akka-typed中已经没有PersistentActor了。取而代之的是带有EventSourcedBehavior的actor,也就是一种专门支持EventSource模式的actor。EventSource的原理和作用在之前的博客里已经有了比较详细的介绍,这里就不再重复了。本篇直接从EventsourcedBehavior actor的具体应用开始介绍。支持EventSource应用的基本数据类型包括 指令Command, 事件Event,状态State。EventSourcing其实就是一个有限状态机fsm finite-state-machine,执行Command,产生Event,改变State,终而复始。下面是一个简单的EventSource类型定义:

trait CborSerializable {}
object Cart {
case class Item(name: String, price: Double) sealed trait Command extends CborSerializable
sealed trait Event //commands
case class AddItem(item: Item) extends Command
case object PayCart extends Command //event
case class ItemAdded(item: Item) extends Event
case object CartPaid extends Event //state
case class CartLoad(load: Set[Item] = Set.empty) val commandHandler: (CartLoad, Command) => Effect[Event,CartLoad] = { (state, cmd) =>
cmd match {
case AddItem(item) =>
Effect.persist(ItemAdded(item))
case PayCart =>
Effect.persist(CartPaid)
}
} val eventHandler: (CartLoad,Event) => CartLoad = { (state,evt) =>
evt match {
case ItemAdded(item) =>
val sts = state.copy(load = state.load+item)
println(s"current cart loading: ${sts}")
sts
case CartPaid =>
val sts = state.copy(load = Set.empty)
println(s"current cart loading: ${sts.load}")
sts
}
} def apply(): Behavior[Command] = EventSourcedBehavior[Command,Event,CartLoad](
persistenceId = PersistenceId("",""),
emptyState = CartLoad(),
commandHandler = commandHandler,
eventHandler = eventHandler
) } object EventSource extends App {
import Cart._
val cart = ActorSystem(Cart(),"shopping-cart")
cart ! Cart.AddItem(Item("banana",11.20))
cart ! Cart.AddItem(Item("watermelon",4.70)) scala.io.StdIn.readLine() cart.terminate() }

首先要搞清楚几件事:EvenSourcedBehavior定义了一个actor。从Behavior[Command]这个结果类型来看,这个actor可以接收并处理Command类型的消息。既然是个actor那么应该具备了receiveMessage,receiveSignal这两项基本能力,但我们又不用自己来定义这些功能。怎么回事呢?看看EventSourcedBehavior的源代码吧:

object EventSourcedBehavior {
...
def apply[Command, Event, State](
persistenceId: PersistenceId,
emptyState: State,
commandHandler: (State, Command) => Effect[Event, State],
eventHandler: (State, Event) => State): EventSourcedBehavior[Command, Event, State] = {
val loggerClass = LoggerClass.detectLoggerClassFromStack(classOf[EventSourcedBehavior[_, _, _]], logPrefixSkipList)
EventSourcedBehaviorImpl(persistenceId, emptyState, commandHandler, eventHandler, loggerClass)
}
...
}

这个EventSourcedBehavior就是某种Behavior。它的所有特殊功能看来应该是在EventSourcedBehaviorsImpl里实现的:

private[akka] final case class EventSourcedBehaviorImpl[Command, Event, State](
persistenceId: PersistenceId,
emptyState: State,
commandHandler: EventSourcedBehavior.CommandHandler[Command, Event, State],
eventHandler: EventSourcedBehavior.EventHandler[State, Event],
loggerClass: Class[_],
...
) extends EventSourcedBehavior[Command, Event, State] {
...
Behaviors
.supervise {
Behaviors.setup[Command] { _ =>
val eventSourcedSetup = new BehaviorSetup(
ctx.asInstanceOf[ActorContext[InternalProtocol]],
persistenceId,
emptyState,
commandHandler,
eventHandler,
WriterIdentity.newIdentity(),
actualSignalHandler,
tagger,
eventAdapter,
snapshotAdapter,
snapshotWhen,
recovery,
retention,
holdingRecoveryPermit = false,
settings = settings,
stashState = stashState) // needs to accept Any since we also can get messages from the journal
// not part of the user facing Command protocol
def interceptor: BehaviorInterceptor[Any, InternalProtocol] = new BehaviorInterceptor[Any, InternalProtocol] { import BehaviorInterceptor._
override def aroundReceive(
ctx: typed.TypedActorContext[Any],
msg: Any,
target: ReceiveTarget[InternalProtocol]): Behavior[InternalProtocol] = {
val innerMsg = msg match {
case res: JournalProtocol.Response => InternalProtocol.JournalResponse(res)
case res: SnapshotProtocol.Response => InternalProtocol.SnapshotterResponse(res)
case RecoveryPermitter.RecoveryPermitGranted => InternalProtocol.RecoveryPermitGranted
case internal: InternalProtocol => internal // such as RecoveryTickEvent
case cmd: Command @unchecked => InternalProtocol.IncomingCommand(cmd)
}
target(ctx, innerMsg)
} override def aroundSignal(
ctx: typed.TypedActorContext[Any],
signal: Signal,
target: SignalTarget[InternalProtocol]): Behavior[InternalProtocol] = {
if (signal == PostStop) {
eventSourcedSetup.cancelRecoveryTimer()
// clear stash to be GC friendly
stashState.clearStashBuffers()
}
target(ctx, signal)
} override def toString: String = "EventSourcedBehaviorInterceptor"
} Behaviors.intercept(() => interceptor)(RequestingRecoveryPermit(eventSourcedSetup)).narrow
} }
.onFailure[JournalFailureException](supervisionStrategy)
}

EventSourcedBehaviorImpl还是一种Behavior[Command],它又是通过一个BehaviorInterceptor实现的。BehaviorInterceptor.aroundReceive和BehaviorInterceptor.aroundSignal可以代替receiveMessage和receiveSignal的工作,这点从这两个函数的结果类型可以得到一些验证:

 /* @tparam Outer The outer message type – the type of messages the intercepting behavior will accept
* @tparam Inner The inner message type - the type of message the wrapped behavior accepts
*
* @see [[BehaviorSignalInterceptor]]
*/
abstract class BehaviorInterceptor[Outer, Inner](val interceptMessageClass: Class[Outer]) {
import BehaviorInterceptor._
... /**
* Intercept a message sent to the running actor. Pass the message on to the next behavior
* in the stack by passing it to `target.apply`, return `Behaviors.same` without invoking `target`
* to filter out the message.
*
* @return The behavior for next message or signal
*/
def aroundReceive(ctx: TypedActorContext[Outer], msg: Outer, target: ReceiveTarget[Inner]): Behavior[Inner] /**
* Override to intercept a signal sent to the running actor. Pass the signal on to the next behavior
* in the stack by passing it to `target.apply`.
*
* @return The behavior for next message or signal
*
* @see [[BehaviorSignalInterceptor]]
*/
def aroundSignal(ctx: TypedActorContext[Outer], signal: Signal, target: SignalTarget[Inner]): Behavior[Inner] ...
}

另外,对于EventSourcedBehavior来说,收到Command, 处理Command方式应该是通过外部提供的这个commandHandler来实现才是最值得注意的:

  final class HandlingCommands(state: RunningState[S])
extends AbstractBehavior[InternalProtocol](setup.context)
with WithSeqNrAccessible { def onMessage(msg: InternalProtocol): Behavior[InternalProtocol] = msg match {
case IncomingCommand(c: C @unchecked) => onCommand(state, c)
case JournalResponse(r) => onDeleteEventsJournalResponse(r, state.state)
case SnapshotterResponse(r) => onDeleteSnapshotResponse(r, state.state)
case get: GetState[S @unchecked] => onGetState(get)
case _ => Behaviors.unhandled
} override def onSignal: PartialFunction[Signal, Behavior[InternalProtocol]] = {
case PoisonPill =>
if (isInternalStashEmpty && !isUnstashAllInProgress) Behaviors.stopped
else new HandlingCommands(state.copy(receivedPoisonPill = true))
case signal =>
if (setup.onSignal(state.state, signal, catchAndLog = false)) this
else Behaviors.unhandled
} def onCommand(state: RunningState[S], cmd: C): Behavior[InternalProtocol] = {
val effect = setup.commandHandler(state.state, cmd)
applyEffects(cmd, state, effect.asInstanceOf[EffectImpl[E, S]]) // TODO can we avoid the cast?
} ...
}

上面这段代码已经足够说明了。根据commandHandler和eventHandler的函数类型可以得出EventSourcedBehavior处理流程 (State, Command) => (State, Event) => new State, 最终输出new State:

object EventSourcedBehavior {
type CommandHandler[Command, Event, State] = (State, Command) => Effect[Event, State]
type EventHandler[State, Event] = (State, Event) => State
...
}

commandHandler返回Effect[Event,State]类型结果,也就是说处理Command过程就是产生Event过程,下面是Effect的各种选项:

object Effect {

  /**
* Persist a single event
*
* Side effects can be chained with `thenRun`
*/
def persist[Event, State](event: Event): EffectBuilder[Event, State] = Persist(event) /**
* Persist multiple events
*
* Side effects can be chained with `thenRun`
*/
def persist[Event, A <: Event, B <: Event, State](evt1: A, evt2: B, events: Event*): EffectBuilder[Event, State] =
persist(evt1 :: evt2 :: events.toList) /**
* Persist multiple events
*
* Side effects can be chained with `thenRun`
*/
def persist[Event, State](events: im.Seq[Event]): EffectBuilder[Event, State] =
PersistAll(events) /**
* Do not persist anything
*
* Side effects can be chained with `thenRun`
*/
def none[Event, State]: EffectBuilder[Event, State] = PersistNothing.asInstanceOf[EffectBuilder[Event, State]] /**
* This command is not handled, but it is not an error that it isn't.
*
* Side effects can be chained with `thenRun`
*/
def unhandled[Event, State]: EffectBuilder[Event, State] = Unhandled.asInstanceOf[EffectBuilder[Event, State]] /**
* Stop this persistent actor
* Side effects can be chained with `thenRun`
*/
def stop[Event, State](): EffectBuilder[Event, State] =
none.thenStop() /**
* Stash the current command. Can be unstashed later with [[Effect.unstashAll]].
*
* Note that the stashed commands are kept in an in-memory buffer, so in case of a crash they will not be
* processed. They will also be discarded if the actor is restarted (or stopped) due to that an exception was
* thrown from processing a command or side effect after persisting. The stash buffer is preserved for persist
* failures if a backoff supervisor strategy is defined with [[EventSourcedBehavior.onPersistFailure]].
*
* Side effects can be chained with `thenRun`
*/
def stash[Event, State](): ReplyEffect[Event, State] =
Stash.asInstanceOf[EffectBuilder[Event, State]].thenNoReply() /**
* Unstash the commands that were stashed with [[Effect.stash]].
*
* It's allowed to stash messages while unstashing. Those newly added
* commands will not be processed by this `unstashAll` effect and have to be unstashed
* by another `unstashAll`.
*
* @see [[EffectBuilder.thenUnstashAll]]
*/
def unstashAll[Event, State](): Effect[Event, State] =
CompositeEffect(none.asInstanceOf[EffectBuilder[Event, State]], SideEffect.unstashAll[State]()) /**
* Send a reply message to the command. The type of the
* reply message must conform to the type specified by the passed replyTo `ActorRef`.
*
* This has the same semantics as `cmd.replyTo.tell`.
*
* It is provided as a convenience (reducing boilerplate) and a way to enforce that replies are not forgotten
* when the `EventSourcedBehavior` is created with [[EventSourcedBehavior.withEnforcedReplies]]. When
* `withEnforcedReplies` is used there will be compilation errors if the returned effect isn't a [[ReplyEffect]].
* The reply message will be sent also if `withEnforcedReplies` isn't used, but then the compiler will not help
* finding mistakes.
*/
def reply[ReplyMessage, Event, State](replyTo: ActorRef[ReplyMessage])(
replyWithMessage: ReplyMessage): ReplyEffect[Event, State] =
none[Event, State].thenReply[ReplyMessage](replyTo)(_ => replyWithMessage) /**
* When [[EventSourcedBehavior.withEnforcedReplies]] is used there will be compilation errors if the returned effect
* isn't a [[ReplyEffect]]. This `noReply` can be used as a conscious decision that a reply shouldn't be
* sent for a specific command or the reply will be sent later.
*/
def noReply[Event, State]: ReplyEffect[Event, State] =
none.thenNoReply() }

接着用handleEvent来根据产生的Event更新State,如下:

    @tailrec def applyEffects(
msg: Any,
state: RunningState[S],
effect: Effect[E, S],
sideEffects: immutable.Seq[SideEffect[S]] = Nil): Behavior[InternalProtocol] = {
if (setup.log.isDebugEnabled && !effect.isInstanceOf[CompositeEffect[_, _]])
setup.log.debugN(
s"Handled command [{}], resulting effect: [{}], side effects: [{}]",
msg.getClass.getName,
effect,
sideEffects.size) effect match {
case CompositeEffect(eff, currentSideEffects) =>
// unwrap and accumulate effects
applyEffects(msg, state, eff, currentSideEffects ++ sideEffects) case Persist(event) =>
// apply the event before persist so that validation exception is handled before persisting
// the invalid event, in case such validation is implemented in the event handler.
// also, ensure that there is an event handler for each single event
val newState = state.applyEvent(setup, event) val eventToPersist = adaptEvent(event)
val eventAdapterManifest = setup.eventAdapter.manifest(event) val newState2 = internalPersist(setup.context, msg, newState, eventToPersist, eventAdapterManifest) val shouldSnapshotAfterPersist = setup.shouldSnapshot(newState2.state, event, newState2.seqNr) persistingEvents(newState2, state, numberOfEvents = , shouldSnapshotAfterPersist, sideEffects) case PersistAll(events) =>
if (events.nonEmpty) {
// apply the event before persist so that validation exception is handled before persisting
// the invalid event, in case such validation is implemented in the event handler.
// also, ensure that there is an event handler for each single event
var seqNr = state.seqNr
val (newState, shouldSnapshotAfterPersist) = events.foldLeft((state, NoSnapshot: SnapshotAfterPersist)) {
case ((currentState, snapshot), event) =>
seqNr +=
val shouldSnapshot =
if (snapshot == NoSnapshot) setup.shouldSnapshot(currentState.state, event, seqNr) else snapshot
(currentState.applyEvent(setup, event), shouldSnapshot)
} val eventsToPersist = events.map(evt => (adaptEvent(evt), setup.eventAdapter.manifest(evt))) val newState2 = internalPersistAll(setup.context, msg, newState, eventsToPersist) persistingEvents(newState2, state, events.size, shouldSnapshotAfterPersist, sideEffects) } else {
// run side-effects even when no events are emitted
tryUnstashOne(applySideEffects(sideEffects, state))
} case _: PersistNothing.type =>
tryUnstashOne(applySideEffects(sideEffects, state)) case _: Unhandled.type =>
import akka.actor.typed.scaladsl.adapter._
setup.context.system.toClassic.eventStream
.publish(UnhandledMessage(msg, setup.context.system.toClassic.deadLetters, setup.context.self.toClassic))
tryUnstashOne(applySideEffects(sideEffects, state)) case _: Stash.type =>
stashUser(IncomingCommand(msg))
tryUnstashOne(applySideEffects(sideEffects, state))
}
}

好了,基本原理都在这了,再挖下去会更肮脏。为上面的例子设了个运行环境,主要是测试persistence-cassandra-plugin的正确设置,如下:

build.sbt

 name := "learn-akka-typed"

version := "0.1"

scalaVersion := "2.13.1"
scalacOptions in Compile ++= Seq("-deprecation", "-feature", "-unchecked", "-Xlog-reflective-calls", "-Xlint")
javacOptions in Compile ++= Seq("-Xlint:unchecked", "-Xlint:deprecation") val AkkaVersion = "2.6.5"
val AkkaPersistenceCassandraVersion = "1.0.0" libraryDependencies ++= Seq(
"com.typesafe.akka" %% "akka-cluster-sharding-typed" % AkkaVersion,
"com.typesafe.akka" %% "akka-persistence-typed" % AkkaVersion,
"com.typesafe.akka" %% "akka-persistence-query" % AkkaVersion,
"com.typesafe.akka" %% "akka-serialization-jackson" % AkkaVersion,
"com.typesafe.akka" %% "akka-persistence-cassandra" % AkkaPersistenceCassandraVersion,
"com.typesafe.akka" %% "akka-slf4j" % AkkaVersion,
"ch.qos.logback" % "logback-classic" % "1.2.3"
)

application.conf

akka.actor.allow-java-serialization = on
akka {
loglevel = DEBUG
actor {
serialization-bindings {
"com.learn.akka.CborSerializable" = jackson-cbor
}
}
# use Cassandra to store both snapshots and the events of the persistent actors
persistence {
journal.plugin = "akka.persistence.cassandra.journal"
snapshot-store.plugin = "akka.persistence.cassandra.snapshot"
} }
akka.persistence.cassandra {
# don't use autocreate in production
journal.keyspace = "poc"
journal.keyspace-autocreate = on
journal.tables-autocreate = on
snapshot.keyspace = "poc_snapshot"
snapshot.keyspace-autocreate = on
snapshot.tables-autocreate = on
} datastax-java-driver {
basic.contact-points = ["192.168.11.189:9042"]
basic.load-balancing-policy.local-datacenter = "datacenter1"
}

akka-typed(3) - PersistentActor has EventSourcedBehavior的更多相关文章

  1. Akka Typed 官方文档之随手记

    ️ 引言 近两年,一直在折腾用FP与OO共存的编程语言Scala,采取以函数式编程为主的方式,结合TDD和BDD的手段,采用Domain Driven Design的方法学,去构造DDDD应用(Dom ...

  2. Akka源码分析-Akka Typed

    对不起,akka typed 我是不准备进行源码分析的,首先这个库的API还没有release,所以会may change,也就意味着其概念和设计包括API都会修改,基本就没有再深入分析源码的意义了. ...

  3. Akka Typed系列:协议&行为

    引言 2019年11月6号LightBend公司发布了AKKA 2.6版本,带来了类型安全的actor,新的Akka Cluster底层通信设施——Artery,带来了更好的稳定性,使用Jackson ...

  4. Akka Essentials - 2

    Actors Defining an actor class MyActor extends Actor { def receive = { } } In Scala, the receive blo ...

  5. 一图看懂Actor Typed

    引言 朋友看罢我之前整理的<Akka Typed 官方文档之随手记>,一人用了诗歌<长城长>作为回赠,另一人则要求推出简化版本.于是抽空整理了几张思维导图,并且用了一些不太恰当 ...

  6. CQRS与Event Sourcing之浅见

    引言 DDD是近年软件设计的热门.CQRS与Event Sourcing作为实施DDD的一种选择,也逐步进入人们的视野.围绕这两个主题,软件开发的大咖[Martin Fowler].[Greg You ...

  7. Lagom 官方文档之随手记

    引言 Lagom是出品Akka的Lightbend公司推出的一个微服务框架,目前最新版本为1.6.2.Lagom一词出自瑞典语,意为"适量". https://www.lagomf ...

  8. Akka(14): 持久化模式:PersistentActor

    Akka程序的特点之一就是高弹性或者强韧性(resilient)的,因为Actor具有自我修复的能力.当Actor模式的程序出现中断情况如:系统崩溃.人为终结等,系统在重启后有关Actor可以恢复之前 ...

  9. akka-typed(4) - EventSourcedBehavior in action

    前面提到过,akka-typed中较重要的改变是加入了EventSourcedBehavior.也就是说增加了一种专门负责EventSource模式的actor, 最终和其它种类的actor一道可以完 ...

随机推荐

  1. Node.js中模块加载机制

    1.模块查找规则-当模块拥有路径但没有后缀时:(require(‘./find’)) require方法根据模块路径查找模块,如果是完整路径,直接引入模块: 如果模块后缀省略,先找同名JS文件,再找同 ...

  2. ztree根据参数动态控制是否显示复选框/单选框(静态JSON数据)

    本文不再更新,可能存在内容过时的情况,实时更新请访问原地址:ztree根据参数动态控制是否显示复选框/单选框(静态JSON数据): 现有全省各地区静态JSON数据,现在想通过Url参数,动态控制是否显 ...

  3. 【Scala】用实例弄清楚scala几种函数的定义和特点

    文章目录 作为参数的函数 匿名函数 柯里化函数(currying) 闭包函数 作为参数的函数 scala> val a1 = Array(1,2,3,4) //这是一个数组 a1: Array[ ...

  4. js console一些常用的功能

    前言 很多时候我们在调试的时候经常用console.log,我感觉其实一个就够了,但是有时候你不可能写一步就去调试下,所以呢,经常用几个console.log,有时候挺难找的,后面翻了翻console ...

  5. AndroidStudio3.6升级后的坑-apk打包

    前段时间尝试了最新版的AndroidStudio3.6,整体来说gradle调试和自带的虚拟机相比较历史版本有了更香的体验. 刚好有个新项目,就直接使用最新版了,这次新版的升级除了保持原有的界面风格, ...

  6. 🏃‍♀️点亮你的Vue技术栈,万字Nuxt.js实践笔记来了~

    前言 作为一位 Vuer(vue开发者),如果还不会这个框架,那么你的 Vue 技术栈还没被点亮. Nuxt.js 是什么 Nuxt.js 官方介绍: Nuxt.js 是一个基于 Vue.js 的通用 ...

  7. 【数论基础】素数判定和Miller Rabin算法

    判断正整数p是否是素数 方法一 朴素的判定   

  8. ES6-10笔记(二)

    class类 Javascript是一种基于对象(object-based)的语言,你遇到的所有东西几乎都是对象.但是,它又不是一种真正的面向对象编程(OOP)语言,因为它的语法中没有class(类) ...

  9. Gitlab 安装、升级、备份、恢复、汉化等

    一.Gitlab安装 1. 基于yum方式安装Gitlab 安装步骤如下 (1)配置yum源 # vim /etc/yum.repos.d/gitlab-ce.repo (2)复制如下内容并保存(注意 ...

  10. React组件proptypes, ref

    一.使用props.children访问嵌套数据 import React from 'react'; class Button extends React.Component { render () ...