Writing complex actors about kotlinx.coroutines

[totemcaf]

We are starting a new project and Actors model fits perfectly. I wonder that using Kotlin Actors requires to mark all code with @ObsoleteCoroutinesApi.

Do we have any progress on the actors implementation replacement?

[nikiJava]

Hi :) Are there any updates about new actors API?

[adam-arold]

What is your suggestion for those who want to use actors in the meantime (until actors come out of the experimental stage)? I'm currently working on a multiplatform library, so using coroutines to enable concurrency is very convenient, but I'm a bit puzzled about the best way (currently) to go about implementing something like a complex actor in your examples (A class which has suspending functions which work on shared mutable state).

[ZakTaccardi]

Once complex actors are implemented, what happens to the existing actor? Is it going to be removed?

Would it be possible to release the existing actor as a separate artifact for legacy support? I'm asking because I would like to get my company to adopt actor usage (I think it's an awesome way to build/update state), but the @ObsoleteApi causes a bit of hesitancy for us to adopt within our company.

One approach we've considered is completely re-implementing the actor ourselves, but I think that risks some subtle bugs around concurrency and exception handling

[elizarov]

The key design consideration here is how a typical actor's code is going to look like. It is not clear in your example. The goal should be to minimize the amount of boilerplate that users of this library code are forced to write. I was thinking about something along the following lines:

abstract class Actor<T>(scope: ActorScope<T>): ActorScope<T> by scope {
    abstract suspend fun onReceive(msg: T)
}

An abstract class lets us share some common state among all actor implementations (in this case, the fact that ActorScope is provided to all the functions defined inside an actor impl). It also makes sense to share "for messages" loop, as every typical actor is going to have one, and leave only onReceive for override.

With this bases class, the specific actor implementation looks like this:

class MyActor(scope: ActorScope<MyMsg>): Actor<MyMsg>(scope) {
    // your private state here
    override suspend fun onReceive(msg: MyMsg) {
        // ... your code here ...
    }
}

There is still some boiler-plate (MyMsg and scope are mentioned twice), but it is not clear how to further reduce it.

Now, you can define a generic actorOf function to start any kind of actor:

fun <T> actorOf(
    context: CoroutineContext,
    init: (ActorScope<T>) -> Actor<T>
): ActorJob<T> = actor<T>(context) {
        val instance = init(this@actor)
        for (msg in channel) instance.onReceive(msg)
    }

Using this function you can start your custom actor with a natural-looking and easy-to-read invocation actorOf(CommonPool, ::MyActor) with ability to further customize construction of your actor impl by providing an explicit lambda for an init block.

[elizarov]

Let me record here a draft design that I current have with respect to complex actors. The idea is to radically reduce boiler-plate that is required for writing actors that accept multiple message types by providing a handy facility that totally avoids the need to write sealed class for messages by following this pattern:

class MyActor : Actor() {
    // actor state is here, must be private

    init {
        // optional state initialization code
    }

    // public actor operation without result
    suspend fun operationA(someParams...) = act {
        // logic here
    }

    // public actor operation with result
    suspend fun operationB(otherParams...): Result = actAndReply {
        // logic here
    }

    private fun helper() { /* regular code, can access state */ }
}

The idea is that all public functions defined on actor must be suspend and must invoke a special function (tentatively called act here) that captures "message parameters" (function parameters) and stores the message in the actor's inbox for processing. When response to the caller is needed, then a separate function (tentatively called actAndReply here) is used.

We can have IDE inspections to help avoid "sharing" pitfalls to verify that all of the actor's state is private and all the public functions are properly written using act pattern. IDE inspection can also help to ensure that only immutable types are passed in/out of actor or the proper defensive copies are made on mutable data (the latter is hard to ensure, though, without complicated reference capability annotations akin to the ones used in Pony language).

The reference to MyActor instance serves as a type-safe reference to an actor and should implement a Job interface. The actor is started lazily on the first invocation of any of its public (acting) functions or by explicitly invoking start(). Lazy start feature prevents races when systems of communicating actors are written like this:

class MySystem { 
    private val actorA = object : Actor() {
         // somewhere sends message to actorB
    }

    private val actorB = object : Actor() {
        // somewhere sends message to actorA
    }

    suspend fun facadeOperation() { actorA.doSomethig() }
}

We shall also consider changing default to CoroutineStart.LAZY for a simple actor { ... } builder, too.

[pablichjenkov]

Eagerly awaiting!

[elizarov]

I'd do something along these lines:

class MyActor {
    // your private state here
    suspend fun onReceive(msg: MyMsg) {
        // ... your code here ...
    }
}

fun myActorJob(): ActorJob<MyMsg> = actor(CommonPool) {
    with(MyActor()) {
        for (msg in channel) onReceive(msg)
    }
}

You can also pass actor's channel to MyActor constructor if it needs access to it.

[elizarov]

Also, if we are starting on a path of generic actor-based programming, then we are inevitably going to be compared with other actor-based frameworks, including comparisons on the basis of performance. It does not mean that we have to worry about performance up-front, but we have to keep our designs, at least, optimizable in the future.

That is my concern for overridable invoke/run method and other potential design decisions. Let me elaborate. A generic actor can, for example, start multiple concurrent actors to process a single mailbox concurrently if the actor's logic is stateless by itself. However, a stateful actor is must typically process its mailbox in serial fashion, e.g. it is a single consumer of its mailbox.

Current implementation of Channel is multi-consumer/multi-producer queue which makes it widely-applicable (it can do both use-cases), but it inevitably costs some performance. One might implement an optimized version of channel for single-producer case to support stateful-actor use-case with better performance, but then we have to ensure that it is not possible for an actor to accidentally launch additional coroutines to read from its mailbox (to make the design less error-prone and more fool-proof). It means that we should actually hide actor's ReceiveChannel from the actor itself.

[pigbayspy]

So will actor api change in future?

[drinkthestars]

what is the latest on the timeline/plans for actor api updates?

[elizarov]

@fvasco Yes, when you ask and actor and want a result back the proper design would be to have a suspend fun with a normal (non-deferred) Result. However, please note that this whole ask & wait pattern is an anti-pattern in actor-based systems, since it limits scalability. Well-designed actor-based system do not work that way.

Internally, actors are still implemented on top of channels.

[gortiz]

The more I use actors, the more necessary I find frameworks specially created to deal with the common actor patterns. I don't think we should create another actor framework but to make easier to integrate the amazing kotlin functionalities with the frameworks we already have.

And there are already two mature frameworks on the JVM world: Akka and Quasar. I didn't use Quasar, but as it uses their own coroutines, they seem to be closer to the Kotlin aproach. I have more experience with Akka, where actors suspendable actors should be useful when dealing with blocking code (because of IO or high computation costs).

My plan right now is to develop a Akka Typed Kotlin DSL that makes easier and nicer to call the Scala/Java APIs from Kotlin. That is not very ambitious and in fact it can be done with a very small project. It is very easy to create suspendable actors that stage messages while the actor is suspended. The main problem there is that it may be not very efficient. What I would really like is to create an Akka Actor System that is able to execute suspendable actors natively, but:

1. Doesn't seem easy.
2. Is more related to Akka itself than Kotlin

By the way, I want to emphasize that Kotlin actors are fantastic for most cases. Frameworks like Akka are needed once you want do more complex things like scheduling (although it wouldn't be difficult to implement in Kotlin actors), clustering (using location transparency) or persistence (for event source systems, for example). I'm thinking on migrate the application I wrote using Kotlin actors to Akka just because of the last two features, but Kotlin actors allow me to rapidly create the first version of it and most importantly it was very easy to think on concurrency and error management on a higher level and I'm sure that by using Kolin actors the application is more resilent that it would be if it was implemented on other non-actor based frameworks .

[pacher]

Currently it is possible for actor to send a message to yourself because ActorScope exposes Channel and not only ReceiveChannel (being fully aware of potential deadlock when channel is at capacity).
It would be awesome if you keep this in mind and make it possible with the new design as well.

[penn5]

I wrote an implementation: https://gitlab.com/snippets/1957446
For one-way requests, it has an almost identical performance to the current actor builder (I compared it against the sample code in https://kotlinlang.org/docs/reference/coroutines/shared-mutable-state-and-concurrency.html with the same optimisations, dispatchers, etc on both sides
For two-way requests, it is very marginally faster than the current actor implementation. However, it is much better than the current implementation because you can do the entire request and response in one line in the main(). It is also much easier to implement it in the actor itself, as there is no when clause. It also supports running with concurrency, when you have no mutability problems (but then why not just use normal functions?). Error handling should be fine too - errors pass through. I could probably make a cancelOnError parameter too.

However, this version makes it very easy to leak the actor scope. So I made https://gitlab.com/snippets/1957455 which is no more restrictive, but uses subclassing to simplify the boilerplate and syntax, and also to prevent leaking, due to the protected fields. However, this can't be used when there is another superclass, because multiple inheritance is forbidden. It also doesn't work with kotlinx.serialization.

I'm putting this here in case I'm going about it completely the wrong way, in which case I won't make a PR.

EDIT: I ran my benchmarks starting all the addition operations in parallel, as follows:

...
coroutineScope {
    repeat(1000000) {
        launch {
            ...
        }
    }
}
...

Upon removing the launch block from the benchmark (for 2-way communication), my solution accelerated to about double the original speed, whereas the current actor slowed down to 0.2 of the original speed, and to 0.5 for 1-way communication. All the benchmarks of my code were 2-way, because it's essential an RPC so making it 1-way doesn't make sense.

EDIT 2: In the first 2 iterations I forgot to make the actual actions suspending. Here is a documented, fixed version: https://gitlab.com/snippets/1957572

[asad-awadia]

Regardless of how the actor piece is brought in- it would be good to have the ability to add supervision strategies like being able to restart on exception

[renatomrcosta]

Hi @elizarov, it's 2024. What about the new API?

Dude doesn't even work for Jetbrains anymore mate

[gortiz]

I have been working with Kotlin actors for months and I think they have very low boilerplate in comparison with untyped Akka actors (where usually you need to extend a class if using the Java dsl, the only one I can use from Kotlin). But with the introduction of Akka Typed, the boilerplate was reduced a lot and they are almost as compact as Kotlin actors.

When I found Kotlin actors problematic is when you try to go beyond the most trivial use cases. I understand that I may have a bias towards Akka because I'm most use of it, but I find quite complex to implement in Kotlin some common actor patterns like send a message to yourself, create a ticker/scheduler, create an adaptor or receive messages from different type hierarchies. Guys from Akka have been trying to solve this problems for 10 years after three attempts, it seems that they have found a nice API that is at the same quite time safe at compile, easy to use and quite compact.

Simple Akka Typed actors are written as a function in a way that is almost equal to current Kotlin actors and the most complex patterns are delegated on a object call ActorContext that is obtained as a parameter.

Akka Typed is not perfect either. There are some use cases that are not easy to implement. For example the common case on which you want to stage all messages received by an actor until some message arrives. On Akka Typed you need to create a behavior to do that. On Kotlin we can implement the ask pattern using suspending functions, suspending the actor until the answer is found.

The Akka Typed has some flaws that make it not completely secure at compile time. For example some functions on ActorContext can be called from any thread but most of them can only be called from the actor thread. That is clearly explain on the javadoc of each method, but it is easy to forget. I have been working on a small Akka Typed Kotlin DSL where can be called from any thread are normal methods, but the others are suspending methods that are executed on the actor thread by using the ask pattern.

I'm not saying Kotlin should copy the Akka Typed API, just learn from it and improve it. There are still problems that are difficult to solve by Akka where Kotlin has a big advantage.

The way to use the actor proposed by Elizarov has its advantages, but I would prefer to have an ActorRef/Channel to send messages and then create extensions functions that call act/tell or actWithReply/ask and being able to define actors (even complex actors!) with behaviors/functions instead of creating classes.

[elizarov]

Waiting until the actor has processed the message is sometimes useful, but is always dangerous and easily leads to deadlocks. I would not recommend to support it.

[dimsuz]

Hmm, baking this framework into the language seems as a bit weird approach to take.

[Kshitij09-sc]

Based on the description of Stateful Actor and the idea of reducing boilerplate, I would like to propose following abstraction (inspired by orbit-mvi's Container). Let me know if anything might go wrong with this implementation.
cc: @elizarov @fvasco

interface ActorContext

abstract class StatefulActor(
    private val parentScope: CoroutineScope,
    bufferCapacity: Int = 64
) {
    private val dispatchChannel = Channel<suspend ActorContext.() -> Unit>(bufferCapacity)
    private val initialized = AtomicBoolean(false)

    private fun initializeIfNeeded() {
        if(initialized.compareAndSet(false, true)) {
            parentScope.launch {
                for (action in dispatchChannel) {
                    launch { context.action() }
                }
            }
        }
    }

    fun shutdown() {
        dispatchChannel.close()
    }

    protected fun act(action: suspend ActorContext.() -> Unit) {
        initializeIfNeeded()
        dispatchChannel.trySendBlocking(action)
    }

    abstract val context: ActorContext
}

Here's one actor implementation using this API

class MyActor(scope: CoroutineScope) : StatefulActor(scope), ActorContext {
    private var clickCounter = 0

    override val context: ActorContext get() = this

    fun increment() = act { clickCounter++ }

    fun printCount() = act { println(clickCounter) }
}

I found it surprisingly fast as well (Unless I'm making some mistake measuring the performance).

Playground with above snippets and measuring setup: https://pl.kotl.in/RHPpCbcUT

[elizarov]

Makes sense. I'd keep an extended version. Maybe rename actor.invoke() to actor.run() (similar to Thread's run method). I'm not sure that an actor base class should be called ActorBase. I'd either keep it an Actor to be short or name it AbstractActor.

[elizarov]

@gortiz All valid points. However, we do not plan to turn kotlinx.coroutines into an actor-based framework with actor refs, actor systems, actor contexts, replies, etc. In terms of kotlinx.coroutines actor is just a code organization pattern -- a coroutine plus an incoming channel. Nothing more. Something more than that has to be living in a separate library.

[elizarov]

Not yet. We are busy with Kotlin Flow for now, which is a higher-priority for us to finish.

[gortiz]

But by using Kotlin's foundation of async programming it would be easier to implement the same in Kotlin than it was for Akka in Scala. Kotlin has by default all primitives for that (channels, queues, back pressures, contexts(coroutines), supervisors).

I think we are not talking about the same thing. One thing is to create a toy project that implement this kinds of things on a biased way and another is to have a product that can be used by different people of different organizations. Of course, it would be easier to create a new Akka on top of Kotlin Coroutines than to create it from zero, but it is still has a huge cost but from my point of view, very little benefits (which are mainly multi platform compilation and to be written in Kotlin instead of Scala because we like the former more).
IMHO it is not worth it and it doesn't seem that Jetbrains is interested on that.

I'd also would like to have something similar as Akka, but due lack of Scala knowledge in our team and unknown future of Scala itself we are avoiding it.

Akka is used on tons of real world open source projects including Apache Flink (for sure) and Spark (if I remember correctly) and commercial (there is a list here)and is developed by a historic company. To be honest, I would trust Akka and Lightbend more than an actor framework created by a individual contributors and/or small companies.

I don't love Scala and I'm not an expert on it, but the documentation is quite good, almost all examples are in Java and the people of Lightbend is very open to help developers even if they don't pay for support.

The actor paradigm is not that popular and between Akka and Quasar there are already too many actor frameworks on the JVM and I'm afraid that there will be even more once Project Loom is released. In my opinion, if there is no very important technical reason to create a new framework, I wouldn't like to create a new one to compete with Akka, that has a big enough community, has been proved on production, is evolving quite fast and, very important, is compatible with the language I love: Kotlin

I think my point is clear and I wouldn't like to tangle even more this thread (which I think I personally shifted too much from their original topic)

[ysm-coder]

I see where Elizarov is coming from, I also think that the pony way for Actors is really awesome but they have the actor support in the language not just in a library.

suspend fun operationA(someParams...) = act {
        // logic here
    }

Would look much better and clearer if it would be

act operationA(someParams...) = logic here

I asume that new keywords are costly and won't get added into the language that easy. But otherwise also the "Class" keyword could be replaced by an "Actor" keyword similar how we have the "Object" keyword does it to (one of the reasons why I prefer Kotlin to Java). I'm writing on mobile and the preview button doesn't work right now, sry if the code is bad formatted.

[LifeIsStrange]

Note that Swift is currently working on actor model support, I wonder if some inspiration could be taken:
Food for thoughts:
https://news.ycombinator.com/item?id=26480922

Edit:
Active review:
https://forums.swift.org/t/se-0306-actors/45734
Review by Chris latner:
https://forums.swift.org/t/se-0306-actors/45734/4

[fvasco]

Hi @elizarov

We can have IDE inspections to help avoid "sharing" pitfalls to verify that all of the actor's state is private and all the public functions are properly written using act pattern.

I considering a lot this phrase, I wish to avoid this type of language support (workaround) around as much is possible.

To enforce this we can encapsulate the actor's state, so the programmer is forced to invoke the act function.

I try to explain better myself: an actor requires: a queue, some message types and an optional state.
I encapsulate the state in a ActorState type, and I can operate on it using some functions, like update (names are only examples).

The follow interface can encapsulate the state

interface ActorState<T> : CoroutineScope {

    suspend fun update(block: T.() -> Unit)
}

so we define a simple builder (may requires more parameters):

fun <T> actorState(state: T): ActorState<T> = TODO()

Finally an actor's example

class Counter /* no superclass required */ {

    private var state = actorState(object {
        var count = 0
    })

    suspend fun increment(amount: Int = 1) = state.update { count += amount }

}

That's all, I cannot more avoid the queue.

[elizarov]

@adam-arold The plan roughly outlined here (for "writing complex actor") is approximately how far it will go as a part of kotinx.coroutines in a foreseeable future. If you want more, I'd suggest to start writing your own coroutine-based actor framework. I'm sure it'll be quite popular if you make it open source.

[gortiz]

Kotlin actors are very low level and this patterns/strategies are not implemented yet and for what @elizarov said, they are not going to be implemented soon. As @Globegitter said, they are not trivial to implement.

As I said earlier, my personal recommendation is using Akka Typed with Kotlin Coroutines. They match so easily that seem to be created one for the other. I proposed a talk about that for KotlinConf, but unfortunately it wasn't chosen. I will try to blog about Akka Typed + Coroutines after the summer. Meanwhile I encourage you to try yourself!

[uberto]

On Tue, 6 Aug 2019 at 09:25, Gonzalo Ortiz Jaureguizar ***@***.***> wrote: Kotlin actors are very low level and this patterns/strategies are not implemented yet and for what @elizarov said, they are not going to be implemented soon. As @Globegitter said, they are not trivial to implement. As I said earlier, my personal recommendation is using Akka Typed with Kotlin Coroutines. They match so easily that seem to be created one for the other. I proposed a talk about that for KotlinConf, but unfortunately it wasn't chosen. I will try to blog about Akka Typed + Coroutines after the summer. Meanwhile I encourage you to try yourself!

Hi Gonzalo, Any chance to see the material of your talk (slide/code/blog?) somewhere? I'd be very interested. Also my proposals for KotlinConf have been rejected but if you present it anytime in London conf let me know. I've used the coroutine.actor(...) in a couple of projects and for our use works nicely in prod. Note that we use it as stateless worker. Since it has been deprecated now I'm tempted to remove it and just create a coroutine in its own thread that reads input from a channel and return an output on another channel. I agree that porting akka to Kotlin would make little sense, but having a lightweight lib in Kotlin to create workers with supervisors etc. would be quite beneficial. Uberto

[elizarov]

Let me voice my concern that adding such functionality to the core is a start of a slippery road of turning it in into an actor-based programming framework. There are lots of issues to be addressed in a large-scale actor-based programming and the corresponding support libraries are only bound to grow over time. Maybe we should think-through what other things needs to be added, beyond an actor base-class and a construction function, and have a separate module to support all of that.

[elizarov]

Even with capacity = 0 the Actor is asynchronous. You can send a message to an Actor and continue working on your code, while actor processes your message concurrently with your code. The Mutex, on the other hand, is always synchronous. No concurrency. That is, conceptually, why solutions based on Mutex/synchronized do not scale well.

[adam-arold]

@elizarov does this mean that I should not wait for improvements to the current actor? The documentation points to this issue and states that the current actor is obsolete and they will get replaced with the introduction of complex actors.

[adam-arold]

I'm planning to work on a Kotlin approach (not a wrapper) because I'm going to write common code which will work on all Kotlin-supported platforms. For this reason, Akka and Quasar are not applicable in my use case.

[fvasco]

Hi, @tristancaron,
pay attention, MutableStateFlow does not provide a CAS mutator (yet), _counter.value++ is not atomic.

[fvasco]

get and set value is thread safe, however _counter.value = _counter.value + 1 is not atomic, so you can successfully write a wrong value.

[fvasco]

Hi, @handstandsam,
I exposed above some classic data race issues.
A single writer does not need synchronization (with other writers).

Obviously multiple concurrent writers, even in the same context, may fail (https://kotlinlang.org/docs/reference/coroutines/shared-mutable-state-and-concurrency.html#mutual-exclusion).

[penn5]

Hmm, baking this framework into the language seems as a bit weird approach to take.

I agree, it should be a library based approach (like it is now)

[chriswu000]

what is the latest on the timeline/plans for actor api updates?

Reddit AMA they said possibly start work in 2022? #485 (comment)

[nowiwr01w]

Hi @elizarov, it's 2024. What about the new API?

[fvasco]

Hi @elizarov,
hi solved this issue in the same way, but an helper class looks like a miss of abstraction.

I want propose to introduce a dedicated interface for actor, something like:

interface Actor : ActorScope {
    fun invoke() : Unit
}

and the new function

fun launch(context: CoroutineContext, ..., actorBuild: (CoroutineContext) -> Actor) : ActorJob

I want to propose a draft implementation in the next week, what do you think about it?

[fvasco]

onReceive method is comfortable to use, but I want to expose you my concerns.

The actor function uses the iterable pattern, instead the Actor class use visitor pattern. Personally I consider the iterator pattern a better choice becouse it is possible to detect actor termination (end of iteration).

Using the explicit itaration in Actor makes it cut-and-paste compatible with the actor function.

I suggest to mantain Actor as public interface and to provide a comfortable implementation

public abstract class ActorBase<E>(actorScope: ActorScope<E>) : Actor<E>, ActorScope<E> by actorScope {

    override suspend fun invoke() {
        for (message in channel)
            onReceive(message)
    }

    public abstract suspend fun onReceive(message: E)
}

Finally actorOf misses of some parameters, we should keep this one, an extended version:

public fun <E> actorOf(
        context: CoroutineContext,
        capacity: Int = 0,
        start: CoroutineStart = CoroutineStart.DEFAULT,
        builder: (ActorScope<E>) -> Actor<E>
): ActorJob<E> = actor(context, capacity, start) {
    val actor = builder(this)
    actor.invoke()
}

or both?

[fvasco]

At this time of development, a stateless concurrent actor can be implemented using a BroadcastChannel to create a different ActorScope for each worker actor.

However your concern sounds right to me, but I suggest to mantain an uniformity of actor function and Actor class

public interface Actor<in E> : CoroutineScope {
    public suspend fun onReceive(message: E)
}

public fun <E> actor(
        context: CoroutineContext,
        capacity: Int = 0,
        start: CoroutineStart = CoroutineStart.DEFAULT,
        onReceive: suspend CoroutineScope.(E) -> Unit
): ActorJob<E>

In such case we deprecate ActorScope in favour of Actor.

[elizarov]

The idea of concurrent stateless actor is to have multiple instance working on the same mailbox in round-robin fashion as opposed to broadcasting messages to all actors. I actually considered adding an additional concurrency: Int = 1 optional parameter to actor builder to cover this use-case, so that you can launch multiple actors under the umbrella of a single ActorJob. However, it is not that really straight-forward as it does require us to start dabbling into supervision strategies, because multiple running coroutines do make it urgent to figure out what you are going to do when one of them crashes.

[fvasco]

Yes obviously, I was wrong.

As you proposed above I consider preferable a different module for actor supervisor and so on.
In my opinion this module should offer a common pool of interfaces for asynchronous programming, an "actor" module should implement more complex use cases, but I hope that the "core" module can interoperate well with existent solutions like Quasar or Akka.

So in my limited scope of view this "core" module shoud propose an "Actor" as a "Producer" dual, all other implementation should extends core's interfaces and provides its own builders.

[fvasco]

Hi @elizarov
I wish expose some consideration about your -interesting- proposal.

Actor, as intended until now, is the dual of Producer, so your design does not help to write complex actor.

Your proposal looks like a fully synchronized, non blocking class, which is equally interesting.
I wrote a dummy implementation using your requirement plus one: avoid abstract superclass; so I wrote an task synchronization helper (TaskQueue here) and a case of use.

class TaskQueue(
        val context: CoroutineContext = DefaultDispatcher,
        val mutex: Mutex = Mutex(),
        lazyInit: (suspend CoroutineScope.() -> Unit)? = null
) {

    private var lazyInit: Deferred<Unit>?

    init {
        this.lazyInit = lazyInit?.let {
            async(context, start = CoroutineStart.LAZY, block = it)
        }
    }

    /**
     * Force lazy initialization
     */
    suspend fun init() {
        lazyInit?.run {
            await()
            lazyInit = null
        }
    }

    suspend operator fun <T> invoke(block: () -> T): T {
        init()
        return mutex.withLock(this) {
            withContext(context) {
                block()
            }
        }
    }
}

class HttpSession {

    val start = Instant.now()

    private lateinit var state: MutableMap<String, String>

    private val taskQueue = TaskQueue {
        state = mutableMapOf()
    }

    suspend fun get(key: String) = taskQueue {
        state[key]
    }

    suspend fun set(key: String, value: String) {
        taskQueue {
            state[key] = value
        }
    }
}

Plus: using this implementation and issue #94 makes easy to implement a read task queue and a write task queue.

[elizarov]

@fvasco Indeed, it does look like a "like a fully synchronized, non blocking class", but it is not one. There are lots of similarities between monitor-based synchronization (like synchronized methods in Java) and actor-based programming model (like behavior functions in Pony). But there are important differences, too. Let me cite Pony documentation here:

A behaviour is like a function, except that functions are synchronous and behaviours are asynchronous.

Let's take a look at it in the context of Kotlin. First of all, we don't need new primitive for a "fully synchronized class". We already have Mutex. So, a fully synchronized class can be written like this:

class HttpSessionSync {
    private val mutex = Mutex()
    // state initialization does not have to happen under lock
    private var state: MutableMap<String, String> = mutableMapOf()
   
    suspend fun set(key: String, value: String) = mutex.withLock {
        state[key] = value
    }

    // etc
}

Notice the boilerplate here. We have to define private val mutex = Mutex() every time we use this pattern, so some kind of out-of-the box Mutex abstract base class might help. Shall we make Mutex open? Anyway, we don't want to promote this pattern, so we will not further discuss it in this thread.

You've made an interesting observation that requiring base class for complex actors is not good idea, so while, IMHO, we should give an option of using one, it should not be a requirement. Let's sketch implementation of a complex actor without having to use a base class:

class HttpSessionActor {
    private val actor = Actor()
    // state initialization does not have to happen under lock
    private var state: MutableMap<String, String> = mutableMapOf()
   
    suspend fun set(key: String, value: String) = actor.act {
        state[key] = value
    }

    // etc
}

What is the difference here as compared to Mutex-based version? The difference is that an Actor has an inbox channel and sending messaged to an actor can be decoupled from their execution. When HttpSessionActor.set is invoked, the invoker can go on with it own work while HttpSessionActor is busy, unless actor's mailbox capacity is exhausted. In the latter case, the invoker will have to wait until mailbox has capacity to store a message, but not longer. This starts to be important for scalability when actors perform long-running asynchronous activities.

[fvasco]

@elizarov thanks for explanation,
in some sense a Mutex act like an actor with capacity = 0 and without coroutine context.

I fix my draft, but I sure that it is possible to implement a better one.

class TaskChannel(
        context: CoroutineContext = DefaultDispatcher,
        capacity: Int = 0,
        lazyInit: (suspend CoroutineScope.() -> Unit)? = null
) {

    private val tasks = Channel<Task<*>>(capacity)

    private var lazyInit: Deferred<*>? = async(context, start = CoroutineStart.LAZY) {
        lazyInit?.invoke(this)

        launch(coroutineContext) {
            tasks.consumeEach { it() }
        }
    }

    /**
     * Force lazy initialization
     */
    suspend fun init() {
        lazyInit?.run {
            await()
            lazyInit = null
        }
    }

    suspend fun <T> act(block: suspend () -> T): Deferred<T> {
        init()
        val task = Task(block)
        tasks.send(task)
        return task
    }

    suspend fun <T> actAndReply(block: suspend () -> T): T = act(block).await()

    private class Task<T>(block: suspend () -> T) : CompletableDeferred<T> by CompletableDeferred() {
        private var block: (suspend () -> T)? = block

        suspend operator fun invoke() {
            try {
                complete(block!!())
            } catch (t: Throwable) {
                completeExceptionally(t)
            } finally {
                block = null
            }
        }
    }
}

[fvasco]

Even with capacity = 0 the Actor is asynchronous
The Mutex ... is always synchronous.

@elizarov can you confirm the follow code snippet?

suspend fun operationB(otherParams...): Result = actAndReply {
    // logic here
}

Is the functions's return typeResult and not Deferred<Result>?

Accordly with #261 it is pretty easy write the act function on top of a single private executor.

[vemilyus]

I've created an implementation of such an "ActorState" (as described by @fvasco)

Maybe it can be of some use, or you could also tell me how it is complete crap ;)

https://gist.github.com/vemilyus/c19f5d75525a37b33c4640bfd61158fe

[fvasco]

you could also tell me how it is complete crap

I am really tempted :)

You should use only a read-write mutex (#94), or a regular channel or mutex.

[gortiz]

The elizarov's proposal looks great in paper, but I would not delegate the sharing check responsibility on the IDE.

I prefer the Akka Typed approach where sharing is forbidden by the compiler and the separation between the channel/mailbox and the behavior is explicit.

[elizarov]

@gortiz At this point we are looking at pure library/IDE solution. Langauge-level changes are out the question for now. As for the explicit separation of behaviour, our goal here is to reduce the amount of boilerplate one need to write to define an actor, not increase it.

[adam-arold]

@elizarov I'll do that! I'm going to wait until your final solution is released then I'll build on top of that. @gortiz do you have plans for doing something similar?

[adam-arold]

@elizarov do you have a rough ETA on the final version of this Issue?

[adam-arold]

Kotlin Flow is gonna be awesome. I'll keep myself busy with something else in the meantime.

[elizarov]

No decision yet on current actors.

[Globegitter]

As this does not really mention much around this topic and I could also not find that much in other issues, what is the current statur around supervisors/supervision of actors, i.e. having strategies for automated restarts if a coroutine crashes. Is that something on the roadmap? If so is this something covered by this issue or would it make sense to create a new issue for this to have one place everyone can find who is interested in this topic?

[elizarov]

@Globegitter Automated restarts has to be a separate issue. Currently, there are no plans nor designs around it. It seems to work pretty well using existing tools when you just define a code that launches your coroutine as a function as then call it on crash to restart. It requires some boiler-plate to write, but due to good higher-order functions in Kotlin you'd need to write the corresponding code only once per project. For example, if you want to restart on any exception, just define:

fun CoroutineScope.launchAndRestart(body: suspend () -> Unit) {
    fun start() = launch { 
        try { body() } 
        catch (e: Throwable) { start() }
    }
    start()
}

Then use it: launchAndRestart { ... some code ... }

Note, that you can easily add your project-specific logging, exception-handling, and timeout logic. It does not seem to be very useful to have this kind of function in the core library, since it is really short.

[Globegitter]

Yeah true that is not too difficult indeed, but then looking at how other languages/frameworks have implemented that they usually offer different strategies, such as OneForOne, AllForOne, RestForOne, with optional exponential backoff, configuring an optional max number of retries etc. and getting that all right becomes imo less trivial all of the sudden. Granted, I am still quite new to this actor pattern and I am still in very early phases of our project, so still figuring out what behaviour makes the most sense etc, but from my point of view, if there was more native support for that, or some mentions in the docs about this it would make experimentation and evaluation easier.

All that said, I love the progress you guys are making and the overall direction this is heading into and coming from a mostly pythong/golang background it has been enjoyable to dive into coroutines/actors with kotlin.

[antanas-arvasevicius]

What about idea to port Akka ideas to Kotlin? Maybe without clustering first just core concepts.

[gortiz]

I would choose to avoid the fragmentation of the ecosystem. It is not trivial to create (and maintain!) a framework like Akka. In fact the worst thing about Akka is that the Java community (almost) ignores it, when it is an amazing framework to work with.

The main advantage of a Kotlin version of Akka is that it could be easily compiled to JS meanwhile connecting scala-js and kotlin-js and compiling Akka to js may be difficult (if possible). At the same time, I don't know web or android developers really need the complex actor patterns implemented in Akka. I would say that in these scenarios Kotlin actors are powerful enough (but I have almost 0 experience in these scenarios).

What I would really like to have is a Akka Kotlin DSL (the same way there is an Akka Java and Scala DSL), which would be almost as easy to implement as the Log4j Kotlin DSL.

[antanas-arvasevicius]

But by using Kotlin's foundation of async programming it would be easier to implement the same in Kotlin than it was for Akka in Scala. Kotlin has by default all primitives for that (channels, queues, back pressures, contexts(coroutines), supervisors).
Maybe with CoroutineContext serialization and deserialization it would be not so hard to implement Akka's Location Transparency ( https://doc.akka.io/docs/akka/current/general/remoting.html )
Sure messaging and clustering is complex thing and for this better to use existing solutions (Spark streaming, Kafka or similar ..) for clustering maybe (HashiCorp Consul - https://learn.hashicorp.com/consul/ )

I'd also would like to have something similar as Akka, but due lack of Scala knowledge in our team and unknown future of Scala itself we are avoiding it.

[penn5]

@elizarov thanks for the upvote, I'll start on the tests for it.

[qwwdfsad]

We are not ready to accept PR for actors right now.

Adding a feature is not only about the code that solves a specific user task or problem.
It is about the feature being consistent with existing features and future ones (that we already discussed and not published anywhere), about the same "library experience", about approachable and non-clashing naming (with existing features, with the-same-name-but-slightly-different-behaviour from other languages etc.), about being performance-proof, about naming bikeshedding (a lot of bikeshedding), about API shape that is hard to misuse and lot more. The actual code is not the biggest issue here :)

It's just a lot of work we are neither ready to do right now for actor nor ready to guide someone through this path via ten iterations of PR reviews.

[penn5]

Okay. I'll make a fork and use that for my project.

[harsh183]

Is there likely going to be any API level changes to how actor works with Coroutines at present?

[elizarov]

No on the near future. We do plan to introduce a more convenient API for writing actors. Existing actor function will be either integrate with new API or deprecated with some clear replacement. Old code using it will continue to work. We don't have a complete design on the table yet, so it is subject to change.

[harsh183]

Where is the new design being discussed? Is a lot of the setup erlang/elixir inspired?

[elizarov]

@harsh183 It is not in active discussions now, but postponed until the most urgent Flow issues are implemented.

Unfortunately, that's not much design inspiration we can take from either Erlang or Elixir, because both are dynamically typed languages, and so they avoid having to deal with the complicated problem of providing type-safe and boilerplate-free actor abstraction, which is our goal here.

[harsh183]

I see. Can you ping here again when it's being talked about again. I'm coming over from the Ruby side where I contribute to Celluloid for actor model there.

…

On Mon, Apr 27, 2020, 5:24 PM Roman Elizarov ***@***.***> wrote: @harsh183 <https://github.com/harsh183> It is not in active discussions now, but postponed until the most urgent Flow issues are implemented. Unfortunately, that's not much design inspiration we can take from either Erlang or Elixir, because both are dynamically typed languages, and so they avoid having to deal with the complicated problem of providing type-safe and boilerplate-free actor abstraction, which is our goal here. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#87 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB3LOZW2LQT7Z263RMLIPVLROVW7DANCNFSM4DUAKLSQ> .

[ZakTaccardi]

Something I’m thinking about is how send(..) on the current actor behaves (as it implements SendChannel<T>).

I believe it might be useful to suspend until the actor has completed its processing of the message.

So then there would be 3 ways to communicate into the actor:

1. non-suspending offer(..)
   2.send(..) that suspends until the actor can guarantee the message will be delivered
2. send(..) that suspends until the actor has processed the message

Under what conditions is number 2 useful?

[tristancaron]

What is the benefit of using Actors instead of StateFlow? Here two implementations of CounterModel:

fun CoroutineScope.counterActor() = actor<CounterMsg> {
    var counter = 0 // actor state
    for (msg in channel) { // iterate over incoming messages
        when (msg) {
            is IncCounter -> counter++
            is GetCounter -> msg.response.complete(counter)
        }
    }
}

and

class CounterModel {
    private val _counter = MutableStateFlow(0) // private mutable state flow
    val counter: StateFlow<Int> get() = _counter // publicly exposed as read-only state flow

    fun inc() {
        _counter.value++
    }
}

[tristancaron]

Thanks for taking the time to answer @fvasco

So MutableStateFlow is not atomic, thus we can have a data race. Am I understanding it correctly?

If so, the documentation seems to imply that data race is not possible. https://kotlin.github.io/kotlinx.coroutines/kotlinx-coroutines-core/kotlinx.coroutines.flow/-state-flow/index.html

Concurrency
All methods of data flow are thread-safe and can be safely invoked from concurrent coroutines without external synchronization.

Unless I misunderstood it.

[handstandsam]

@fvasco - Since the get and set methods of StateFlow are thread-safe, isn't there no need for _counter.value to be atomic? All get and set requests will be processed in order meaning that there would be no race conditions. If you ended up doing background work inside the model to modify _counter, I could see an issue, but as long as you are doing computations within that context you should be fine, right?

[adam-arold]

What should I use in the meantime if I don't want to use a Mutex (because I'm afraid of deadlocks)? What I have right now is a super primitive wrapper class that can get/transform its state:

class Atom<T : Any>(initialValue: T) {

    @Volatile
    private var value: T = initialValue

    override fun get(): T = value

    @Synchronized
    override fun transform(transformer: (T) -> T): T {
        value = transformer(value)
        return value
    }

}

but with this I can still shoot myself in the foot if two of these Atoms cause a deadlock. What i'm looking for is a way to calculate a new value for shared mutable state (transform here) that's relatively fast. I scrapped my other implementation that was using actor when it became obsolete, but I'm not happy with the current solution.

[LifeIsStrange]

@dimsuz I don't know but at least the pros and cons of doing this should be outlined in their RFC

[fvasco]

@Kshitij09-sc
You are mixing blocking (dispatchChannel.trySendBlocking(action)) and asynchronous (action: suspend ActorContext.() -> Unit) code, this is a delicate choice and may lead to errors.
The Job in initializeIfNeeded executes actions concurrently (two nested launch), so printCount will print a number concurrently with other tasks.
You are using an AtomicBoolean instead of a lazy launch.

[Kshitij09-sc]

@fvasco Agree with latter 2 points. Regarding mixing trySendBlocking(action) with suspending action: suspend ActorContext.() -> Unit, I considered this to be able to detect deadlocks. trySendBlocking will return failure once queue is full, queue might get full due to less buffer capacity or some action went into infinite loop (deadlock) which could be thrown as exception in debug environment.