Scala Akka and Protocol Buffers

Question

I'm playing around with Akka's remoting and serialization facilities and want to understand a few things to get started. I've read over the documentation on serialization here:

http://doc.akka.io/docs/akka/2.3.4/scala/serialization.html

According to the documentation, it seems that it would be enough to just supply these things in my application.conf, under:

akka.actor {

 serializers {
   java = "akka.serialization.JavaSerializer"
   proto = "akka.remote.serialization.ProtobufSerializer"
}

 serialization-bindings {
    "com.mycompany.messages.MyMessage" = proto
 }
}

And let's assume I have a case class under that package, such as:

package com.mycompany.messages;
case class MyMessage(name: String, year: Int)

And then in my actors, I can simly do something like this:

class ClientActor extends Actor {

     def receive = {
        case x: MyMessage => ...
     }
}

Would this configuration be enough, or would I need to do something more? I've looked at an external serializer mentioned from the documentation here: https://github.com/romix/akka-protostuff-serialization

This looks really promising, but I was looking for something standard that comes out of the box from Akka.

I'm also looking into testing for message version compatibility. Let's say Actor A talks to Actor B with MessageX

MessageX initially might contain fields like this:

a: String, b: String, c: String

Now let's say Actor B upgrades its version of Message X, lets call it Message X +1

Message X +1 now includes another field, like so:

a: String, b: String, c: String, d: String

But Actor A is still sending the older version of the message, just simply Message X... would Actor B still know how to deserialize the old message?

Thanks for the help.

Answer 1

The Protobuf serializer can only serialize Protobuf messages. So for things to work as you want you need to make MyMessage a protobuf message.

Answer 2

Just an additional information on your versioning concerns;

Protobuf can serialise/ deserialize different versions of the same message types. You have to sustain the already present fields' indexes and add the additional new ones accordingly, in your proto file descriptor.