The document discusses the design and performance motivations behind ZIO streams and their comparison with other streaming libraries like Scalaz and Akka Streams. It highlights the efficiency and flexibility of ZIO streams for handling various data processing tasks while also addressing issues like leak-free data processing and incremental computation. The document includes a variety of practical examples demonstrating the API and functionalities of ZIO streams.

1. John A. De Goes
@jdegoes - degoes.net
SCALAZ STREAM
REBIRTH
Scale By The Bay · November 16 · San Francisco
Itamar Ravid
@itrvd - iravid.com

2. #ZIO
#SCALEBYTHEBAY

3. PERFORMANCE
OVERVIEW
MOTIVATION API DESIGN

4. 1 Why Streams & Why Scalaz Stream
MOTIVATION

5. Why Streams?
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Infinite Input
Sets
Leak-Free Data
Processing
Incremental
Computation
∞

6. File & Socket
Processing
Graph
Processing
Reactive &
Dataflow
Akka Streams FS2
Scalaz ReactiveGraphJet
Why ZIO Stream?
PERFORMANCEMOTIVATION API DESIGN SUMMARY

7. Why ZIO Stream?
File & Socket
Processing
Graph
Processing
Reactive &
Dataflow
Akka Streams FS2
Scalaz ReactiveGraphJet
ZIO STREAM
PERFORMANCEMOTIVATION API DESIGN SUMMARY

8. Why ZIO Stream?
PERFORMANCEMOTIVATION API DESIGN SUMMARY
𐄂 Cycles𐄂 Cons 𐄂 Uncons

9. ZIO Stream Old Scalaz Stream Akka Streams FS2
Lazy ✓ 𐄂 𐄂 ✓
Leak-Free ✓ 𐄂 𐄂 On a good day
Uniform ✓ NA NA 𐄂
Type Inferred ✓ 𐄂 ✓ 𐄂
Economical ✓ 𐄂 𐄂 ✓
ZIO Integration ✓ 𐄂 𐄂 𐄂
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Why ZIO Stream?

10. 2 What Scalaz Stream Looks Like
API

11. PERFORMANCEMOTIVATION API DESIGN SUMMARY
AkkaStream AbruptTerminationException AbstractShape ActorAttributes ActorMaterializer ActorMaterializerSettings
AmorphousShape Attributes BidiShape BindFailedException BufferOverflowException Client ClosedShape
ConnectionException DelayOverflowStrategy EagerClose FanInShape FanInShape10 FanInShape11 FanInShape12
FanInShape13 FanInShape14 FanInShape15 FanInShape16 FanInShape17 FanInShape18 FanInShape19 FanInShape1N
FanInShape2 FanInShape20 FanInShape21 FanInShape22 FanInShape3 FanInShape4 FanInShape5 FanInShape6
FanInShape7 FanInShape8 FanInShape9 FanOutShape FanOutShape10 FanOutShape11 FanOutShape12 FanOutShape13
FanOutShape14 FanOutShape15 FanOutShape16 FanOutShape17 FanOutShape18 FanOutShape19 FanOutShape2
FanOutShape20 FanOutShape21 FanOutShape22 FanOutShape3 FanOutShape4 FanOutShape5 FanOutShape6
FanOutShape7 FanOutShape8 FanOutShape9 FlowMonitor FlowMonitorState FlowShape Fusing Graph IgnoreBoth
IgnoreCancel IgnoreComplete Inlet InPort IOResult KillSwitch KillSwitches MaterializationContext MaterializationException
Materializer MaterializerLoggingProvider Outlet OutPort OverflowStrategy QueueOfferResult RateExceededException
Server Shape SharedKillSwitch SinkShape SourceShape StreamLimitReachedException
StreamSubscriptionTimeoutSettings StreamSubscriptionTimeoutTerminationMode StreamTcpException
SubstreamCancelStrategy Supervision ThrottleMode TLSClientAuth TLSClosing TLSProtocol TLSRole UniformFanInShape
UniformFanOutShape UniqueKillSwitch

12. PERFORMANCEMOTIVATION API DESIGN SUMMARY
AkkaStream AbruptTerminationException AbstractShape ActorAttributes ActorMaterializer ActorMaterializerSettings
AmorphousShape Attributes BidiShape BindFailedException BufferOverflowException Client ClosedShape
ConnectionException DelayOverflowStrategy EagerClose FanInShape FanInShape10 FanInShape11 FanInShape12
FanInShape13 FanInShape14 FanInShape15 FanInShape16 FanInShape17 FanInShape18 FanInShape19 FanInShape1N
FanInShape2 FanInShape20 FanInShape21 FanInShape22 FanInShape3 FanInShape4 FanInShape5 FanInShape6
FanInShape7 FanInShape8 FanInShape9 FanOutShape FanOutShape10 FanOutShape11 FanOutShape12 FanOutShape13
FanOutShape14 FanOutShape15 FanOutShape16 FanOutShape17 FanOutShape18 FanOutShape19 FanOutShape2
FanOutShape20 FanOutShape21 FanOutShape22 FanOutShape3 FanOutShape4 FanOutShape5 FanOutShape6
FanOutShape7 FanOutShape8 FanOutShape9 FlowMonitor FlowMonitorState FlowShape Fusing Graph IgnoreBoth
IgnoreCancel IgnoreComplete Inlet InPort IOResult KillSwitch KillSwitches MaterializationContext MaterializationException
Materializer MaterializerLoggingProvider Outlet OutPort OverflowStrategy QueueOfferResult RateExceededException
Server Shape SharedKillSwitch SinkShape SourceShape StreamLimitReachedException
StreamSubscriptionTimeoutSettings StreamSubscriptionTimeoutTerminationMode StreamTcpException
SubstreamCancelStrategy Supervision ThrottleMode TLSClientAuth TLSClosing TLSProtocol TLSRole UniformFanInShape
UniformFanOutShape UniqueKillSwitch

13. Stream Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
ZIO STREAM

14. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Stream[E, A]
May fail with a checked error of type E
May effectfully produce zero or more values of type A

15. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Monoid
append
zero
Applicative
map
point
ap
Monad
bind
Bifunctor
leftMap
bimap
Alternative
alt
empty

16. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)
val empty: Stream[Nothing, Nothing] =
Stream.empty

17. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val singleton: Stream[Nothing, Int] =
Stream.point(42)
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

18. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val hundred: Stream[Nothing, Int] =
Stream.fromIterable(0 to 100)
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

19. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val userIn: Stream[IOException, String] =
Stream.liftIO(getStrLn)
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

20. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val intQueue: Queue[Int] = ...
val intStream: Stream[Nothing, Int] =
Stream.fromQueue(intQueue)
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

21. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

22. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val latency: Stream[Nothing, Long] =
Stream.unfoldM(0 -> 0) {
case (n, total) =>
api.measureLatency(l =>
(n + 1) -> (total + n))
}
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

23. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val fileBytes: Stream[IOException, Byte] =
Stream.bracket(openFile)(closeFile(_)) {
file => file.readByte.attempt.toOption
}
Stream.empty
Stream.point(v)
Stream.fromIterable(list)
Stream.liftIO(io)
Stream.fromQueue(queue)
Stream.unfold(s)(f)
Stream.unfoldM(s)(f)
Stream.bracket(open)(close)(read)

24. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val integerDigits: Stream[Nothing, Int] =
integers.map(_.toString).map(_.length)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

25. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val users: Stream[Exception, User] =
...
val usersWithProfile:
Stream[Exception, (User, Profile)] =
for {
user <- users
prof <- Stream.liftIO(getProfile(user.id))
} yield (user, prof)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

26. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val usersWithProfile:
Stream[Exception, (User, Profile)] = …
val targetUsers: Stream[Exception, User] =
usersWithProfile.filter(user =>
user.age >= 18 &&
user.age <= 32 &&
user.city == "San Francisco" &&
user.interests.contains("technology"))
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

27. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
type File = Stream[IOException, Report]
val monthlyFiles: List[File] = …
val allFiles =
monthlyFiles.foldLeft[File]
(Stream.empty)(_ ++ _)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

28. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val stream:
Stream[IOException, Chunk[Byte]] =
Stream.bracket(openFile)(
closeFile)(readChunk)
val usersJ: Stream[IOException, Json] =
stream.transduce(toJson)
.filter(isUserRecord)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

29. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val averages: Stream[Nothing, Int] =
integers.scan(0 -> 0) {
case ((sum, count), int) =>
(sum + int, count + 1) ->
if (count == 0) Int.MaxValue
else sum / count
}
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

30. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val first100Evens: Stream[Nothing, Int] =
integers.filter(_ % 2 == 0).take(100)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

31. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val evensUnder100: Stream[Nothing, Int] =
integers.filter(_ % 2 == 0)
.takeWhile(_ < 100)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

32. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val evensAfter100: Stream[Nothing, Int] =
integers.filter(_ % 2 == 0).drop(100)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

33. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val evensOver100: Stream[Nothing, Int] =
integers.filter(_ % 2 == 0)
.dropWhile(_ < 100)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

34. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val inputs: Stream[IOException, String] =
Stream.liftIO(getStrLn).forever
val echo: IO[IOException, Unit] =
inputs.foreach(putStrLn)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

35. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val integers: Stream[Nothing, Int] =
Stream.unfold(0)(i => (i, i + 1))
val users: Stream[Exception, User] = …
val usersWithId:
Stream[Exception, (Int, User)] =
integers.zip(users)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

36. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val stream1: Stream[IOException, JSON] =
kafka.stream("events1")
val stream2: Stream[IOException, JSON] =
kafka.stream("events2")
val both: Stream[IOException, JSON] =
stream1.merge(stream2)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

37. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val file1:
Stream[IOException, User] =
readFile("file1.data")
.transduce(toUser)
val file2:
Stream[IOException, User] =
readFile("file2.data")
.transduce(toUser)
val fastest: Stream[IOException, User] =
file1.joinWith(file2)(_ race _)
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

38. Stream
PERFORMANCEMOTIVATION API DESIGN SUMMARY
type StreamIO[A] = Stream[IOException, A]
val bytes: StreamIO[Byte] = …
val headersAndContent:
Managed[IOException,
(ContentType, StreamIO[Byte])] =
bytes.peel(parseContentType)
headersAndContent.use {
case (ContentType(Json), content) =>
content.transduce(toJson)...
case (ContentType(Xml), content) =>
content.transduce(toXml)...
}
stream.map(f)
stream.flatMap(f)
stream.filter(f)
stream1 ++ stream2
stream.transduce(what)
stream.scan(z)(f)
stream.take(n)
stream.takeWhile(f)
stream.drop(n)
stream.dropWhile(f)
stream.foreach(f)
stream1.zip(stream2)
stream1.merge(stream2)
stream1.joinWith(stream2)(f)
stream.peel(what)

39. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Sink[E, A0, A, B]
May fail with a checked error of type E
May effectfully produce a result of type B
Consumes values of type A
May possibly output a remainder of type A0

40. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
trait Stream[E, A] {
def run[R](sink: Sink[E, A, A, R]): IO[E, R]
}
myStream.run(Sink.drain) // IO[E, Unit]
myStream.run(decodeRequest) // IO[E, Request]
myStream.run(Sink.collect) // IO[E, List[A]]

41. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Monoid
append
zero
Applicative
map
point
ap
Monad
bind
Bifunctor
leftMap
bimap
Alternative
alt
empty
Category
id
compose
Profunctor
lmap
rmap
dimap
Strong
first
second
Choice
left
right

42. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val answer: Sink[Nothing, Nothing,
Any, Int] =
Sink.point(42)
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

43. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val lifted: Sink[Throwable, Nothing,
Any, Array[Byte]] =
Sink.liftIO(IO.syncThrowable(dangerousOp))
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

44. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val squared: Sink[Unit, Nothing,
Int, Int] =
Sink.lift(i => i * i)
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

45. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val failure: Sink[DomainError, Nothing,
Any, Nothing] =
Sink.fail(BadOperation)
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

46. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val json: Sink[Unit, Nothing,
Json, Json] =
Sink.await
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

47. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val drain: Sink[Nothing, Nothing,
Any, Unit] =
Sink.drain
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

48. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val collected: Sink[Nothing, Nothing,
Json, List[Json]] =
Sink.collect[Json]
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

49. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val parser: Sink[Nothing, Char,
Char, Either[Error, Int]] =
Sink.fold((ParserState.initial, List[Int]()) {
case (ParserState.Digits, c) if c.isDigit =>
// ...
}
Sink.point(value)
Sink.liftIO(ioValue)
Sink.lift(f)
Sink.fail(e)
Sink.await
Sink.drain
Sink.collect
Sink.fold(init)(step)

50. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val digits: Sink[Nothing, Char,
Char, List[Char]] =
Sink.readWhile[Char](_.isDigit)
val number: Sink[Nothing, Char,
Char, Int] =
Sink.map(_.mkString.toInt)
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

51. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val summing: Sink[Nothing, Int, Int, Int] =
Sink.fold(0)(_ + _)
case class Person(name: String, salary: Int)
val salarySum:
Sink[Nothing, Int, Person, Int] =
Sink.contramap[Person](_.salary)
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

52. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val err: Sink[DomainError, Nothing,
Any, Nothing] =
Sink.fail(DomainError)
val errs: Sink[AppError, Nothing,
Any, Nothing] =
Sink.mapError(AppError(_))
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

53. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val sumOfConsecutive:
Sink[Nothing, Nothing,
Int, Int] =
for {
a <- Sink.await
b <- Sink.await
} yield a + b
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

54. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val parseHeader: Sink[Err, Byte,
Byte, Header] =
…
val parseBody: Sink[Err, Byte,
Byte, Body] =
…
val parseHeaderAndBody:
Sink[Err, Byte, Byte, (Header, Body)] =
parseHeader ~ parseBody
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

55. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val parseHeader: Sink[Err, Byte,
Byte, Header] =
…
val maybeParseHeader:
Sink[Nothing, Byte,
Byte, Option[Header]] =
parseHeader ?
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

56. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val request1: Sink[Err, Byte,
Byte, Req1] = …
val request2: Sink[Err, Byte,
Byte, Req2] = …
val request:
Sink[Err, Byte,
Byte, Either[Req1, Req2]] =
request1 raceBoth request2
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

57. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val request1: Sink[Err, Byte,
Byte, Req1] = …
val request2: Sink[Err, Byte,
Byte, Req2] = …
val request:
Sink[Err, Byte,
Byte, Either[Req1, Req2]] =
request1 orElse request2
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

58. Sink
PERFORMANCEMOTIVATION API DESIGN SUMMARY
val parseUser: Sink[Err, Byte,
Byte, Person] = …
val parsePeople: Sink[Err, Byte,
Byte, List[Person]] =
parseUser repeat
sink.map(f)
sink.contramap(f)
sink.mapError(f)
sink.flatMap(f)
sink ~ otherSink
sink.?
sink.raceBoth(other)
sink.orElse(other)
sink.repeat

59. 3 How Fast Is Scalaz Stream
PERFORMANCE

60. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Start with stream of integers (chunked)
Filter for only even integers
Convert to longs
Sum all integers

61. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def akkaChunkFilterMapSum = {
val chunks = (1 to chunkCount).flatMap(i =>
Array.fill(chunkSize)(i))
val program = AkkaSource
.fromIterator(() => chunks.iterator)
.filter(_ % 2 == 0)
.map(_.toLong)
.toMat(AkkaSink.fold(0L)(_ + _))(Keep.right)
Await.result(program.run, Duration.Inf)
}

62. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY

63. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def fs2ChunkFilterMapSum = {
val chunks = (1 to chunkCount).map(i =>
FS2Chunk.array(Array.fill(chunkSize)(i)))
val stream = FS2Stream(chunks: _*)
.flatMap(FS2Stream.chunk(_))
.filter(_ % 2 == 0)
.map(_.toLong)
.covary[CatsIO]
.compile
.fold(0L)(_ + _)
stream.unsafeRunSync
}

64. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
FS2

65. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def scalazChunkFilterMapSum = {
val chunks = (1 to chunkCount).map(i =>
Chunk.fromArray(Array.fill(chunkSize)(i)))
val stream = StreamChunk
.fromChunks(chunks: _*)
.filter(_ % 2 == 0)
.map(_.toLong)
val sink = Sink.foldLeft(0L)((s, c) => c.foldl(s)(_ + _))
unsafeRun(stream.run(sink))
}

66. Filter · Map · Sum
PERFORMANCEMOTIVATION API DESIGN SUMMARY
FS2

67. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Start with stream of characters (chunked)
Identify CSV separators and columns
Convert character stream into token stream

68. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def akkaCsvTokenize() = {
val chunks = genCsvChunks
val program = AkkaSource
.fromIterator(() => chunks.flatten.iterator)
.scan((Vector.empty[Char], Vector.empty[CSV.Token])) {
case ((acc, _), char) =>
if (char == CSV.ColumnSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Vector(CSV.Column(acc.mkString))
else Vector.empty[CSV.Token]) ++
Vector(CSV.NewCol))
} else if (char == CSV.RowSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Vector(CSV.Column(acc.mkString))
else Vector.empty[CSV.Token]) ++
Vector(CSV.NewCol))
} else (acc :+ char) -> Vector.empty[CSV.Token]
}
.flatMapConcat(t => AkkaSource(t._2))
.toMat(AkkaSink.ignore)(Keep.right)
Await.result(program.run, Duration.Inf)
}

69. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY

70. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def fs2CsvTokenize() = {
val chunks = genCsvChunks.map(FS2Chunk.array(_))
val stream = FS2Stream(chunks: _*)
.flatMap(FS2Stream.chunk(_))
.scan((Vector.empty[Char], Vector.empty[CSV.Token])) {
case ((acc, _), char) =>
if (char == CSV.ColumnSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Vector(CSV.Column(acc.mkString))
else Vector.empty[CSV.Token]) ++
Vector(CSV.NewCol))
} else if (char == CSV.RowSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Vector(CSV.Column(acc.mkString))
else Vector.empty[CSV.Token]) ++
Vector(CSV.NewCol))
} else (acc :+ char) -> Vector.empty[CSV.Token]
}
.flatMap(t => FS2Stream(t._2))
.covary[CatsIO]
.compile
.drain
stream.unsafeRunSync
}

71. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
FS2

72. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
def scalazCsvTokenize() = {
val chunks = genCsvChunks.map(Chunk.fromArray(_))
val stream = ChunkedStream
.fromChunks(chunks: _*)
.scan[Vector[Char], Chunk[CSV.Token]](Vector.empty[Char]) {
case (acc, char) =>
if (char == CSV.ColumnSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Chunk(CSV.Column(acc.mkString))
else Chunk.empty) ++
Chunk(CSV.NewCol))
} else if (char == CSV.RowSep) {
Vector.empty[Char] ->
((if (acc.length > 0)
Chunk(CSV.Column(acc.mkString))
else Chunk.empty) ++
Chunk(CSV.NewCol))
} else (acc :+ char) -> Chunk.empty
}
.mapConcat(identity(_))
unsafeRun(stream.run(Sink.drain))
}

73. CSV Tokenize
PERFORMANCEMOTIVATION API DESIGN SUMMARY
FS2

74. 4 The Inner Workings of Scalaz Stream
DESIGN

75. BUILD YOUR
OWN STREAM
Initial Encoding
PERFORMANCEMOTIVATION API DESIGN SUMMARY

76. Initial Encoding
sealed trait Stream[+A] Type
Emission
Mapping
Joining
Merging
Folding
Effect
...
PERFORMANCEMOTIVATION API DESIGN SUMMARY

77. Initial Encoding
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

78. Initial Encoding
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
case class Map[A0, A](s: Stream[A0],
f: A0 => A) extends Stream[A]
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

79. Initial Encoding
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
case class Map[A0, A](s: Stream[A0],
f: A0 => A) extends Stream[A]
case class Join[A](s: Stream[Stream[A]])
extends Stream[A]
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

80. Initial Encoding
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
case class Map[A0, A](s: Stream[A0],
f: A0 => A) extends Stream[A]
case class Join[A](s: Stream[Stream[A]])
extends Stream[A]
case class Merge[A](l: Stream[A],
r: Stream[A]) extends Stream[A]
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

81. Initial Encoding
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
case class Map[A0, A](s: Stream[A0],
f: A0 => A) extends Stream[A]
case class Join[A](s: Stream[Stream[A]])
extends Stream[A]
case class Merge[A](l: Stream[A],
r: Stream[A]) extends Stream[A]
case class Fold[A0, A](a: A,
s: Stream[A0],
f: (A, A0) => A) extends Stream[A]
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

82. Initial Encoding
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...
sealed trait Stream[+A]
case class Emit[A](a: A) extends Stream[A]
case class Map[A0, A](s: Stream[A0],
f: A0 => A) extends Stream[A]
case class Join[A](s: Stream[Stream[A]])
extends Stream[A]
case class Merge[A](l: Stream[A],
r: Stream[A]) extends Stream[A]
case class Fold[A0, A](a: A,
s: Stream[A0],
f: (A, A0) => A) extends Stream[A]
case class Effect[A](io: IO[A]) extends Stream[A]

83. Initial Encoding
def drain[A](s: Stream[A]): IO[Unit] =
s match {
case Emit(a) => …
...
}
Interpreter
PERFORMANCEMOTIVATION API DESIGN SUMMARY

84. BUILD YOUR
OWN STREAM
Final Encoding
PERFORMANCEMOTIVATION API DESIGN SUMMARY

85. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit])
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

86. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit])
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

87. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit]) {
def map[B](f: A => B): Stream[B] =
Stream(r => run(r.compose(f)))
}
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

88. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit]) {
def map[B](f: A => B): Stream[B] =
Stream(r => run(r.compose(f)))
}
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
def join[A](s: Stream[Stream[A]]) =
Stream(r => s.run(sa => sa.run(r)))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

89. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit]) {
def map[B](f: A => B): Stream[B] =
Stream(r => run(r.compose(f)))
def merge[A1 >: A](that: Stream[A1]) =
Stream[A1](r =>
run(r).par(that.run(r)).void)
}
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
def join[A](s: Stream[Stream[A]]) =
Stream(r => s.run(sa => sa.run(r)))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

90. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit]) {
def map[B](f: A => B): Stream[B] =
Stream(r => run(r.compose(f)))
def merge[A1 >: A](that: Stream[A1]) =
Stream[A1](r =>
run(r).par(that.run(r)).void)
def fold[S](s: S)(f: (S, A) => S) = ...
}
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
def join[A](s: Stream[Stream[A]]) =
Stream(r => s.run(sa => sa.run(r)))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

91. Final Encoding
case class Stream[+A](
run: (A => IO[Unit]) => IO[Unit]) {
def map[B](f: A => B): Stream[B] =
Stream(r => run(r.compose(f)))
def merge[A1 >: A](that: Stream[A1]) =
Stream[A1](r =>
run(r).par(that.run(r)).void)
def fold[S](s: S)(f: (S, A) => S) = ...
}
object Stream {
def emit[A](a: => A): Stream[A] =
Stream[A](read => read(a))
def join[A](s: Stream[Stream[A]]) =
Stream(r => s.run(sa => sa.run(r)))
def liftIO[A](io: IO[A]): Stream[A] =
Stream[A](io.flatMap(_))
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Type
Emission
Mapping
Joining
Merging
Folding
Effect
...

92. Final
Flexible
Lookahead optimizations
Ad Hoc
Non-uniform
Initial
Initial vs Final
✓
✓
𐄂
PERFORMANCEMOTIVATION API DESIGN SUMMARY
𐄂
Inflexible
No lookahead optimizations
Structured
Uniform
𐄂
𐄂
✓
✓

93. Constant space processing
Discovered in 2003
Aggressively lazy
ITERATEES
PERFORMANCEMOTIVATION API DESIGN SUMMARY
Final encoding-friendly
Automatically leak-free
“Left folds”

94. Iteratees 101
sealed trait Iteratee[-A, +B]
case class More[A, B](f: A => Iteratee[A, B]) extends Iteratee[A, B]
case class Done[B](value: B) extends Iteratee[Any, B]
sealed trait Enumerator[+A] { self =>
def run[A1 >: A, B](iteratee: Iteratee[A1, B]): IO[B]
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

95. Iteratees 101
sealed trait Enumerator[+A] { self =>
final def run[A1 >: A, B](iteratee: Iteratee[A1, B]): IO[B] =
fold(...)(...)
def fold[B, S](s: S)(f: (S, A) => Either[S, B]): IO[B]
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

96. Iteratees 101
sealed trait Enumerator[+A] { self =>
final def run[A1 >: A, B](iteratee: Iteratee[A1, B]): IO[B] =
fold(...)(...)
def fold[B, S](s: S)(f: (S, A) => Either[S, B]): IO[B]
final def map[B](f0: A => B): Enumerator[B] =
new Enumerator[B] {
def fold[C, S](s: S)(f: (S, B) => Either[S, C]): IO[C] =
self.fold(s)((s, a) => f(s, f0(a)))
}
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

97. Iteratees 101
sealed trait Enumerator[+A] { self =>
final def run[A1 >: A, B](iteratee: Iteratee[A1, B]): IO[B] =
fold(...)(...)
def fold[B, S](s: S)(f: (S, A) => Either[S, B]): IO[B]
final def map[B](f0: A => B): Enumerator[B] =
new Enumerator[B] {
def fold[C, S](s: S)(f: (S, B) => Either[S, C]): IO[C] =
self.fold(s)((s, a) => f(s, f0(a)))
}
final def filter(f0: A => Boolean): Enumerator[A] =
new Enumerator[A] {
def fold[B, S](s: S)(f: (S, A) => Either[S, B]): IO[B] =
self.fold(s)((s, a) => if (f0(a)) f(s, a) else Left(s))
}
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

98. Iteratees 101
𐄂 Stack Safe𐄂 Combining 𐄂 Leftovers
PERFORMANCEMOTIVATION API DESIGN SUMMARY

99. Iteratees 101
𐄂 Stack Safe?𐄂 Combining? 𐄂 Leftovers?
PERFORMANCEMOTIVATION API DESIGN SUMMARY

100. Iteratees 101
PERFORMANCEMOTIVATION API DESIGN SUMMARY

101. Iteratees 101
PERFORMANCEMOTIVATION API DESIGN SUMMARY

102. Iteratees 101
PERFORMANCEMOTIVATION API DESIGN SUMMARY

103. Iteratees 101
sealed trait IterateeT[F[_], -A, +B]
case class More[F[_], A, B](f: A => IterateeT[F, A, B]) extends IterateeT[F, A, B]
case class Done[F[_], B](value: B) extends IterateeT[F, Any, B]
case class Effect[F[_], B](fx: F[B]) extends IterateeT[F, Any, B]
PERFORMANCEMOTIVATION API DESIGN SUMMARY

104. Iteratees 101
sealed trait IterateeT[F[_], -A, +B]
case class More[F[_], A, B](f: A => IterateeT[F, A, B]) extends IterateeT[F, A, B]
case class Done[F[_], B](value: B) extends IterateeT[F, Any, B]
case class Effect[F[_], B](fx: F[B]) extends IterateeT[F, Any, B]
type Parallel[F[_], A1, A2, B] =
IterateeT[IterateeT[F, A1, ?], A2, B]
def lift[F[_], A1, A2, B](it: IterateeT[F, A1, B]): Parallel[F, A1, A2, B] =
Effect[IterateeT[F, A1, ?], B](it)
PERFORMANCEMOTIVATION API DESIGN SUMMARY

105. Iteratees 101
sealed trait IterateeT[F[_], -A, +B]
case class More[F[_], A, B](f: A => IterateeT[F, A, B]) extends IterateeT[F, A, B]
case class Done[F[_], B](value: B) extends IterateeT[F, Any, B]
case class Effect[F[_], B](fx: F[B]) extends IterateeT[F, Any, B]
type Parallel[F[_], A1, A2, B] =
IterateeT[IterateeT[F, A1, ?], A2, B]
def lift[F[_], A1, A2, B](it: IterateeT[F, A1, B]): Parallel[F, A1, A2, B] =
Effect[IterateeT[F, A1, ?], B](it)
def readLeft[F[_], A1, A2]: Parallel[F, A1, A2, A1] =
lift(More(a => Done(a)))
def readRight[F[_], A1, A2]: Parallel[F, A1, A2, A2] =
More(a => Done(a))
PERFORMANCEMOTIVATION API DESIGN SUMMARY

106. Iteratees 101
sealed trait IterateeT[F[_], -A, +B]
case class More[F[_], A, B](f: A => IterateeT[F, A, B]) extends IterateeT[F, A, B]
case class Done[F[_], B](value: B) extends IterateeT[F, Any, B]
case class Effect[F[_], B](fx: F[B]) extends IterateeT[F, Any, B]
type Parallel[F[_], A1, A2, B] =
IterateeT[IterateeT[F, A1, ?], A2, B]
def lift[F[_], A1, A2, B](it: IterateeT[F, A1, B]): Parallel[F, A1, A2, B] =
Effect[IterateeT[F, A1, ?], B](it)
def readLeft[F[_], A1, A2]: Parallel[F, A1, A2, A1] =
lift(More(a => Done(a)))
def readRight[F[_], A1, A2]: Parallel[F, A1, A2, A2] =
More(a => Done(a))
PERFORMANCEMOTIVATION API DESIGN SUMMARY

107. ZIO Stream
trait Stream[E, A] {
def fold[S](s: S)(f: (S, A) => IO[E, Step[S]]): IO[E, Step[S]]
}
object Stream {
sealed trait Step[S]
object Step {
case class Stop[S](value: S) extends Step[S]
case class Cont[S](value: S) extends Step[S]
}
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

108. ZIO Stream
trait Stream[E, A] {
def fold[S](s: S)(f: (S, A) => IO[E, Step[S]]): IO[E, Step[S]]
}
object Stream {
sealed trait Step[S]
object Step {
case class Stop[S](value: S) extends Step[S]
case class Cont[S](value: S) extends Step[S]
}
}
Continuation Monad!
PERFORMANCEMOTIVATION API DESIGN SUMMARY

109. ZIO Stream
trait Stream[E, A] {
def fold[S](s: S)(f: (S, A) => IO[E, Step[S]]): IO[E, Step[S]]
def toQueue(capacity: Int = 1): Managed[Nothing, Queue[Take[E, A]]] =
for {
queue <- Managed.liftIO(Queue.bounded[Take[E, A]](capacity))
offerVal = (a: A) => queue.offer(Take.Value(a)).void
offerErr = (e: E) => queue.offer(Take.Fail(e))
enqueuer = (self.foreach[E, A](offerVal).catchAll(offerErr) *>
queue.offer(Take.End).forever).fork
_ <- Managed(enqueuer)(_.interrupt)
} yield queue
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

110. ZIO Stream
trait Stream[E, A] {
def fold[S](s: S)(f: (S, A) => IO[E, Step[S]]): IO[E, Step[S]]
def zip[B, C](that: Stream[E, B]): Stream[E, (A, B)] =
new Stream[E, (A, B)] {
type Q[A] = Queue[Take[E, A]]
def fold[S](s: S)(f: (S, (A, B)) => IO[E, Step[S]]): IO[E, Step[S]] = {
def loop(q1: Q[A], q2: Q[B], s: S): IO[E, Step[S]] =
Take.option(q1.take).seq(Take.option(q2.take)).flatMap {
case (Some(a), Some(b)) => f(s, (a, b))
case _ => IO.now(Step.Cont(s))
}
self.toQueue().use(q1 => that.toQueue().use(q2 => loop(q1, q2, s)))
}
}
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

111. ZIO Stream
𐄂 Stack Safe✓ Combining 𐄂 Leftovers
PERFORMANCEMOTIVATION API DESIGN SUMMARY

112. ZIO Stream
trait Stream[E, A] {
def fold[S](s: S)(f: (S, A) => IO[E, Step[S]]): IO[E, Step[S]]
def peel[R](sink: Sink[E, A, A, R]): Managed[E, (R, Stream[E, A])] =
...
}
Compose all leading sinks
Peel off sink (e.g. headers)
Work with stream remainder
PERFORMANCEMOTIVATION API DESIGN SUMMARY

113. ZIO Stream
𐄂 Stack Safe✓ Combining ✓ Leftovers
PERFORMANCEMOTIVATION API DESIGN SUMMARY

114. ZIO Stream
trait Stream[E, A] {
def fold[S]: IO[Nothing, (S, (S, A) => IO[E, Step[S]]) => IO[E, Step[S]]]
// trampolined: F[A => F[B]]
}
PERFORMANCEMOTIVATION API DESIGN SUMMARY

115. ZIO Stream
✓ Stack Safe✓ Combining ✓ Leftovers
PERFORMANCEMOTIVATION API DESIGN SUMMARY

116. 5 Where To From Here
SUMMARY

117. Fusion, Queue &
complex combinators
Performance
Outstanding Work
Comprehensive test
suites and great docs
Tests & Docs
Combinators for
every reason
Combinators
PERFORMANCEMOTIVATION API DESIGN SUMMARY

118. Eric Torreborre — Beautiful Folds
The FS2 team — Michael Pilquist, Fabio Labella, Pavel Chlupacek, et al
Special thanks to all people who made ZIO Stream possible.
Credits
Oleg Kiselyov — Iteratees 2003
PERFORMANCEMOTIVATION API DESIGN SUMMARY

119. Any questions?
Thanks!
John A. De Goes
@jdegoes - degoes.net
Itamar Ravid
@itrvd - iravid.com
PERFORMANCEMOTIVATION API DESIGN SUMMARY