Akka（27）： Stream：Use case-Connecting Slick-dbStream & Scalaz-stream-fs2

在以前的博文中我们介绍了Slick，它是一种FRM（Functional Relation Mapper）。有别于ORM，FRM的特点是函数式的语法可以支持灵活的对象组合（Query Composition）实现大规模的代码重复利用，但同时这些特点又影响了编程人员群体对FRM的接受程度，阻碍了FRM成为广为流行的一种数据库编程方式。所以我们只能从小众心态来探讨如何改善Slick现状，希望通过与某些Stream库集成，在Slick FRM的基础上恢复一些人们熟悉的Recordset数据库光标（cursor）操作方式，希望如此可以降低FRM数据库编程对函数式编程水平要求，能够吸引更多的编程人员接受FRM。刚好，在这篇讨论里我们希望能介绍一些Akka-Stream和外部系统集成对接的实际用例，把Slick数据库数据载入连接到Akka-Stream形成streaming-dataset应该是一个挺好的想法。Slick和Akka-Stream可以说是自然匹配的一对，它们都是同一个公司产品，都支持Reactive-Specification。Reactive系统的集成对象之间是通过公共界面Publisher来实现对接的。Slick提供了个Dababase.stream函数可以构建这个Publisher：

 /** Create a `Publisher` for Reactive Streams which, when subscribed to, will run the specified
      * `DBIOAction` and return the result directly as a stream without buffering everything first.
      * This method is only supported for streaming actions.
      *
      * The Publisher itself is just a stub that holds a reference to the action and this Database.
      * The action does not actually start to run until the call to `onSubscribe` returns, after
      * which the Subscriber is responsible for reading the full response or cancelling the
      * Subscription. The created Publisher can be reused to serve a multiple Subscribers,
      * each time triggering a new execution of the action.
      *
      * For the purpose of combinators such as `cleanup` which can run after a stream has been
      * produced, cancellation of a stream by the Subscriber is not considered an error. For
      * example, there is no way for the Subscriber to cause a rollback when streaming the
      * results of `someQuery.result.transactionally`.
      *
      * When using a JDBC back-end, all `onNext` calls are done synchronously and the ResultSet row
      * is not advanced before `onNext` returns. This allows the Subscriber to access LOB pointers
      * from within `onNext`. If streaming is interrupted due to back-pressure signaling, the next
      * row will be prefetched (in order to buffer the next result page from the server when a page
      * boundary has been reached). */
    final def stream[T](a: DBIOAction[_, Streaming[T], Nothing]): DatabasePublisher[T] = streamInternal(a, false)

这个DatabasePublisher[T]就是一个Publisher[T]：

/** A Reactive Streams `Publisher` for database Actions. */
abstract class DatabasePublisher[T] extends Publisher[T] { self =>
...
}

然后Akka-Stream可以通过Source.fromPublisher(publisher)构建Akka Source构件：

  /**
   * Helper to create [[Source]] from `Publisher`.
   *
   * Construct a transformation starting with given publisher. The transformation steps
   * are executed by a series of [[org.reactivestreams.Processor]] instances
   * that mediate the flow of elements downstream and the propagation of
   * back-pressure upstream.
   */
  def fromPublisher[T](publisher: Publisher[T]): Source[T, NotUsed] =
    fromGraph(new PublisherSource(publisher, DefaultAttributes.publisherSource, shape("PublisherSource")))

理论上Source.fromPublisher(db.stream(query))就可以构建一个Reactive-Stream-Source了。下面我们就建了例子来做示范：首先是Slick的铺垫代码boiler-code：

  val aqmraw = Models.AQMRawQuery
  val db = Database.forConfig("h2db")
  // aqmQuery.result returns Seq[(String,String,String,String)]
  val aqmQuery = aqmraw.map {r => (r.year,r.state,r.county,r.value)}
  // type alias
  type RowType = (String,String,String,String)
  // user designed strong typed resultset type. must extend FDAROW
  case class TypedRow(year: String, state: String, county: String, value: String) extends FDAROW
  // strong typed resultset conversion function. declared implicit to remind during compilation
  implicit def toTypedRow(row: RowType): TypedRow =
    TypedRow(row._1,row._2,row._3,row._4)

我们需要的其实就是aqmQuery，用它来构建DatabasePublisher：

  // construct DatabasePublisher from db.stream
  val dbPublisher: DatabasePublisher[RowType] = db.stream[RowType](aqmQuery.result)
  // construct akka source
  val source: Source[RowType,NotUsed] = Source.fromPublisher[RowType](dbPublisher)

有了dbPublisher就可以用Source.fromPublisher函数构建source了。现在我们试着运算这个Akka-Stream：

  implicit val actorSys = ActorSystem("actor-system")
  implicit val ec = actorSys.dispatcher
  implicit val mat = ActorMaterializer()

  source.take().map{row => toTypedRow(row)}.runWith(
    Sink.foreach(qmr => {
      println(s"州名: ${qmr.state}")
      println(s"县名：${qmr.county}")
      println(s"年份：${qmr.year}")
      println(s"取值：${qmr.value}")
      println("-------------")
    }))

  scala.io.StdIn.readLine()
  actorSys.terminate()

下面是运算结果：

州名: Alabama
县名：Elmore
年份：
取值：
-------------
州名: Alabama
县名：Jefferson
年份：
取值：
-------------
州名: Alabama
县名：Lawrence
年份：
取值：
-------------
州名: Alabama
县名：Madison
年份：
取值：
-------------
州名: Alabama
县名：Mobile
年份：
取值：
-------------
州名: Alabama
县名：Montgomery
年份：
取值：
-------------

显示我们已经成功的连接了Slick和Akka-Stream。

现在我们有了Reactive stream source，它是个akka-stream，该如何对接处于下游的scalaz-stream-fs2呢？我们知道：akka-stream是Reactive stream，而scalaz-stream-fs2是纯“拖式”pull-model stream，也就是说上面这个Reactive stream source必须被动等待下游的scalaz-stream-fs2来读取数据。按照Reactive-Stream规范，下游必须通过backpressure信号来知会上游是否可以发送数据状态，也就是说我们需要scalaz-stream-fs2来产生backpressure。scalaz-stream-fs2 async包里有个Queue结构：

/**
 * Asynchronous queue interface. Operations are all nonblocking in their
 * implementations, but may be 'semantically' blocking. For instance,
 * a queue may have a bound on its size, in which case enqueuing may
 * block until there is an offsetting dequeue.
 */
trait Queue[F[_], A] { self =>
  /**
   * Enqueues one element in this `Queue`.
   * If the queue is `full` this waits until queue is empty.
   *
   * This completes after `a`  has been successfully enqueued to this `Queue`
   */
  def enqueue1(a: A): F[Unit]

  /**
   * Enqueues each element of the input stream to this `Queue` by
   * calling `enqueue1` on each element.
   */
  def enqueue: Sink[F, A] = _.evalMap(enqueue1)
  /** Dequeues one `A` from this queue. Completes once one is ready. */
  def dequeue1: F[A]
  /** Repeatedly calls `dequeue1` forever. */
  def dequeue: Stream[F, A] = Stream.bracket(cancellableDequeue1)(d => Stream.eval(d._1), d => d._2).repeat
...
}

这个结构支持多线程操作，也就是说enqueue和dequeue可以在不同的线程里操作。值得关注的是：enqueue会block，只有在完成了dequeue后才能继续。这个dequeue就变成了抵消backpressure的有效方法了。具体操作方法是：上游在一个线程里用enqueue发送一个数据元素，然后等待下游完成在另一个线程里的dequeue操作，完成这个循环后再进行下一个元素的enqueue。enqueue代表akka-stream向scalaz-stream-fs2发送数据，可以用akka-stream的Sink构件来实现：

 class FS2Gate[T](q: fs2.async.mutable.Queue[Task,Option[T]]) extends GraphStage[SinkShape[T]] {
  val in = Inlet[T]("inport")
  val shape = SinkShape.of(in)

  override def createLogic(inheritedAttributes: Attributes): GraphStageLogic =
    new GraphStageLogic(shape) with InHandler {
      override def preStart(): Unit = {
        pull(in)          //initiate stream elements movement
        super.preStart()
      }

      override def onPush(): Unit = {
        q.enqueue1(Some(grab(in))).unsafeRun()
        pull(in)
      }

      override def onUpstreamFinish(): Unit = {
        q.enqueue1(None).unsafeRun()
        println("the end of stream !")
        completeStage()
      }

      override def onUpstreamFailure(ex: Throwable): Unit = {
        q.enqueue1(None).unsafeRun()
        completeStage()
      }

      setHandler(in,this)

    }
}

以上这个akka-stream GraphStage描述了对上游每一个元素的enqueue动作。我们可以用scalaz-stream-fs2的flatMap来序列化运算两个线程里的enqueue和dequeue：

   val fs2Stream: Stream[Task,RowType] = Stream.eval(async.boundedQueue[Task,Option[RowType]]())
     .flatMap { q =>
       Task(source.to(new FS2Gate[RowType](q)).run).unsafeRunAsyncFuture  //enqueue Task(new thread)
       pipe.unNoneTerminate(q.dequeue)      //dequeue in current thread
     }

这个函数返回fs2.Stream[Task,RowType]，是一种运算方案，我们必须run来实际运算：

  fs2Stream.map{row => toTypedRow(row)}
      .map(qmr => {
      println(s"州名: ${qmr.state}")
      println(s"县名：${qmr.county}")
      println(s"年份：${qmr.year}")
      println(s"取值：${qmr.value}")
      println("-------------")
    }).run.unsafeRun

通过测试运行，我们成功的为scalaz-stream-fs2实现了data streaming。

下面是本次示范的源代码：

import slick.jdbc.H2Profile.api._
import com.bayakala.funda._
import api._

import scala.language.implicitConversions
import scala.concurrent.duration._
import akka.actor._
import akka.stream._
import akka.stream.scaladsl._
import akka.stream.stage._
import slick.basic.DatabasePublisher
import akka._
import fs2._
import akka.stream.stage.{GraphStage, GraphStageLogic}

 class FS2Gate[T](q: fs2.async.mutable.Queue[Task,Option[T]]) extends GraphStage[SinkShape[T]] {
  val in = Inlet[T]("inport")
  val shape = SinkShape.of(in)

  override def createLogic(inheritedAttributes: Attributes): GraphStageLogic =
    new GraphStageLogic(shape) with InHandler {
      override def preStart(): Unit = {
        pull(in)          //initiate stream elements movement
        super.preStart()
      }

      override def onPush(): Unit = {
        q.enqueue1(Some(grab(in))).unsafeRun()
        pull(in)
      }

      override def onUpstreamFinish(): Unit = {
        q.enqueue1(None).unsafeRun()
        println("end of stream !!!!!!!")
        completeStage()
      }

      override def onUpstreamFailure(ex: Throwable): Unit = {
        q.enqueue1(None).unsafeRun()
        completeStage()
      }

      setHandler(in,this)

    }
}

object AkkaStreamSource extends App {

  val aqmraw = Models.AQMRawQuery
  val db = Database.forConfig("h2db")
  // aqmQuery.result returns Seq[(String,String,String,String)]
  val aqmQuery = aqmraw.map {r => (r.year,r.state,r.county,r.value)}
  // type alias
  type RowType = (String,String,String,String)
  // user designed strong typed resultset type. must extend FDAROW
  case class TypedRow(year: String, state: String, county: String, value: String) extends FDAROW
  // strong typed resultset conversion function. declared implicit to remind during compilation
  implicit def toTypedRow(row: RowType): TypedRow =
    TypedRow(row._1,row._2,row._3,row._4)
  // construct DatabasePublisher from db.stream
  val dbPublisher: DatabasePublisher[RowType] = db.stream[RowType](aqmQuery.result)
  // construct akka source
  val source: Source[RowType,NotUsed] = Source.fromPublisher[RowType](dbPublisher)

  implicit val actorSys = ActorSystem("actor-system")
  implicit val ec = actorSys.dispatcher
  implicit val mat = ActorMaterializer()

  /*
  source.take(10).map{row => toTypedRow(row)}.runWith(
    Sink.foreach(qmr => {
      println(s"州名: ${qmr.state}")
      println(s"县名：${qmr.county}")
      println(s"年份：${qmr.year}")
      println(s"取值：${qmr.value}")
      println("-------------")
    })) */

   val fs2Stream: Stream[Task,RowType] = Stream.eval(async.boundedQueue[Task,Option[RowType]]())
     .flatMap { q =>
       Task(source.to(new FS2Gate[RowType](q)).run).unsafeRunAsyncFuture  //enqueue Task(new thread)
       pipe.unNoneTerminate(q.dequeue)      //dequeue in current thread
     }

  fs2Stream.map{row => toTypedRow(row)}
      .map(qmr => {
      println(s"州名: ${qmr.state}")
      println(s"县名：${qmr.county}")
      println(s"年份：${qmr.year}")
      println(s"取值：${qmr.value}")
      println("-------------")
    }).run.unsafeRun

  scala.io.StdIn.readLine()
  actorSys.terminate()

}