Akka（16）： 持久化模式：PersistentFSM-可以自动修复的状态机器

前面我们讨论过FSM，一种专门为维护内部状态而设计的Actor，它的特点是一套特殊的DSL能很方便地进行状态转换。FSM的状态转换模式特别适合对应现实中的业务流程，因为它那套DSL可以更形象的描述业务功能。为了实现FSM的可用性，就必须为FSM再增加自我修复能力，PersistentFSM是FSM和PersistentActor的合并，是在状态机器模式的基础上再增加了状态转变事件的持久化，从而实现内部状态的自我修复功能的。在FSM结构基础上，PersistentFSM又增加了领域事件（domain-event）这一元素，也就是事件来源（event-sourcing）模式里持久化的目标。PersistentFSM trait是如下定义的：

/**
 * A FSM implementation with persistent state.
 *
 * Supports the usual [[akka.actor.FSM]] functionality with additional persistence features.
 * `PersistentFSM` is identified by 'persistenceId' value.
 * State changes are persisted atomically together with domain events, which means that either both succeed or both fail,
 * i.e. a state transition event will not be stored if persistence of an event related to that change fails.
 * Persistence execution order is: persist -> wait for ack -> apply state.
 * Incoming messages are deferred until the state is applied.
 * State Data is constructed based on domain events, according to user's implementation of applyEvent function.
 *
 */
trait PersistentFSM[S <: FSMState, D, E] extends PersistentActor with PersistentFSMBase[S, D, E] with ActorLogging {...}

我们看到：PersistentFSM继承了PersistentActor，代表它具备了事件来源模式中的事件持久化和日志恢复能力。继承的另一个类型PersistentFSMBase是FSM trait的重新定义，针对状态机器增加的持久化特性设计了一套持久化状态转换的DSL。PersistentFSM trait的三个类参数S,D,E分别代表状态类型（State）、状态数据（Data）、领域事件（event）。与FSM比较：PersistentFSM除增加了event参数外，State类型是以FSMState类型为基础的，方便对State进行序列化（serialization）：

 /**
   * FSM state and data snapshot
   *
   * @param stateIdentifier FSM state identifier
   * @param data FSM state data
   * @param timeout FSM state timeout
   * @tparam D state data type
   */
  @InternalApi
  private[persistence] case class PersistentFSMSnapshot[D](stateIdentifier: String, data: D, timeout: Option[FiniteDuration]) extends Message

  /**
   * FSMState base trait, makes possible for simple default serialization by conversion to String
   */
  trait FSMState {
    def identifier: String
  }

PersistentFSM程序结构与FSM相似：

class PersistentFSMActor extends PersistentFSM[StateType,DataType,EventType] {

  startWith(initState,initData)  //起始状态

  when(stateA) {...}             //处理各种状态
  when(stateB) {...}

  whenUnhandled {...}            //处理共性状态

  onTransition {...}             //状态转变跟踪

 }

从这个程序结构来看，日志恢复（recovery）receiveRecovery函数实现应该隐含在类型定义里面：

 /**
   * After recovery events are handled as in usual FSM actor
   */
  override def receiveCommand: Receive = {
    super[PersistentFSMBase].receive
  }

  /**
   * Discover the latest recorded state
   */
  override def receiveRecover: Receive = {
    case domainEventTag(event) ⇒ startWith(stateName, applyEvent(event, stateData))
    case StateChangeEvent(stateIdentifier, timeout) ⇒ startWith(statesMap(stateIdentifier), stateData, timeout)
    case SnapshotOffer(_, PersistentFSMSnapshot(stateIdentifier, data: D, timeout)) ⇒ startWith(statesMap(stateIdentifier), data, timeout)
    case RecoveryCompleted ⇒
      initialize()
      onRecoveryCompleted()
  }

注意initialize已经过时，不要再用，我们可以重写onRecoveryCompleted()来实现一些初始化工作。那么事件写入日志又放在哪里了呢：

  /**
   * Persist FSM State and FSM State Data
   */
  override private[akka] def applyState(nextState: State): Unit = {
    var eventsToPersist: immutable.Seq[Any] = nextState.domainEvents.toList

    //Prevent StateChangeEvent persistence when staying in the same state, except when state defines a timeout
    if (nextState.notifies || nextState.timeout.nonEmpty) {
      eventsToPersist = eventsToPersist :+ StateChangeEvent(nextState.stateName.identifier, nextState.timeout)
    }

    if (eventsToPersist.isEmpty) {
      //If there are no events to persist, just apply the state
      super.applyState(nextState)
    } else {
      //Persist the events and apply the new state after all event handlers were executed
      var nextData: D = stateData
      var handlersExecutedCounter = 

      def applyStateOnLastHandler() = {
        handlersExecutedCounter +=
        if (handlersExecutedCounter == eventsToPersist.size) {
          super.applyState(nextState using nextData)
          currentStateTimeout = nextState.timeout
          nextState.afterTransitionDo(stateData)
        }
      }

      persistAll[Any](eventsToPersist) {
        case domainEventTag(event) ⇒
          nextData = applyEvent(event, nextData)
          applyStateOnLastHandler()
        case StateChangeEvent(stateIdentifier, timeout) ⇒
          applyStateOnLastHandler()
      }
    }
  }

注意这个内部函数applyState重写（override）了父辈PersistentFSMBase中的applyState：

  /*
   * *******************************************
   *       Main actor receive() method
   * *******************************************
   */
  override def receive: Receive = {
    case TimeoutMarker(gen) ⇒
      if (generation == gen) {
        processMsg(StateTimeout, "state timeout")
      }
    case t @ Timer(name, msg, repeat, gen) ⇒
      if ((timers contains name) && (timers(name).generation == gen)) {
        if (timeoutFuture.isDefined) {
          timeoutFuture.get.cancel()
          timeoutFuture = None
        }
        generation +=
        if (!repeat) {
          timers -= name
        }
        processMsg(msg, t)
      }
    case SubscribeTransitionCallBack(actorRef) ⇒
      // TODO Use context.watch(actor) and receive Terminated(actor) to clean up list
      listeners.add(actorRef)
      // send current state back as reference point
      actorRef ! CurrentState(self, currentState.stateName, currentState.timeout)
    case Listen(actorRef) ⇒
      // TODO Use context.watch(actor) and receive Terminated(actor) to clean up list
      listeners.add(actorRef)
      // send current state back as reference point
      actorRef ! CurrentState(self, currentState.stateName, currentState.timeout)
    case UnsubscribeTransitionCallBack(actorRef) ⇒
      listeners.remove(actorRef)
    case Deafen(actorRef) ⇒
      listeners.remove(actorRef)
    case value ⇒
      if (timeoutFuture.isDefined) {
        timeoutFuture.get.cancel()
        timeoutFuture = None
      }
      generation +=
      processMsg(value, sender())
  }

  private def processMsg(value: Any, source: AnyRef): Unit = {
    val event = Event(value, currentState.stateData)
    processEvent(event, source)
  }

  private[akka] def processEvent(event: Event, source: AnyRef): Unit = {
    val stateFunc = stateFunctions(currentState.stateName)
    val nextState = if (stateFunc isDefinedAt event) {
      stateFunc(event)
    } else {
      // handleEventDefault ensures that this is always defined
      handleEvent(event)
    }
    applyState(nextState)
  }

  private[akka] def applyState(nextState: State): Unit = {
    nextState.stopReason match {
      case None ⇒ makeTransition(nextState)
      case _ ⇒
        nextState.replies.reverse foreach { r ⇒ sender() ! r }
        terminate(nextState)
        context.stop(self)
    }
  }

在PersistentFSM trait中的抽象函数receiveCommand在实现时直接调用了PersistentFSMBase中的receive：

 /**
   * After recovery events are handled as in usual FSM actor
   */
  override def receiveCommand: Receive = {
    super[PersistentFSMBase].receive
  }

PersistentFSM还需要实现抽象函数applyEvent：

  /**
   * Override this handler to define the action on Domain Event
   *
   * @param domainEvent domain event to apply
   * @param currentData state data of the previous state
   * @return updated state data
   */
  def applyEvent(domainEvent: E, currentData: D): D

这个函数的主要功能是针对发生的事件进行当前状态数据的转换。另一个需要实现的抽象函数是domainEventClassTag。这是一个ClassTag[E]实例，用来解决泛型E的模式匹配问题（由scala语言类型擦拭type-erasure造成）：

  /**
   * Enables to pass a ClassTag of a domain event base type from the implementing class
   *
   * @return [[scala.reflect.ClassTag]] of domain event base type
   */
  implicit def domainEventClassTag: ClassTag[E]

  /**
   * Domain event's [[scala.reflect.ClassTag]]
   * Used for identifying domain events during recovery
   */
  val domainEventTag = domainEventClassTag
...
  /**
   * Discover the latest recorded state
   */
  override def receiveRecover: Receive = {
    case domainEventTag(event) ⇒ startWith(stateName, applyEvent(event, stateData))
...
   persistAll[Any](eventsToPersist) {
        case domainEventTag(event) ⇒
          nextData = applyEvent(event, nextData)
          applyStateOnLastHandler()
        case StateChangeEvent(stateIdentifier, timeout) ⇒
          applyStateOnLastHandler()
      }

akka-persistentFSM官方文档中的例子挺有代表性，下面我就根据这个例子来进行示范。这是个电商购物车的例子。用PersistentFSM来实现最大的优点就是在任何情况下都可以保证购物车内容的一致性。而且可以自动保存电商用户所有的历史选购过程方便将来大数据分析-这已经是一种潮流了，甚至对中途暂时放弃了的购物车也可以在下次登陆时自动恢复。好了，我们先来研究一下这个例子：首先是数据结构：

import akka.persistence.fsm.PersistentFSM._

object WebShopping {
  sealed trait UserState extends FSMState  //状态类型
  case object LookingAround extends UserState {  //浏览状态，可转Shopping状态
    override def identifier: String = "Looking Around"
  }
  case object Shopping extends UserState {  //拣选状态，可转到Paid状态或超时变Inactive
    override def identifier: String = "Shopping"
  }
  case object Inactive extends UserState {  //停滞状态，可转回Shopping状态
    override def identifier: String = "Inactive"
  }
  case object Paid extends UserState {   //结账完成购物，只能查询购物结果，或退出
    override def identifier: String = "Paid"
  }

  case class Item(id: String, name: String, price: Float)
  //state data
  sealed trait ShoppingCart {  //true functional structure
    def addItem(item: Item): ShoppingCart
    def removeItem(id: String): ShoppingCart
    def empty(): ShoppingCart
  }
  case class LoadedCart(items: Seq[Item]) extends ShoppingCart {
    override def addItem(item: Item): ShoppingCart = LoadedCart(items :+ item)
    override def removeItem(id: String): ShoppingCart = {
      val newItems = items.filter {item => item.id != id}
      if (newItems.length > )
        LoadedCart(newItems)
      else
        EmptyCart
    }
    override def empty() = EmptyCart
  }
  case object EmptyCart extends ShoppingCart {
    override def addItem(item: Item) = LoadedCart(item :: Nil)
    override def empty() = this
    override def removeItem(id: String): ShoppingCart = this
  }

}

UserState是FSM的当前状态。状态代表FSM的流程，每种状态运行它自己的业务流程：

  when(LookingAround) {...}             //处理各种状态
  when(Shopping) {...}
  when(Inactive) {...}
  when(Paid) {...}
...

ShoppingCart代表FSM当前状态的数据。每种状态都有可能具备不同的数据。注意ShoppingCart是典型的函数式数据结构：不可变结构，任何更新操作都返回新的结构。StateData ShoppingCart是在抽象函数applyEvent里更新的。再看看applyEvent的函数款式：

  /**
   * Override this handler to define the action on Domain Event
   *
   * @param domainEvent domain event to apply
   * @param currentData state data of the previous state
   * @return updated state data
   */
  def applyEvent(domainEvent: E, currentData: D): D

要求用户提供这个函数的实现：根据发生的事件及当前状态数据产生新的状态数据。applyEvent函数是如下调用的：

 override def receiveRecover: Receive = {
    case domainEventTag(event) ⇒ startWith(stateName, applyEvent(event, stateData))
    case StateChangeEvent(stateIdentifier, timeout) ⇒ startWith(statesMap(stateIdentifier), stateData, timeout)
    case SnapshotOffer(_, PersistentFSMSnapshot(stateIdentifier, data: D, timeout)) ⇒ startWith(statesMap(stateIdentifier), data, timeout)
    case RecoveryCompleted ⇒
      initialize()
      onRecoveryCompleted()
  }
...
 /**
   * Persist FSM State and FSM State Data
   */
  override private[akka] def applyState(nextState: State): Unit = {
    var eventsToPersist: immutable.Seq[Any] = nextState.domainEvents.toList

    //Prevent StateChangeEvent persistence when staying in the same state, except when state defines a timeout
    if (nextState.notifies || nextState.timeout.nonEmpty) {
      eventsToPersist = eventsToPersist :+ StateChangeEvent(nextState.stateName.identifier, nextState.timeout)
    }

    if (eventsToPersist.isEmpty) {
      //If there are no events to persist, just apply the state
      super.applyState(nextState)
    } else {
      //Persist the events and apply the new state after all event handlers were executed
      var nextData: D = stateData
      var handlersExecutedCounter = 

      def applyStateOnLastHandler() = {
        handlersExecutedCounter +=
        if (handlersExecutedCounter == eventsToPersist.size) {
          super.applyState(nextState using nextData)
          currentStateTimeout = nextState.timeout
          nextState.afterTransitionDo(stateData)
        }
      }

      persistAll[Any](eventsToPersist) {
        case domainEventTag(event) ⇒
          nextData = applyEvent(event, nextData)
          applyStateOnLastHandler()
        case StateChangeEvent(stateIdentifier, timeout) ⇒
          applyStateOnLastHandler()
      }
    }
  }

状态转换是通过stay, goto,stop实现的：

 /**
   * Produce transition to other state.
   * Return this from a state function in order to effect the transition.
   *
   * This method always triggers transition events, even for `A -> A` transitions.
   * If you want to stay in the same state without triggering an state transition event use [[#stay]] instead.
   *
   * @param nextStateName state designator for the next state
   * @return state transition descriptor
   */
  final def goto(nextStateName: S): State = PersistentFSM.State(nextStateName, currentState.stateData)()

  /**
   * Produce "empty" transition descriptor.
   * Return this from a state function when no state change is to be effected.
   *
   * No transition event will be triggered by [[#stay]].
   * If you want to trigger an event like `S -> S` for `onTransition` to handle use `goto` instead.
   *
   * @return descriptor for staying in current state
   */
  final def stay(): State = goto(currentState.stateName).withNotification(false) // cannot directly use currentState because of the timeout field

  /**
   * Produce change descriptor to stop this FSM actor with reason "Normal".
   */
  final def stop(): State = stop(Normal)

状态数据转换是用applying实现的：

 /**
     * Specify domain events to be applied when transitioning to the new state.
     */
    @varargs def applying(events: E*): State[S, D, E] = {
      copy(domainEvents = domainEvents ++ events)
    }

    /**
     * Register a handler to be triggered after the state has been persisted successfully
     */
    def andThen(handler: D ⇒ Unit): State[S, D, E] = {
      copy(afterTransitionDo = handler)
    }

applying对State[S,D,E]类型进行操作，State[S,D,E]的定义如下：

 /**
   * This captures all of the managed state of the [[akka.actor.FSM]]: the state
   * name, the state data, possibly custom timeout, stop reason, replies
   * accumulated while processing the last message, possibly domain event and handler
   * to be executed after FSM moves to the new state (also triggered when staying in the same state)
   */
  final case class State[S, D, E](
    stateName:         S,
    stateData:         D,
    timeout:           Option[FiniteDuration] = None,
    stopReason:        Option[Reason]         = None,
    replies:           List[Any]              = Nil,
    domainEvents:      Seq[E]                 = Nil,
    afterTransitionDo: D ⇒ Unit               = { _: D ⇒ })(private[akka] val notifies: Boolean = true) {

    /**
     * Copy object and update values if needed.
     */
    @InternalApi
    private[akka] def copy(stateName: S = stateName, stateData: D = stateData, timeout: Option[FiniteDuration] = timeout, stopReason: Option[Reason] = stopReason, replies: List[Any] = replies, notifies: Boolean = notifies, domainEvents: Seq[E] = domainEvents, afterTransitionDo: D ⇒ Unit = afterTransitionDo): State[S, D, E] = {
      State(stateName, stateData, timeout, stopReason, replies, domainEvents, afterTransitionDo)(notifies)
    }

applying实际上是把发生事件存入一个清单domainEvents，然后在调用applyState函数时再施用：

 /**
   * Persist FSM State and FSM State Data
   */
  override private[akka] def applyState(nextState: State): Unit = {
    var eventsToPersist: immutable.Seq[Any] = nextState.domainEvents.toList
...

PersistentFSM继承了PersistentActor事件来源（event-sourcing）模式。下面是command和event的类型定义：

  sealed trait Command
  case class AddItem(item: Item) extends Command
  case class RemoveItem(id: String) extends Command
  case object Buy extends Command
  case object Leave extends Command
  case object GetCart extends Command

  sealed trait DomainEvent
  case class ItemAdded(item: Item) extends DomainEvent
  case class ItemRemoved(id: String) extends DomainEvent
  case object OrderClosed extends DomainEvent

我们知道：DomainEvent将会被写入日志，它与Command的关系是：运算某些Command时会产生DomainEvent，然后这些产生的DomainEvent会被写入日志。

我们开始设计这个PersistentFSM：

class WebShopping(webUserId: String) extends PersistentFSM[UserState,ShoppingCart,DomainEvent] {
  override def persistenceId: String = webUserId
  override def domainEventClassTag: ClassTag[DomainEvent] = classTag[DomainEvent]

  override def applyEvent(domainEvent: DomainEvent, currentCart: ShoppingCart): ShoppingCart =
     domainEvent match {
       case ItemAdded(item) => currentCart.addItem(item)
       case ItemRemoved(id) => currentCart.removeItem(id)
       case OrderClosed => currentCart.empty()
     }
}

我们首先实现了trait中的抽象函数。其中persistenceId代表了当前购物者的userid。这样我们就可以把用户的购物过程写入日志。试想想这里面的意义：我们用一个独立的Actor来处理一个用户的购物过程。Actor对资源要求很低，但运算能力却高效强大，一个服务器上如果有足够内存就可以轻松负载几十万甚至百万级Actor实例，如果再使用akka-cluster的话不知不觉我们已经实现了可以容纳百万级用户的电商网站了。

好了，现在我们看看这个PersistentFSM的完整业务流程：

class WebShopping(webUserId: String) extends PersistentFSM[UserState,ShoppingCart,DomainEvent] {
  override def persistenceId: String = webUserId
  override def domainEventClassTag: ClassTag[DomainEvent] = classTag[DomainEvent]

  override def applyEvent(event: DomainEvent, currentCart: ShoppingCart): ShoppingCart =
     event match {
       case ItemAdded(item) => currentCart.addItem(item)
       case ItemRemoved(id) => currentCart.removeItem(id)
       case OrderClosed => currentCart.empty()  //买单成功后清空ShoppingCart
     }

  startWith(LookingAround,EmptyCart)  //初次登陆购物状态

  when(LookingAround) {   //浏览时可以加入购物车转到Shopping状态
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"LookingAround-Adding Item: $item",currentCart))
      goto(Shopping) applying ItemAdded(item) forMax( second)
    case Event(GetCart,currentCart) =>
      stay replying currentCart
  }

  when(Shopping) {
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Adding Item: $item",currentCart))
      stay applying ItemAdded(item) forMax ( second) andThen {
        case cart @ _ =>
          context.system.eventStream.publish(CurrentCart(s"Shopping-after adding Item: $item",cart))
      }
    case Event(RemoveItem(id),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Removing Item: $id",currentCart))
      stay applying ItemRemoved(id) forMax ( second) andThen {
        case cart @ _ =>
          context.system.eventStream.publish(CurrentCart(s"Shopping-after removing Item: $id",cart))
     }
    case Event(Buy,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Buying",currentCart))
      goto(Paid) applying OrderClosed forMax ( second) andThen {
        case cart @ _ => saveStateSnapshot()
          context.system.eventStream.publish(CurrentCart(s"Shopping-after paid",cart))
      }

    case Event(Leave,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Leaving",currentCart))
      stop()
    case Event(StateTimeout,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Timeout",currentCart))
      goto(Inactive) forMax( second)
    case Event(GetCart,currentCart) =>
      stay replying currentCart
  }

  when(Inactive) {
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Inactive-Adding Item: $item",currentCart))
      goto(Shopping) applying ItemAdded(item) forMax( second)
    case Event(StateTimeout,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Inactive-Timeout",currentCart))
      stop()
  }

  when(Paid) {
    case Event(Leave,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Paid-Leaving",currentCart))
      stop()
    case Event(GetCart,currentCart) =>
      stay replying currentCart
  }
}

我们看到通过FSM的DSL，PersistentActor和FSM的具体技术特征和细节被隐藏了，呈现给编程人员的是一段对业务流程的描述，这样可以使整段代码代表的功能更贴近现实应用，容易理解。

下面是有关数据快照、日志维护以及过程跟踪等方法的示范：

  whenUnhandled {
    case Event(SaveSnapshotSuccess(metadata),currentCart) =>
      context.system.eventStream.publish(CurrentCart("Successfully saved snapshot",currentCart))
      //假如不需要保存历史购物过程，可以清理日志和旧快照
      deleteSnapshots(SnapshotSelectionCriteria(maxSequenceNr = metadata.sequenceNr - ))
      deleteMessages(metadata.sequenceNr)
      stay()
    case Event(SaveSnapshotFailure(metadata, reason),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Fail to save snapshot for $reason",currentCart))
      stay()
    case Event(DeleteMessagesSuccess(toSeq),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Succefully deleted journal upto: $toSeq",currentCart))
      stay()
    case Event(DeleteMessagesFailure(cause,toSeq),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Failed to delete journal upto: $toSeq because: $cause",currentCart))
      stay()
    case Event(DeleteSnapshotsSuccess(crit),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Successfully deleted snapshots for $crit",currentCart))
      stay()
    case Event(DeleteSnapshotsFailure(crit,cause),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Failed to delete snapshots $crit because: $cause",currentCart))
      stay()
  }

  onTransition {
    case LookingAround -> Shopping =>
       context.system.eventStream.publish(CurrentCart("LookingAround -> Shopping",stateData))
    case Shopping -> Inactive =>
      context.system.eventStream.publish(CurrentCart("Shopping -> Inactive",stateData))
    case Shopping -> Paid =>
      context.system.eventStream.publish(CurrentCart("Shopping -> Paid",stateData))
    case Inactive -> Shopping =>
      context.system.eventStream.publish(CurrentCart("Inactive -> Shopping",stateData))
  }
  override def onRecoveryCompleted(): Unit =
    context.system.eventStream.publish(CurrentCart("OnRecoveryCompleted",stateData))

  override def onPersistFailure(cause: Throwable, event: Any, seqNr: Long): Unit =
    context.system.eventStream.publish(CurrentCart(s"onPersistFailure ${cause.getMessage}",stateData))

  override def onPersistRejected(cause: Throwable, event: Any, seqNr: Long): Unit =
    context.system.eventStream.publish(CurrentCart(s"onPersistRejected ${cause.getMessage}",stateData))

  override def onRecoveryFailure(cause: Throwable, event: Option[Any]): Unit =
    context.system.eventStream.publish(CurrentCart(s"onRecoveryFailure ${cause.getMessage}",stateData))

下面是过程跟踪器的设计代码：

package persistentfsm.tracker
import akka.actor._
import persistentfsm.cart.WebShopping
object EventTracker {
  def props = Props(new EventTracker)
}
class EventTracker extends Actor {
  override def preStart(): Unit = {
    context.system.eventStream.subscribe(self,classOf[WebShopping.CurrentCart])
    super.preStart()
  }

  override def postStop(): Unit = {
    context.system.eventStream.unsubscribe(self)
    super.postStop()
  }

  override def receive: Receive = {
    case WebShopping.CurrentCart(loc,cart) =>
      println(loc)
      cart match {
        case WebShopping.EmptyCart => println("empty cart!")
        case WebShopping.LoadedCart(items) => println(s"Current content in cart: $items")
      }
  }

}

用下面的代码来测试运行：

package persistentfsm.demo
import persistentfsm.cart._
import persistentfsm.tracker._
import akka.actor._
import WebShopping._

object PersistentFSMDemo extends App {
   val pfSystem = ActorSystem("persistentfsm-system")
   val trackerActor = pfSystem.actorOf(EventTracker.props,"tracker")
   val cart123 = pfSystem.actorOf(WebShopping.props(""))

   cart123 ! GetCart
   cart123 ! AddItem(Item("","Cigar",12.50))
   cart123 ! AddItem(Item("","Wine",18.30))
   cart123 ! AddItem(Item("","Coffee",5.50))
   cart123 ! GetCart
   cart123 ! RemoveItem("")
   cart123 ! Buy
   cart123 ! GetCart
   cart123 ! AddItem(Item("","Bread",3.25))
   cart123 ! AddItem(Item("","Cake",5.25))

   scala.io.StdIn.readLine()

   pfSystem.terminate()

}

重复运算可以得出：结账后选购的商品可以恢复。如果中途异常退出，购物车中已经选购的商品任然保留。

下面是本次示范的完整源代码：

build.sbt

name := "persistent-fsm"

version := "1.0"

scalaVersion := "2.11.9"

sbtVersion := "0.13.5"

libraryDependencies ++= Seq(
  "com.typesafe.akka"           %% "akka-actor"       % "2.5.3",
  "com.typesafe.akka"           %% "akka-persistence" % "2.5.3",
  "ch.qos.logback" % "logback-classic" % "1.1.7",
  "com.typesafe.akka" %% "akka-persistence-cassandra" % "0.54",
  "com.typesafe.akka" %% "akka-persistence-cassandra-launcher" % "0.54" % Test
)

application.conf

akka {
  persistence {
    journal.plugin = "cassandra-journal"
    snapshot-store.plugin = "cassandra-snapshot-store"
    fsm {
      snapshot-after =
    }
  }
}
akka.actor.warn-about-java-serializer-usage = off

WebShopping.scala

package persistentfsm.cart
import WebShopping._
import akka.persistence.fsm._
import akka.persistence.fsm.PersistentFSM._
import akka.persistence._
import akka.actor._
import scala.concurrent.duration._
import scala.reflect._

object WebShopping {
  sealed trait UserState extends FSMState  //状态类型
  case object LookingAround extends UserState {  //浏览状态，可转Shopping状态
    override def identifier: String = "Looking Around"
  }
  case object Shopping extends UserState {  //拣选状态，可转到Paid状态或超时变Inactive
    override def identifier: String = "Shopping"
  }
  case object Inactive extends UserState {  //停滞状态，可转回Shopping状态
    override def identifier: String = "Inactive"
  }
  case object Paid extends UserState {   //结账完成购物，只能查询购物结果，或退出
    override def identifier: String = "Paid"
  }

  case class Item(id: String, name: String, price: Double)
  //state data
  sealed trait ShoppingCart {  //true functional structure
    def addItem(item: Item): ShoppingCart
    def removeItem(id: String): ShoppingCart
    def empty(): ShoppingCart
  }
  case class LoadedCart(items: Seq[Item]) extends ShoppingCart {
    override def addItem(item: Item): ShoppingCart = LoadedCart(items :+ item)
    override def removeItem(id: String): ShoppingCart = {
      val newItems = items.filter {item => item.id != id}
      if (newItems.length > )
        LoadedCart(newItems)
      else
        EmptyCart
    }
    override def empty() = EmptyCart
  }
  case object EmptyCart extends ShoppingCart {
    override def addItem(item: Item) = LoadedCart(item :: Nil)
    override def empty() = this
    override def removeItem(id: String): ShoppingCart = this
  }

  sealed trait Command
  case class AddItem(item: Item) extends Command
  case class RemoveItem(id: String) extends Command
  case object Buy extends Command
  case object Leave extends Command
  case object GetCart extends Command

  sealed trait DomainEvent
  case class ItemAdded(item: Item) extends DomainEvent
  case class ItemRemoved(id: String) extends DomainEvent
  case object OrderClosed extends DomainEvent
//logging message type
  case class CurrentCart(location: String, cart: ShoppingCart)

  def props(uid: String) = Props(new WebShopping(uid))

}
class WebShopping(webUserId: String) extends PersistentFSM[UserState,ShoppingCart,DomainEvent] {
  override def persistenceId: String = webUserId
  override def domainEventClassTag: ClassTag[DomainEvent] = classTag[DomainEvent]

  override def applyEvent(event: DomainEvent, currentCart: ShoppingCart): ShoppingCart =
     event match {
       case ItemAdded(item) => currentCart.addItem(item)
       case ItemRemoved(id) => currentCart.removeItem(id)
       case OrderClosed => currentCart.empty()  //买单成功后清空ShoppingCart
     }

  startWith(LookingAround,EmptyCart)  //初次登陆购物状态

  when(LookingAround) {   //浏览时可以加入购物车转到Shopping状态
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"LookingAround-Adding Item: $item",currentCart))
      goto(Shopping) applying ItemAdded(item) forMax( second)
    case Event(GetCart,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"LookingAround-Showing",currentCart))
      stay replying currentCart
  }

  when(Shopping) {
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Adding Item: $item",currentCart))
      stay applying ItemAdded(item) forMax ( second) andThen {
        case cart @ _ =>
          context.system.eventStream.publish(CurrentCart(s"Shopping-after adding Item: $item",cart))
      }
    case Event(RemoveItem(id),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Removing Item: $id",currentCart))
      stay applying ItemRemoved(id) forMax ( second) andThen {
        case cart @ _ =>
          context.system.eventStream.publish(CurrentCart(s"Shopping-after removing Item: $id",cart))
     }
    case Event(Buy,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Buying",currentCart))
      goto(Paid) applying OrderClosed forMax ( second) andThen {
        case cart @ _ => saveStateSnapshot()
          context.system.eventStream.publish(CurrentCart(s"Shopping-after paid",cart))
      }

    case Event(Leave,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Leaving",currentCart))
      stop()
    case Event(StateTimeout,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Shopping-Timeout",currentCart))
      goto(Inactive) forMax( second)
    case Event(GetCart,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"LookingAround-Showing",currentCart))
      stay replying currentCart
  }

  when(Inactive) {
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Inactive-Adding Item: $item",currentCart))
      goto(Shopping) applying ItemAdded(item) forMax( second)
    case Event(StateTimeout,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Inactive-Timeout",currentCart))
      stop()
  }

  when(Paid) {
    case Event(Leave,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Paid-Leaving",currentCart))
      stop()
    case Event(GetCart,currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Paid-Showing",currentCart))
      stay replying currentCart
    case Event(AddItem(item),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Paid-Adding Item: $item",currentCart))
      goto(Shopping) applying ItemAdded(item) forMax( second)
  }

  whenUnhandled {
    case Event(SaveSnapshotSuccess(metadata),currentCart) =>
      context.system.eventStream.publish(CurrentCart("Successfully saved snapshot",currentCart))
      //假如不需要保存历史购物过程，可以清理日志和旧快照
      deleteSnapshots(SnapshotSelectionCriteria(maxSequenceNr = metadata.sequenceNr - ))
      deleteMessages(metadata.sequenceNr)
      stay()
    case Event(SaveSnapshotFailure(metadata, reason),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Fail to save snapshot for $reason",currentCart))
      stay()
    case Event(DeleteMessagesSuccess(toSeq),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Succefully deleted journal upto: $toSeq",currentCart))
      stay()
    case Event(DeleteMessagesFailure(cause,toSeq),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Failed to delete journal upto: $toSeq because: $cause",currentCart))
      stay()
    case Event(DeleteSnapshotsSuccess(crit),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Successfully deleted snapshots for $crit",currentCart))
      stay()
    case Event(DeleteSnapshotsFailure(crit,cause),currentCart) =>
      context.system.eventStream.publish(CurrentCart(s"Failed to delete snapshots $crit because: $cause",currentCart))
      stay()
    case _ => goto(LookingAround)
  }

  onTransition {
    case LookingAround -> Shopping =>
       context.system.eventStream.publish(CurrentCart("State changed: LookingAround -> Shopping",stateData))
    case Shopping -> Inactive =>
      context.system.eventStream.publish(CurrentCart("State changed: Shopping -> Inactive",stateData))
    case Shopping -> Paid =>
      context.system.eventStream.publish(CurrentCart("State changed: Shopping -> Paid",stateData))
    case Inactive -> Shopping =>
      context.system.eventStream.publish(CurrentCart("State changed: Inactive -> Shopping",stateData))
    case Paid -> LookingAround =>
      context.system.eventStream.publish(CurrentCart("State changed: Paid -> LookingAround",stateData))
  }
  override def onRecoveryCompleted(): Unit = {
    context.system.eventStream.publish(CurrentCart("OnRecoveryCompleted", stateData))
  }

  override def onPersistFailure(cause: Throwable, event: Any, seqNr: Long): Unit =
    context.system.eventStream.publish(CurrentCart(s"onPersistFailure ${cause.getMessage}",stateData))

  override def onPersistRejected(cause: Throwable, event: Any, seqNr: Long): Unit =
    context.system.eventStream.publish(CurrentCart(s"onPersistRejected ${cause.getMessage}",stateData))

  override def onRecoveryFailure(cause: Throwable, event: Option[Any]): Unit =
    context.system.eventStream.publish(CurrentCart(s"onRecoveryFailure ${cause.getMessage}",stateData))

}

EventTracker.scala

package persistentfsm.tracker
import akka.actor._
import persistentfsm.cart.WebShopping
object EventTracker {
  def props = Props(new EventTracker)
}
class EventTracker extends Actor {
  override def preStart(): Unit = {
    context.system.eventStream.subscribe(self,classOf[WebShopping.CurrentCart])
    super.preStart()
  }

  override def postStop(): Unit = {
    context.system.eventStream.unsubscribe(self)
    super.postStop()
  }

  override def receive: Receive = {
    case WebShopping.CurrentCart(loc,cart) =>
      println(loc)
      cart match {
        case WebShopping.EmptyCart => println("empty cart!")
        case WebShopping.LoadedCart(items) => println(s"Current content in cart: $items")
      }
  }

}

PersistentFSMDemo.scala

package persistentfsm.demo
import persistentfsm.cart._
import persistentfsm.tracker._
import akka.actor._
import WebShopping._

object PersistentFSMDemo extends App {
   val pfSystem = ActorSystem("persistentfsm-system")
   val trackerActor = pfSystem.actorOf(EventTracker.props,"tracker")
   val cart123 = pfSystem.actorOf(WebShopping.props(""))

   cart123 ! GetCart
   cart123 ! AddItem(Item("","Cigar",12.50))
   cart123 ! AddItem(Item("","Wine",18.30))
   cart123 ! AddItem(Item("","Coffee",5.50))
   cart123 ! GetCart
   cart123 ! RemoveItem("")
   cart123 ! Buy
   cart123 ! GetCart
   cart123 ! AddItem(Item("","Bread",3.25))
   cart123 ! AddItem(Item("","Cake",5.25))

   scala.io.StdIn.readLine()

   pfSystem.terminate()

}