Fs2 + Effects

Streaming applications

About me

Effects

cats.effect.IO[A] 

A pure abstraction representing the intention to perform a side effect, where the result of that side effect may be obtained synchronously (via return) or asynchronously (via callback).

trait Effect[F[_]] extends Async[F]

A monad that can suspend side effects into the `F` context and that supports lazy and potentially asynchronous evaluation.

Effects

f(println("hi"), println("hi"))
val x = println("hi")
f(x, x)
import cats.effect.IO
import scala.io.StdIn
  
val program = for {
  _     <- IO { println("Please enter your name:") }
  name  <- IO { StdIn.readLine }
  _     <- IO { println(s"Hi $name!") }
} yield ()

program.unsafeRunSync()

"If replacing expression x by its value produces the same behavior, then x is referentially transparent"

def putStrLn(line: String) = IO { println(line) }
val x = putStrLn("hi")
f(x, x) == f(putStrLn("hi"), putStrLn("hi"))

Fs2 concepts

Stream[F,O]

 It represents a discrete stream of O values which may request evaluation of F effects.

Pipe[F,I,O]
= Stream[F,I] => Stream[F,O]
Sink[F,I]
= Pipe[F,I,Unit]

Just a streaming transformation!

Its sole purpose is to run effects.

A bit of history

Process[F,O]
F[A] = scalaz.Task[A]

Case of Study

Order Generator

Pricer Service

Case of Study

Order Generator

Pricer Service

Case of Study

import cats.effect.IO
import com.gvolpe.fs2.streams.model._
import fs2.{Pipe, Sink, Stream}

import scala.concurrent.ExecutionContext

class PricerFlow() {

  def flow(consumer: Stream[IO, Order],
           logger: Sink[IO, Order],
           storage: OrderStorage,
           pricer: Pipe[IO, Order, Order],
           publisher: Sink[IO, Order])
          (implicit ec: ExecutionContext): Stream[IO, Unit] = {
    Stream(
      consumer      observe logger to storage.write,
      storage.read  through pricer to publisher
    ).join(2)
  }

}

Pricer Service

Case of Study

class OrderGeneratorFlow()(implicit ec: ExecutionContext, R: Scheduler) {

  private def defaultOrderGen: Pipe[IO, Int, Order] = { orderIds =>
    val tickInterrupter = time.sleep[IO](11.seconds).map(_ => true)
    val orderTick       = time.awakeEvery[IO](2.seconds).interruptWhen(tickInterrupter)
    (orderIds zip orderTick) flatMap { case (id, _) =>
      val itemId    = Random.nextInt(500).toLong
      val itemPrice = Random.nextInt(10000).toDouble
      val newOrder  = Order(id.toLong, List(Item(itemId, s"laptop-$id", itemPrice)))
      Stream.emit(newOrder)
    }
  }

  def flow(source: Sink[IO, Order], 
           orderGen: Pipe[IO, Int, Order] = defaultOrderGen): Stream[IO, Unit] = {
    Stream.range(1, 10).covary[IO] through orderGen to source
  }

}

Pricer Service

Case of Study

implicit val R = fs2.Scheduler.fromFixedDaemonPool(2, "generator-scheduler")
implicit val S = scala.concurrent.ExecutionContext.Implicits.global

val pricerProgram = for {
  kafkaTopic    <- Stream.eval(async.topic[IO, Order](Order.Empty))
  rabbitQueue   <- Stream.eval(async.boundedQueue[IO, Order](100))
  dbQueue       <- Stream.eval(async.boundedQueue[IO, Order](100))
  kafkaBroker   = new OrderKafkaBroker(kafkaTopic)
  rabbitBroker  = new OrderRabbitMqBroker(rabbitQueue)
  db            = new OrderDb(dbQueue)
  pricerFlow    = new PricerFlow().flow(kafkaBroker.consume, logger, 
                                        OrderStorage(db.read, db.persist), 
                                        pricer, rabbitBroker.produce)
  orderGenFlow  = new OrderGeneratorFlow().flow(kafkaBroker.produce)
  program       <- pricerFlow mergeHaltBoth orderGenFlow
} yield program

pricerProgram.run.unsafeRunSync()

Pricer Service

Case of Study

protected def acquireConnection[F[_] : Effect]: F[(Connection, Channel)] =
  Effect[F].delay {
    val conn    = factory.newConnection
    val channel = conn.createChannel
    (conn, channel)
  }

/**
  * Creates a connection and a channel in a safe way using Stream.bracket.
  * In case of failure, the resources will be cleaned up properly.
  *
  * @return An effectful [[fs2.Stream]] of type [[Channel]]
  * */
def createConnectionChannel[F[_] : Effect](): Stream[F, Channel] =
  Stream.bracket(acquireConnection)(
    cc => asyncF[F, Channel](cc._2),
    cc => Effect[F].delay {
      val (conn, channel) = cc
      log.info(s"Releasing connection: $conn previously acquired.")
      if (channel.isOpen) channel.close()
      if (conn.isOpen) conn.close()
    }
  )

Fs2 Rabbit library

Case of Study

def jsonEncode[F[_] : Effect, A : Encoder]: Pipe[F, AmqpMessage[A], AmqpMessage[String]] =
  streamMsg =>
    for {
      amqpMsg <- streamMsg
      json    <- asyncF[F, String](amqpMsg.payload.asJson.noSpaces)
    } yield AmqpMessage(json, amqpMsg.properties)

private val log = LoggerFactory.getLogger(getClass)

def jsonDecode[F[_] : Effect, A : Decoder]: 
  Pipe[F, AmqpEnvelope, (Either[Error, A], DeliveryTag)] =
    streamMsg =>
      for {
        amqpMsg <- streamMsg
        parsed  <- asyncF[F, Either[Error, A]](decode[A](amqpMsg.payload))
        _       <- asyncF[F, Unit](log.debug(s"Parsed: $parsed"))
      } yield (parsed, amqpMsg.deliveryTag)

// asyncF(body) is an alias for Stream.eval(F.delay(body))

Fs2 Rabbit library

Comparison

Monix vs Akka Stream vs Fs2

Resources

Questions?

Made with Slides.com