, and may therefor affect performance.
   */
  def monitor: ReprMat[Out, (Mat, FlowMonitor[Out])] =
    monitorMat(Keep.both)

}