Proper usage of Futures in parallel calculations - scala

I'm implementing an algorithm that can be easily parallelized but can't figure out how to create a proper number of futures and how to abort early. At the moment the outline of the code is along these lines
def solve: Boolean = {
var result = false
while(!result && i < iterations) {
val futures = (1 to threads) map { _ => solveIter(geInitialValues()) }
val loopResult = Future.fold(futures)(false)((acc, r) => acc || r )
result = Await.result(loopResult, Duration.Inf)
i+=1
}
}
def solveIter(initialValues: Values): Future[Boolean] = Future {
/* Takes a lot of time */
}
The obvious problem is explicitly set level of parallelism that may or may not be suitable for current execution context. If all futures are created at once how to make Future.fold to abort early?

You can't cancel a Future, because Futures are read-only. But you can use a Promise which is "the write part from a Future".
Example code:
this code timeouts after 5 seconds because the futures are not done (solveIter never completes)
to complete a promise, remove comment which adds a completed promise to the 'promises' - the other futures will be cancel
remove 'promises.foreach(_.trySuccess(false))' and you get the timeout again, because the other futures doesn't getting cancel
import scala.concurrent.ExecutionContext.Implicits.global
import scala.concurrent.duration._
import scala.concurrent.{Await, Future, Promise}
import scala.util.Try
// create a bunch of promises
val promises = ((1 to 10) map { _ =>
val p = Promise[Boolean]()
p.completeWith(solveIter())
p
}) // :+ Promise().success(true)
// ^^ REMOVE THIS COMMENT TO ADD A PROMISE WHICH COMPLETES
// get the futures from the promises
val futures = promises.map(_.future)
// loop over all futures
futures.foreach(oneFuture =>
// register callback when future is done
oneFuture.foreach{
case true =>
println("future with 'true' result found")
// stop others
promises.foreach(_.trySuccess(false))
case _ => // future completes with false
})
// wait at most 5 seconds till all futures are done
Try(Await.ready(Future.sequence(futures), 5.seconds)).recover { case _ =>
println("TIMEOUT")
}
def solveIter(): Future[Boolean] = Future {
/* Takes a VERY VERY VERY .... lot of time */
Try(Await.ready(Promise().future, Duration.Inf))
false
}

Related

Eagerly-evaluate-and-forget behavior for Cats Effect IO

I'm converting Future code to IO. I have code similar to this
def doSomething: Future[Foo] = {
Future {
//some code the result of which we don't care about
}
Future {
//Foo
}
}
And then at the end of the program, I doSomething.unsafeRunSync. How do I convert these Futures to IOs while maintaining the fire-and-forget functionality of the first Future? In using IO's async API, I am worried about accidentally blocking the thread when I later call unsafeRunSync on doSomething.
A solution that uses only cats-effect could use IO.start. This, combined with the fact that you will then never join the resulting Fiber, will look something like this:
import cats.effect._
import cats.implicits._
import scala.concurrent.ExecutionContext.Implicits.global
import scala.concurrent.duration._
object ExampleApp extends App{
val fireAndForget =
IO(println("Side effect pre-sleep")) *>
IO.sleep(2.seconds) *>
IO(println("Side effect post-sleep"))
val toBeUsed = IO{
println("Inside second one")
42
}
val result = for {
fiber <- IO.shift *> fireAndForget.start
res <- toBeUsed.handleErrorWith { error =>
// This is just in case you 'toBeUsed' can actually fail,
// and you might want to cancel the original side-effecting IO
fiber.cancel *> IO.raiseError(error) }
} yield res
println(result.unsafeRunSync())
println("Waiting 3 seconds...")
IO.sleep(3.seconds).unsafeRunSync()
println("Done")
}
This will print (most of the times) something similar to:
Side effect pre-sleep
Inside second one
42 // Up until here, will be printed right away
Waiting 3 seconds... // It will then be waiting a while
Side effect post-sleep // ...at which point the side effecting code terminates
Done
Finally, here are more details about Fiber and IO.shift
I believe that you need to wrap the first Future in such a way that it completes immediately. We ignore exeptions, or catch them however, but they are contained within its own thread. The parameter cb is the promise that needs to complete; so we short-circuit the completion by providing a value immediately.
def firstFuture(implicit ec: ExecutionContext): IO[Unit] = {
IO.async[Unit] { cb =>
ec.execute(() => {
try {
//some code the result of which we don't care about
} catch {
}
})
cb(Right(()))
}
}
In the for-comprehension, the firstFuture will complete immediately even though its thread will have a long-running task active on it.
def doSomething(implicit ec: ExecutionContext): IO[Foo] = {
for {
_ <- firstFuture
IO.async[Foo] { fb =>
// Foo
}
}
}

Get actual value in Future Scala [duplicate]

I am a newbie to scala futures and I have a doubt regarding the return value of scala futures.
So, generally syntax for a scala future is
def downloadPage(url: URL) = Future[List[Int]] {
}
I want to know how to access the List[Int] from some other method which calls this method.
In other words,
val result = downloadPage("localhost")
then what should be the approach to get List[Int] out of the future ?
I have tried using map method but not able to do this successfully.`
The case of Success(listInt) => I want to return the listInt and I am not able to figure out how to do that.
The best practice is that you don't return the value. Instead you just pass the future (or a version transformed with map, flatMap, etc.) to everyone who needs this value and they can add their own onComplete.
If you really need to return it (e.g. when implementing a legacy method), then the only thing you can do is to block (e.g. with Await.result) and you need to decide how long to await.
You need to wait for the future to complete to get the result given some timespan, here's something that would work:
import scala.concurrent.duration._
def downloadPage(url: URL) = Future[List[Int]] {
List(1,2,3)
}
val result = downloadPage("localhost")
val myListInt = result.result(10 seconds)
Ideally, if you're using a Future, you don't want to block the executing thread, so you would move your logic that deals with the result of your Future into the onComplete method, something like this:
result.onComplete({
case Success(listInt) => {
//Do something with my list
}
case Failure(exception) => {
//Do something with my error
}
})
I hope you already solved this since it was asked in 2013 but maybe my answer can help someone else:
If you are using Play Framework, it support async Actions (actually all Actions are async inside). An easy way to create an async Action is using Action.async(). You need to provide a Future[Result]to this function.
Now you can just make transformations from your Future[List[Int]] to Future[Result] using Scala's map, flatMap, for-comprehension or async/await. Here an example from Play Framework documentation.
import play.api.libs.concurrent.Execution.Implicits.defaultContext
def index = Action.async {
val futureInt = scala.concurrent.Future { intensiveComputation() }
futureInt.map(i => Ok("Got result: " + i))
}
You can do something like that. If The wait time that is given in Await.result method is less than it takes the awaitable to execute, you will have a TimeoutException, and you need to handle the error (or any other error).
import scala.concurrent._
import ExecutionContext.Implicits.global
import scala.util.{Try, Success, Failure}
import scala.concurrent.duration._
object MyObject {
def main(args: Array[String]) {
val myVal: Future[String] = Future { silly() }
// values less than 5 seconds will go to
// Failure case, because silly() will not be done yet
Try(Await.result(myVal, 10 seconds)) match {
case Success(extractedVal) => { println("Success Happened: " + extractedVal) }
case Failure(_) => { println("Failure Happened") }
case _ => { println("Very Strange") }
}
}
def silly(): String = {
Thread.sleep(5000)
"Hello from silly"
}
}
The best way I’ve found to think of a Future is a box that will, at some point, contain the thing that you want. The key thing with a Future is that you never open the box. Trying to force open the box will lead you to blocking and grief. Instead, you put the Future in another, larger box, typically using the map method.
Here’s an example of a Future that contains a String. When the Future completes, then Console.println is called:
import scala.concurrent.Future
import scala.concurrent.ExecutionContext.Implicits.global
object Main {
def main(args:Array[String]) : Unit = {
val stringFuture: Future[String] = Future.successful("hello world!")
stringFuture.map {
someString =>
// if you use .foreach you avoid creating an extra Future, but we are proving
// the concept here...
Console.println(someString)
}
}
}
Note that in this case, we’re calling the main method and then… finishing. The string’s Future, provided by the global ExecutionContext, does the work of calling Console.println. This is great, because when we give up control over when someString is going to be there and when Console.println is going to be called, we let the system manage itself. In constrast, look what happens when we try to force the box open:
val stringFuture: Future[String] = Future.successful("hello world!")
val someString = Future.await(stringFuture)
In this case, we have to wait — keep a thread twiddling its thumbs — until we get someString back. We’ve opened the box, but we’ve had to commandeer the system’s resources to get at it.
It wasn't yet mentioned, so I want to emphasize the point of using Future with for-comprehension and the difference of sequential and parallel execution.
For example, for sequential execution:
object FuturesSequential extends App {
def job(n: Int) = Future {
Thread.sleep(1000)
println(s"Job $n")
}
val f = for {
f1 <- job(1)
f2 <- job(2)
f3 <- job(3)
f4 <- job(4)
f5 <- job(5)
} yield List(f1, f2, f3, f4, f5)
f.map(res => println(s"Done. ${res.size} jobs run"))
Thread.sleep(6000) // We need to prevent main thread from quitting too early
}
And for parallel execution (note that the Future are before the for-comprehension):
object FuturesParallel extends App {
def job(n: Int) = Future {
Thread.sleep(1000)
println(s"Job $n")
}
val j1 = job(1)
val j2 = job(2)
val j3 = job(3)
val j4 = job(4)
val j5 = job(5)
val f = for {
f1 <- j1
f2 <- j2
f3 <- j3
f4 <- j4
f5 <- j5
} yield List(f1, f2, f3, f4, f5)
f.map(res => println(s"Done. ${res.size} jobs run"))
Thread.sleep(6000) // We need to prevent main thread from quitting too early
}

`akka.pattern.after` - Understanding `duration` Argument?

Using Scala 2.11.8, given a Future that sleeps 5 seconds before returning (), I called akka.pattern.ask with a time-out of 1 second:
import scala.concurrent.duration._
import scala.concurrent._
import akka.actor.ActorSystem
import scala.concurrent.ExecutionContext.Implicits.global
import akka.pattern.after
val system = ActorSystem.create()
after( FiniteDuration(1, SECONDS), system.scheduler)( Future{ Thread.sleep( 5000 ); () } )
res15: scala.concurrent.Future[Unit] = List()
// roughly 5 seconds later, I see:
scala> res15
res22: scala.concurrent.Future[Unit] = Success(())
Why doesn't res15 return a Failure since the Future does not complete until 5 seconds, however its timeout is 1 second?
After method will run your future after some delay. duration is not timeout, its the starting delay. After method internally uses scheduleOnce method of actorSystem schedular.
After is very interesting method which starts evaluating/executing your future expression (i.e value) only after the certain given duration.
If the given duration is very small then its starts executing the value immediately.
As value is a call by name parameter the value is not immediately evaluated. value will be evaluated only when explicitly called by the name value.
Here is the code from akka lib.
def after[T](duration: FiniteDuration, using: Scheduler)(value: ⇒ Future[T])(implicit ec: ExecutionContext): Future[T] =
if (duration.isFinite() && duration.length < 1) {
try value catch { case NonFatal(t) ⇒ Future.failed(t) }
} else {
val p = Promise[T]()
using.scheduleOnce(duration) { p completeWith { try value catch { case NonFatal(t) ⇒ Future.failed(t) } } }
p.future
}
Basically your duration will trigger the following code
using.scheduleOnce(duration) { p completeWith { try value catch { case NonFatal(t) ⇒ Future.failed(t) } } }
The above code will start executing the future after delay(duration is the delay.).
Important observations:
note that value is a call by name parameter. value will be evaluated when called by name that means your future will not be start executing until and unless value is called explicitly. value is called inside try because value could be a multi line expression which contains lot of code and which returns a future finally.
Promise (see in the code) will make you wait in a non blocking fashion till the future is completed its execution. promise.future will only complete when given future completes.

Scala waiting for sequence of futures

I was hoping code like follows would wait for both futures, but it does not.
object Fiddle {
val f1 = Future {
throw new Throwable("baaa") // emulating a future that bumped into an exception
}
val f2 = Future {
Thread.sleep(3000L) // emulating a future that takes a bit longer to complete
2
}
val lf = List(f1, f2) // in the general case, this would be a dynamically sized list
val seq = Future.sequence(lf)
seq.onComplete {
_ => lf.foreach(f => println(f.isCompleted))
}
}
val a = FuturesSequence
I assumed seq.onComplete would wait for them all to complete before completing itself, but not so; it results in:
true
false
.sequence was a bit hard to follow in the source of scala.concurrent.Future, I wonder how I would implement a parallel that waits for all original futures of a (dynamically sized) sequence, or what might be the problem here.
Edit: A related question: https://worldbuilding.stackexchange.com/questions/12348/how-do-you-prove-youre-from-the-future :)
One common approach to waiting for all results (failed or not) is to "lift" failures into a new representation inside the future, so that all futures complete with some result (although they may complete with a result that represents failure). One natural way to get that is lifting to a Try.
Twitter's implementation of futures provides a liftToTry method that makes this trivial, but you can do something similar with the standard library's implementation:
import scala.util.{ Failure, Success, Try }
val lifted: List[Future[Try[Int]]] = List(f1, f2).map(
_.map(Success(_)).recover { case t => Failure(t) }
)
Now Future.sequence(lifted) will be completed when every future is completed, and will represent successes and failures using Try.
And so, a generic solution for waiting on all original futures of a sequence of futures may look as follows, assuming an execution context is of course implicitly available.
import scala.util.{ Failure, Success, Try }
private def lift[T](futures: Seq[Future[T]]) =
futures.map(_.map { Success(_) }.recover { case t => Failure(t) })
def waitAll[T](futures: Seq[Future[T]]) =
Future.sequence(lift(futures)) // having neutralized exception completions through the lifting, .sequence can now be used
waitAll(SeqOfFutures).map {
// do whatever with the completed futures
}
A Future produced by Future.sequence completes when either:
all the futures have completed successfully, or
one of the futures has failed
The second point is what's happening in your case, and it makes sense to complete as soon as one of the wrapped Future has failed, because the wrapping Future can only hold a single Throwable in the failure case. There's no point in waiting for the other futures because the result will be the same failure.
This is an example that supports the previous answer. There is an easy way to do this using just the standard Scala APIs.
In the example, I am creating 3 futures. These will complete at 5, 7, and 9 seconds respectively. The call to Await.result will block until all futures have resolved. Once all 3 futures have completed, a will be set to List(5,7,9) and execution will continue.
Additionally, if an exception is thrown in any of the futures, Await.result will immediately unblock and throw the exception. Uncomment the Exception(...) line to see this in action.
try {
val a = Await.result(Future.sequence(Seq(
Future({
blocking {
Thread.sleep(5000)
}
System.err.println("A")
5
}),
Future({
blocking {
Thread.sleep(7000)
}
System.err.println("B")
7
//throw new Exception("Ha!")
}),
Future({
blocking {
Thread.sleep(9000)
}
System.err.println("C")
9
}))),
Duration("100 sec"))
System.err.println(a)
} catch {
case e: Exception ⇒
e.printStackTrace()
}
Even though it is quite old question But this is how I got it running in recent time.
object Fiddle {
val f1 = Future {
throw new Throwable("baaa") // emulating a future that bumped into an exception
}
val f2 = Future {
Thread.sleep(3000L) // emulating a future that takes a bit longer to complete
2
}
val lf = List(f1, f2) // in the general case, this would be a dynamically sized list
val seq = Future.sequence(lf)
import scala.concurrent.duration._
Await.result(seq, Duration.Inf)
}
This won't get completed and will wait till all the future gets completed. You can change the waiting time as per your use case. I have kept it to infinite and that was required in my case.
We can enrich Seq[Future[T]] with its own onComplete method through an implicit class:
def lift[T](f: Future[T])(implicit ec: ExecutionContext): Future[Try[T]] =
f map { Success(_) } recover { case e => Failure(e) }
def lift[T](fs: Seq[Future[T]])(implicit ec: ExecutionContext): Seq[Future[Try[T]]] =
fs map { lift(_) }
implicit class RichSeqFuture[+T](val fs: Seq[Future[T]]) extends AnyVal {
def onComplete[U](f: Seq[Try[T]] => U)(implicit ec: ExecutionContext) = {
Future.sequence(lift(fs)) onComplete {
case Success(s) => f(s)
case Failure(e) => throw e // will never happen, because of the Try lifting
}
}
}
Then, in your particular MWE, you can do:
val f1 = Future {
throw new Throwable("baaa") // emulating a future that bumped into an exception
}
val f2 = Future {
Thread.sleep(3000L) // emulating a future that takes a bit longer to complete
2
}
val lf = List(f1, f2)
lf onComplete { _ map {
case Success(v) => ???
case Failure(e) => ???
}}
This solution has the advantage of allowing you to call an onComplete on a sequence of futures as you would on a single future.
Create the Future with a Try to avoid extra hoops.
implicit val ec = ExecutionContext.global
val f1 = Future {
Try {
throw new Throwable("kaboom")
}
}
val f2 = Future {
Try {
Thread.sleep(1000L)
2
}
}
Await.result(
Future.sequence(Seq(f1, f2)), Duration("2 sec")
) foreach {
case Success(res) => println(s"Success. $res")
case Failure(e) => println(s"Failure. ${e.getMessage}")
}

Cancellation with Future and Promise in Scala

This is a followup to my previous question.
Suppose I have a task, which executes an interruptible blocking call. I would like to run it as a Future and cancel it with failure method of Promise.
I would like the cancel to work as follows:
If one cancels the task before it finished I would like the task to finish "immediately", interrupting the blocking call if it has already started and I would like the Future to invoke onFailure.
If one cancels the task after the task finished I would like to get a status saying that the cancel failed since the task already finished.
Does it make sense? Is it possible to implement in Scala? Are there any examples of such implementations?
scala.concurrent.Future is read-only, so one reader cannot mess things up for the other readers.
It seems like you should be able to implement what you want as follows:
def cancellableFuture[T](fun: Future[T] => T)(implicit ex: ExecutionContext): (Future[T], () => Boolean) = {
val p = Promise[T]()
val f = p.future
p tryCompleteWith Future(fun(f))
(f, () => p.tryFailure(new CancellationException))
}
val (f, cancel) = cancellableFuture( future => {
while(!future.isCompleted) continueCalculation // isCompleted acts as our interrupted-flag
result // when we're done, return some result
})
val wasCancelled = cancel() // cancels the Future (sets its result to be a CancellationException conditionally)
Here is the interruptable version of Victor's code per his comments (Victor, please correct me if I misinterpreted).
object CancellableFuture extends App {
def interruptableFuture[T](fun: () => T)(implicit ex: ExecutionContext): (Future[T], () => Boolean) = {
val p = Promise[T]()
val f = p.future
val aref = new AtomicReference[Thread](null)
p tryCompleteWith Future {
val thread = Thread.currentThread
aref.synchronized { aref.set(thread) }
try fun() finally {
val wasInterrupted = (aref.synchronized { aref getAndSet null }) ne thread
//Deal with interrupted flag of this thread in desired
}
}
(f, () => {
aref.synchronized { Option(aref getAndSet null) foreach { _.interrupt() } }
p.tryFailure(new CancellationException)
})
}
val (f, cancel) = interruptableFuture[Int] { () =>
val latch = new CountDownLatch(1)
latch.await(5, TimeUnit.SECONDS) // Blocks for 5 sec, is interruptable
println("latch timed out")
42 // Completed
}
f.onFailure { case ex => println(ex.getClass) }
f.onSuccess { case i => println(i) }
Thread.sleep(6000) // Set to less than 5000 to cancel
val wasCancelled = cancel()
println("wasCancelled: " + wasCancelled)
}
With Thread.sleep(6000) the output is:
latch timed out
42
wasCancelled: false
With Thread.sleep(1000) the output is:
wasCancelled: true
class java.util.concurrent.CancellationException
Twitter's futures implement cancellation. Have a look here:
https://github.com/twitter/util/blob/master/util-core/src/main/scala/com/twitter/util/Future.scala
Line 563 shows the abstract method responsible for this. Scala's futures currently do not support cancellation.
You can use Monix library instead of Future
https://monix.io