Actors (Scala)

Defining an Actor class

Actor classes are implemented by extending the Actor class and implementing the receive method. The receive method should define a series of case statements (which has the type PartialFunction[Any, Unit]) that defines which messages your Actor can handle, using standard Scala pattern matching, along with the implementation of how the messages should be processed.

Here is an example:

import akka.actor.Actor
import akka.event.EventHandler

class MyActor extends Actor {
  def receive = {
    case "test" => EventHandler.info(this, "received test")
    case _ => EventHandler.info(this, "received unknown message")
  }
}

Please note that the Akka Actor receive message loop is exhaustive, which is different compared to Erlang and Scala Actors. This means that you need to provide a pattern match for all messages that it can accept and if you want to be able to handle unknown messages then you need to have a default case as in the example above.

Creating Actors

val myActor = Actor.actorOf[MyActor]
myActor.start()

Normally you would want to import the actorOf method like this:

import akka.actor.Actor._

val myActor = actorOf[MyActor]

To avoid prefixing it with Actor every time you use it.

You can also start it in the same statement:

val myActor = actorOf[MyActor].start()

The call to actorOf returns an instance of ActorRef. This is a handle to the Actor instance which you can use to interact with the Actor. The ActorRef is immutable and has a one to one relationship with the Actor it represents. The ActorRef is also serializable and network-aware. This means that you can serialize it, send it over the wire and use it on a remote host and it will still be representing the same Actor on the original node, across the network.

Creating Actors with non-default constructor

If your Actor has a constructor that takes parameters then you can’t create it using actorOf[TYPE]. Instead you can use a variant of actorOf that takes a call-by-name block in which you can create the Actor in any way you like.

Here is an example:

val a = actorOf(new MyActor(..)).start() // allows passing in arguments into the MyActor constructor

Creating Actors using anonymous classes

When spawning actors for specific sub-tasks from within an actor, it may be convenient to include the code to be executed directly in place, using an anonymous class:

def receive = {
  case m: DoIt =>
    actorOf(new Actor {
      def receive = {
        case DoIt(msg) =>
          val replyMsg = doSomeDangerousWork(msg)
          self.reply(replyMsg)
          self.stop()
      }
      def doSomeDangerousWork(msg: Message) = { ... }
    }).start() ! m
}

Running a block of code asynchronously

Here we create a light-weight actor-based thread, that can be used to spawn off a task. Code blocks spawned up like this are always implicitly started, shut down and made eligible for garbage collection. The actor that is created “under the hood” is not reachable from the outside and there is no way of sending messages to it. It being an actor is only an implementation detail. It will only run the block in an event-based thread and exit once the block has run to completion.

spawn {
  ... // do stuff
}

Start Hook

Right after starting the actor, its preStart method is invoked. This is guaranteed to happen before the first message from external sources is queued to the actor’s mailbox.

override def preStart {
  // e.g. send initial message to self
  self ! GetMeStarted
  // or do any other stuff, e.g. registering with other actors
  someService ! Register(self)
}

Restart Hooks

A supervised actor, i.e. one which is linked to another actor with a fault handling strategy, will be restarted in case an exception is thrown while processing a message. This restart involves four of the hooks mentioned above:

1. The old actor is informed by calling preRestart with the exception which caused the restart and the message which triggered that exception; the latter may be None if the restart was not caused by processing a message, e.g. when a supervisor does not trap the exception and is restarted in turn by its supervisor. This method is the best place for cleaning up, preparing hand-over to the fresh actor instance, etc.
2. The old actor’s freshInstance factory method is invoked, which may optionally produce the new actor instance which will replace this actor. If this method returns None or throws an exception, the initial factory from the Actor.actorOf call is used to produce the fresh instance.
3. The new actor’s preStart method is invoked, just as in the normal start-up case.
4. The new actor’s postRestart method is called with the exception which caused the restart.

An actor restart replaces only the actual actor object; the contents of the mailbox and the hotswap stack are unaffected by the restart, so processing of messages will resume after the postRestart hook returns. Any message sent to an actor while it is being restarted will be queued to its mailbox as usual.

Stop Hook

After stopping an actor, its postStop hook is called, which may be used e.g. for deregistering this actor from other services. This hook is guaranteed to run after message queuing has been disabled for this actor, i.e. sending messages would fail with an IllegalActorStateException.

Fire-forget

This is the preferred way of sending messages. No blocking waiting for a message. This gives the best concurrency and scalability characteristics.

actor ! "Hello"

If invoked from within an Actor, then the sending actor reference will be implicitly passed along with the message and available to the receiving Actor in its channel: UntypedChannel member field. The target actor can use this to reply to the original sender, e.g. by using the self.reply(message: Any) method.

If invoked from an instance that is not an Actor there will be no implicit sender passed along with the message and you will get an IllegalActorStateException when calling self.reply(...).

Send-And-Receive-Future

Using ? will send a message to the receiving Actor asynchronously and will return a Future:

val future = actor ? "Hello"

The receiving actor should reply to this message, which will complete the future with the reply message as value; if the actor throws an exception while processing the invocation, this exception will also complete the future. If the actor does not complete the future, it will expire after the timeout period, which is taken from one of the following three locations in order of precedence:

1. explicitly given timeout as in actor.?("hello")(timeout = 12 millis)

2. implicit argument of type Actor.Timeout, e.g.

   implicit val timeout = Actor.Timeout(12 millis)
   val future = actor ? "hello"
   

3. default timeout from akka.conf

See Futures (Scala) for more information on how to await or query a future.

Send-And-Receive-Eventually

The future returned from the ? method can conveniently be passed around or chained with further processing steps, but sometimes you just need the value, even if that entails waiting for it (but keep in mind that waiting inside an actor is prone to dead-locks, e.g. if obtaining the result depends on processing another message on this actor).

For this purpose, there is the method Future.as[T] which waits until either the future is completed or its timeout expires, whichever comes first. The result is then inspected and returned as Some[T] if it was normally completed and the answer’s runtime type matches the desired type; if the future contains an exception or the value cannot be cast to the desired type, it will throw the exception or a ClassCastException (if you want to get None in the latter case, use Future.asSilently[T]). In case of a timeout, None is returned.

(actor ? msg).as[String] match {
  case Some(answer) => ...
  case None         => ...
}

val resultOption = (actor ? msg).as[String]
if (resultOption.isDefined) ... else ...

for (x <- (actor ? msg).as[Int]) yield { 2 * x }

Forward message

You can forward a message from one actor to another. This means that the original sender address/reference is maintained even though the message is going through a ‘mediator’. This can be useful when writing actors that work as routers, load-balancers, replicators etc.

actor.forward(message)

Reply using the channel

If you want to have a handle to an object to whom you can reply to the message, you can use the Channel abstraction. Simply call self.channel and then you can forward that to others, store it away or otherwise until you want to reply, which you do by channel ! response:

case request =>
    val result = process(request)
    self.channel ! result       // will throw an exception if there is no sender information
    self.channel tryTell result // will return Boolean whether reply succeeded

The Channel trait is contravariant in the expected message type. Since self.channel is subtype of Channel[Any], you may specialise your return channel to allow the compiler to check your replies:

class MyActor extends Actor {
  def doIt(channel: Channel[String], x: Any) = { channel ! x.toString }
  def receive = {
    case x => doIt(self.channel, x)
  }
}

case request =>
    friend forward self.channel

We recommend that you as first choice use the channel abstraction instead of the other ways described in the following sections.

Reply using the reply and reply_? methods

If you want to send a message back to the original sender of the message you just received then you can use the reply(..) method.

case request =>
  val result = process(request)
  self.reply(result)

In this case the result will be send back to the Actor that sent the request.

The reply method throws an IllegalStateException if unable to determine what to reply to, e.g. the sender is not an actor. You can also use the more forgiving tryReply method which returns true if reply was sent, and false if unable to determine what to reply to.

case request =>
  val result = process(request)
  if (self.tryReply(result)) ...// success
  else ... // handle failure

Summary of reply semantics and options

• self.reply(...) can be used to reply to an Actor or a Future from within an actor; the current actor will be passed as reply channel if the current channel supports this.
• self.channel is a reference providing an abstraction for the reply channel; this reference may be passed to other actors or used by non-actor code.

channel match {
  case ref : ActorRef => ...
  case f : ActorCompletableFuture => ...
}

Upgrade

Akka supports hotswapping the Actor’s message loop (e.g. its implementation) at runtime. There are two ways you can do that:

• Send a HotSwap message to the Actor.
• Invoke the become method from within the Actor.

Both of these takes a ActorRef => PartialFunction[Any, Unit] that implements the new message handler. The hotswapped code is kept in a Stack which can be pushed and popped.

To hotswap the Actor body using the HotSwap message:

actor ! HotSwap( self => {
  case message => self.reply("hotswapped body")
})

Using the HotSwap message for hotswapping has its limitations. You can not replace it with any code that uses the Actor’s self reference. If you need to do that the the become method is better.

To hotswap the Actor using become:

def angry: Receive = {
  case "foo" => self reply "I am already angry!!!"
  case "bar" => become(happy)
}

def happy: Receive = {
  case "bar" => self reply "I am already happy :-)"
  case "foo" => become(angry)
}

def receive = {
  case "foo" => become(angry)
  case "bar" => become(happy)
}

The become method is useful for many different things, but a particular nice example of it is in example where it is used to implement a Finite State Machine (FSM): Dining Hakkers

Here is another little cute example of become and unbecome in action:

case object Swap
class Swapper extends Actor {
 def receive = {
   case Swap =>
     println("Hi")
     become {
       case Swap =>
         println("Ho")
         unbecome() // resets the latest 'become' (just for fun)
     }
 }
}

val swap = actorOf[Swapper].start()

swap ! Swap // prints Hi
swap ! Swap // prints Ho
swap ! Swap // prints Hi
swap ! Swap // prints Ho
swap ! Swap // prints Hi
swap ! Swap // prints Ho

Downgrade

Since the hotswapped code is pushed to a Stack you can downgrade the code as well. There are two ways you can do that:

• Send the Actor a RevertHotswap message
• Invoke the unbecome method from within the Actor.

Both of these will pop the Stack and replace the Actor’s implementation with the PartialFunction[Any, Unit] that is at the top of the Stack.

Revert the Actor body using the RevertHotSwap message:

actor ! RevertHotSwap

Revert the Actor body using the unbecome method:

def receive: Receive = {
  case "revert" => unbecome()
}