// GET api/values/5/6
public async Task <string> Get(int arg1, int arg2)
{
ServiceUriBuilder actorServiceUriBuilder = new ServiceUriBuilder(CalculatorActorService);
// This only creates a proxy object, it does not activate an actor or invoke any methods yet.
ICalculatorActor calculatorActor = ActorProxy.Create <ICalculatorActor>(new ActorId(0), actorServiceUriBuilder.ToUri());
// This will invoke a method on the actor. If an actor with the given ID does not exist, it will be activated by this method call.
var result = await calculatorActor.GetSumAsync(arg1, arg2);
return(result.ToString());
}
private static double Add(ICalculatorActor calculator, double x, double y)
{
return calculator.Add(x, y).Result;
}
private static double Add(ICalculatorActor calculator, double x, double y)
{
return(calculator.Add(x, y).Result);
}
static void Main(string[] args)
{
// Note: In samples below, .Result is used to force eager evaluation, however,
// async / await patterns could be used as well.
// Creating and starting server proxy for an actor is easy.
// Returned reference can be used to stop server.
var actorServer = ActorServerProxy.Create <CalculatorActor>("tcp://*:4632");
// To create client proxy and connect to the server just pass an interface which describes
// what method actor supports. Returned object implements given interface,
// so one can use one interface to call local and remote actors.
// Returned task is signalled when actor successfully connects to the server.
ICalculatorActor calculator = ActorClientProxy.CreateActor <ICalculatorActor>("tcp://localhost:4632").Result;
{
// Task will throw an exception if it won't be able to connect to the server.
try
{
ICalculatorActor wrongAddress = ActorClientProxy.CreateActor <ICalculatorActor>("tcp://127.0.0.1:45662").Result;
}
catch (AggregateException exc)
{
Console.WriteLine();
Console.WriteLine("Could not connect to server: " + exc.InnerException.Message);
}
}
{
// Messages can be sent just by calling methods.
// Returned tasks will be notified when network response will be received
// from server.
Console.WriteLine();
Console.WriteLine("5 + 4 = " + calculator.Add(5, 4).Result);
}
{
// Primitive types are serialized automatically, if custom classes are to be used
// they should be implemented with ProtoContract and ProtoMember attributes,
// so that Protobuf-net will know how to serialize them.
// Complex objects can be used both as parameters and results.
var rectInfo = calculator.CalculateRectangle(new Rectangle {
A = 5, B = 6
}).Result;
Console.WriteLine();
Console.WriteLine("Recangle 5 x 6 has field equal = {0} and perimeter = {1}",
rectInfo.Field, rectInfo.Perimeter);
}
{
// Because actor called on the server side is always the same instance
// (until client has been reconnected, which gives no warranty on state of server actor)
// it can be used to persist state just as local actor can.
// Note: You don't need to wait for task to finish, messages are invoked on the server
// in the order they were called by client proxy.
calculator.PushNumber(12.0);
calculator.PushNumber(15.0);
var x = calculator.PopNumber();
var y = calculator.PopNumber();
Console.WriteLine();
Console.WriteLine("15 - 12 = " + calculator.Subtract(x.Result, y.Result).Result);
}
{
// Any errors will be serialized and sent back to the client proxy.
// No number on stack, should throw.
// Note, that Task aggregates exception, however, when using async/await pattern
// they will be automatically unwrapped.
var error = calculator.PopNumber();
try
{
error.Wait();
}
catch (AggregateException exc)
{
Console.WriteLine();
Console.WriteLine("Error while popping from stack: " + exc.InnerException.Message);
}
}
{
// Remote actor still acts as an actor when it comes to message execution, which is serial.
// This means, as with local actors, you can queue messages safetly, without worrying
// about order execution and multithreading errors.
Console.WriteLine();
Console.WriteLine("Pinging...");
var p1 = calculator.Ping();
var p2 = calculator.Ping();
var p3 = calculator.Ping();
Task.WhenAll(p1, p2, p3).Wait();
Console.WriteLine("Pinging complete");
}
{
// Arrays and even enumerables can be used as parameters
var mean = calculator.Mean(new[] { 5.0, 4.0, 3.0, 1.0, 14.0 }).Result;
var mean2 = calculator.MeanEnum((new List <double> {
5.0, 4.0, 3.0, 1.0, 14.0
}).AsEnumerable()).Result;
Console.WriteLine();
Console.WriteLine("Mean of [1, 3, 4, 5, 14] equals " + mean);
Console.WriteLine("Double check: " + mean2);
}
{
// Because server actors behave just like local ones, you can
// choose not to use actor context on called methods
// and they will run synchronously, so custom synchronization
// (or no synchronization) can be applied.
// Note, that this might be an unexpected behaviour
// for users, so this functionality should be used
// with care.
// Long processing methods should be run on
// task pool, if actor context won't be used, otherwise
// they will block network communication for all
// connected clients.
Console.WriteLine();
Console.WriteLine("Pinging without actor context...");
var p1 = calculator.PingAsync();
var p2 = calculator.PingAsync();
var p3 = calculator.PingAsync();
Task.WhenAll(p1, p2, p3).Wait();
Console.WriteLine("Pinging complete");
}
{
// Many client actors may connect to one server.
// Their method calls will be scheduled as they would originate
// from single client, which means, serial execution will be enforced
// automatically.
Console.WriteLine();
Console.WriteLine("Pinging with 2 clients");
var calculator2 = ActorClientProxy.CreateActor <ICalculatorActor>("tcp://localhost:4632").Result;
var p1 = calculator.Ping();
var p2 = calculator2.Ping();
Task.WhenAll(p1, p2).Wait();
Console.WriteLine("Pinging done");
// Because 'simpler' method for creating an actor was used,
// reference to an actor interface was received, instead of to wrapper object.
// If wrapper object with control method is needed, returned object
// can be cast to IActorClientProxy or IActorClientProxy<Actor Interface>
((IActorClientProxy)calculator2).Close();
}
{
// Created actor proxy is, from the user's perspective just an interface,
// which can also be implemented by local actor, therefore,
// where the code executes is fully transparent for other methods.
ICalculatorActor localCalc = new CalculatorActor();
Console.WriteLine();
Console.WriteLine("Remote: 4 + 8 = " + Add(calculator, 4, 8));
Console.WriteLine("Local: 4 + 8 = " + Add(localCalc, 4, 8));
}
// Properties with IObservable<T> types can be used to
// broadcast messages to clients without receiving any
// requests first.
Console.WriteLine();
Console.WriteLine("Random generator: ");
var rng = calculator.Rng.Subscribe(r => Console.WriteLine(r));
Thread.Sleep(4000);
rng.Dispose();
Console.WriteLine();
calculator.Messages.Subscribe(m =>
{
Console.WriteLine("A: " + m.X + "; B: " + m.Y + ". Area: " + m.Info.Field);
});
Thread.Sleep(3000);
// When more control over proxy is needed, CreateProxy method can be used.
// This will return the same object as CreateActor method would, but returned
// type will be of a wrapper type.
var proxyCalc = ActorClientProxy.CreateProxy <ICalculatorActor>("tcp://localhost:4632").Result;
// Actual actor implementation can be accessed through .Actor property,
// or by simply casting whole proxy to actor interface.
// Both method will return the same object.
ICalculatorActor proxyCalcActor1 = proxyCalc.Actor;
ICalculatorActor proxyCalcActor2 = (ICalculatorActor)proxyCalc;
Console.WriteLine();
Console.WriteLine("Two methods of accessing proxy actor return the same object - " +
object.ReferenceEquals(proxyCalcActor1, proxyCalcActor2));
actorServer.Stop();
((IActorClientProxy)calculator).Close();
Console.WriteLine();
Console.WriteLine("Sample finished");
Console.Read();
}
