FS2 - Fake it till you make it

class: center, middle, inverse

# Fake it till you make it with FS2

James Collier @JRCCollier

![FS2 Logo](./assets/fs2-logo.png)

---

## Playback Session

Telemetry events will be emitted from devices.

```scala
import java.time.Instant
import java.util.UUID

sealed trait PlaybackEvent {
  def sessionId: UUID
}
final case class Started(timestamp: Instant, sessionId: UUID) extends PlaybackEvent
final case class Paused(timestamp: Instant, sessionId: UUID) extends PlaybackEvent
final case class Resumed(timestamp: Instant, sessionId: UUID) extends PlaybackEvent
final case class Ended(timestamp: Instant, sessionId: UUID) extends PlaybackEvent
```

---

## 1000ft view

![device to cloud](./assets/device-to-cloud.png)

---
class: inverse

## The catch

😱 it's going to be a while till devices are instrumented!

The show must go on!

Need to develop while application teams are instrumenting.

---

## Fake it till we instrument it

We have the ~~technology~~ schema, we can build the events.

![gen to cloud](./assets/gen-to-cloud.png)

---

## Let's just use scalacheck!

```scala
import org.scalacheck.ScalacheckShapeless._
import org.scalacheck.{Arbitrary, Gen}

private implicit val arbInstant: Arbitrary[Instant] =
    Arbitrary(Gen.calendar.map(_.toInstant))
val arbPlaybackEvt = implicitly[Arbitrary[PlaybackEvent]]
```

Now let's generate some events...

```scala
List.fill(10)(arbPlaybackEvt.arbitrary.sample).flatten
// res0: List[PlaybackEvent] = List(
//   Resumed(+91560516-12-31T00:00:00Z, 56d509d0-793f-48d9-bf00-a083667702c6),
//   Ended(3914-06-30T09:02:45.926Z, b98faff0-1da7-4dcd-a441-89a91b5c5a30),
//   Paused(2979-04-10T23:59:59.059Z, 022eb5a0-60e0-478b-9bc3-b649603533b7),
//   Resumed(3014-06-21T21:57:33.788Z, fd420381-3ee5-4c3b-af56-994fce5656a8),
//   Paused(1592-10-27T00:00:00Z, a64e75d1-c98d-4449-a8e5-b37bfe877937),
//   Resumed(1969-12-31T23:59:59.999Z, 78bb87ae-93b8-48be-9703-416d8bd1cb92),
//   Started(4000-11-03T20:17:29.116Z, 48de9b7b-821d-4864-8f48-587f6b5e58fe),
//   Paused(1258-09-23T12:00:29.469Z, b63d855f-b086-4d4a-a242-fda7d767edb8),
//   Started(0166-03-14T10:43:00.095Z, 60b91ad1-300e-4da7-9dae-5cfc01ae4f63),
//   Resumed(0123-12-31T23:19:58.852Z, a9d854e4-90cd-4681-851c-951f4ebab8ae)
// )
```

😱 not useful!

???

* timestamp is nonsense.
* each event has it's own session.
* not a representative lifecycle.

---

## Requirements

* Sessions modelled.
* Representative delays between events.
* Configure the number of concurrent sessions.

Maybe this can be modelled as streams?

Step in FS2:

![FS2 Logo](./assets/fs2-logo.png)

```scala
import cats.effect._
import fs2.Stream
import scala.concurrent.ExecutionContext
import scala.concurrent.duration._
```

---

## Model a playback scenario

```scala
def session: Stream[IO, PlaybackEvent] = for {
  now <- Stream.emit(Instant.parse("2019-11-26T12:00:00Z"))
  sessionId <- Stream.eval(IO(UUID.randomUUID()))
  s <- Stream.emits(List(
    Started(now, sessionId),
    Ended(now.plusSeconds(60), sessionId)
  ))
} yield s
```

```scala
session.compile.toList.unsafeRunSync()
// res2: List[PlaybackEvent] = List(
//   Started(2019-11-26T12:00:00Z, c4abd44d-c1f8-4a76-a81c-9a885aa57d89),
//   Ended(2019-11-26T12:01:00Z, c4abd44d-c1f8-4a76-a81c-9a885aa57d89)
// )
```

---

## Nap between events 💤💤💤

Useful to model the progress of time between emitting
events for a session.

```scala
def nowS: Stream[IO, Instant] =
  Stream.eval(timer.clock.realTime(MILLISECONDS).map(Instant.ofEpochMilli))

def idealSession: Stream[IO, PlaybackEvent] = for {
  sessionId <- Stream.eval(IO(UUID.randomUUID()))
  s <- nowS.map(Started(_, sessionId)) ++
    nowS.map(Ended(_, sessionId)).delayBy(30.minutes)
} yield s
```

```scala
idealSession
  .evalMap(event => IO(println(event)))
  .compile
  .drain
  .unsafeRunAsyncAndForget()
// Started(2018-12-20T00:00:00Z,6d03b9a5-4e5c-422b-a7dc-a726878730e7)
testContext.tick(1.hour)
// Ended(2018-12-20T00:30:00Z,6d03b9a5-4e5c-422b-a7dc-a726878730e7)
```

---

## A stream of sessions

![Session streams](./assets/session-streams.png)

???

Par join to choose how many concurrent sessions to have.

---

## Concurrent sessions

```scala
val numConcurrent = 3

Stream.emit(idealSession)
  .repeat
  .take(5) // limit to 5 for example
  .parJoin(numConcurrent)
  .evalMap(event => IO(println(event)))
  .compile
  .drain
  .unsafeRunAsyncAndForget()
```

```scala
testContext.tick(1.hour)
// Started(2018-12-20T01:00:00Z,a1f0f374-84b0-4ad5-a716-ab631948a177)
// Started(2018-12-20T01:00:00Z,d51c085f-af3a-4de9-ac34-1f09010f7e5b)
// Started(2018-12-20T01:00:00Z,79a86935-87fe-48b4-a55d-0cf8dbf28dd9)
// Ended(2018-12-20T01:30:00Z,a1f0f374-84b0-4ad5-a716-ab631948a177)
// Ended(2018-12-20T01:30:00Z,79a86935-87fe-48b4-a55d-0cf8dbf28dd9)
// Ended(2018-12-20T01:30:00Z,d51c085f-af3a-4de9-ac34-1f09010f7e5b)
// Started(2018-12-20T01:30:00Z,69a8b6ff-1ca3-4eed-9411-0ffc31775791)
// Started(2018-12-20T01:30:00Z,3fecff2a-db5b-42f9-bfb5-a525f1948d1b)
// Ended(2018-12-20T02:00:00Z,69a8b6ff-1ca3-4eed-9411-0ffc31775791)
// Ended(2018-12-20T02:00:00Z,3fecff2a-db5b-42f9-bfb5-a525f1948d1b)
```

???

Sessions are repeated forever.

---

## Add another scenario!

```scala
def pauseSession: Stream[IO, PlaybackEvent] = for {
  sessionId <- Stream.eval(IO(UUID.randomUUID()))
  s <- nowS.map(Started(_, sessionId)) ++
    nowS.map(Paused(_, sessionId)).delayBy(20.seconds) ++
    nowS.map(Resumed(_, sessionId)).delayBy(5.seconds) ++
    nowS.map(Ended(_, sessionId)).delayBy(20.minutes)
} yield s
```

---

## Running multiple scenarios

```scala
Stream.emits(List(idealSession, pauseSession))
  .repeat
  .take(4) // limit to 4 for example
  .parJoin(numConcurrent)
  .evalMap(event => IO(println(event)))
  .compile
  .drain
  .unsafeRunAsyncAndForget()
```

```scala
testContext.tick(1.hour)
// Started(2018-12-20T02:00:00Z,b66585ea-b3d3-456e-9342-deb207cb9925)
// Started(2018-12-20T02:00:00Z,242f6e38-50f3-4fd2-ac6d-0951d77fa66a)
// Started(2018-12-20T02:00:00Z,6c42a426-8aa4-4b69-8e4f-413ed09502b5)
// Paused(2018-12-20T02:00:20Z,b66585ea-b3d3-456e-9342-deb207cb9925)
// Resumed(2018-12-20T02:00:25Z,b66585ea-b3d3-456e-9342-deb207cb9925)
// Ended(2018-12-20T02:20:25Z,b66585ea-b3d3-456e-9342-deb207cb9925)
// Started(2018-12-20T02:20:25Z,43ecebca-c30a-499d-9dfd-5c1425783555)
// Paused(2018-12-20T02:20:45Z,43ecebca-c30a-499d-9dfd-5c1425783555)
// Resumed(2018-12-20T02:20:50Z,43ecebca-c30a-499d-9dfd-5c1425783555)
// Ended(2018-12-20T02:30:00Z,242f6e38-50f3-4fd2-ac6d-0951d77fa66a)
// Ended(2018-12-20T02:30:00Z,6c42a426-8aa4-4b69-8e4f-413ed09502b5)
// Ended(2018-12-20T02:40:50Z,43ecebca-c30a-499d-9dfd-5c1425783555)
```

---

## Testing

Can use `cats.effect.laws.util.TestContext` to simulate time.

```scala
val f = idealSession
  .compile
  .toList
  .unsafeToFuture()
```

```scala
f.value
// res9: Option[scala.util.Try[List[PlaybackEvent]]] = None
```

```scala
testContext.tick(1.hour)
f.value
// res11: Option[scala.util.Try[List[PlaybackEvent]]] = Some(
//   Success(
//     List(
//       Started(2018-12-20T03:00:00Z, 1dfa54a3-fb6d-45ca-a8a1-bc0b4a64e3b5),
//       Ended(2018-12-20T03:30:00Z, 1dfa54a3-fb6d-45ca-a8a1-bc0b4a64e3b5)
//     )
//   )
// )
```

---

## Counting sessions

```scala
val f2 = Stream.emit(idealSession)
  .repeat
  .parJoin(1000)
  .interruptAfter(1.hour)
  .compile
  .toList
  // count sessions
  .map(_.groupBy(_.sessionId).keySet.size)
  .unsafeToFuture()
```

`idealSession` lasts 30 minutes, so after 1 hour there should be ~2000 sessions.

```scala
testContext.tick(1.hour)
f2.value
// res13: Option[scala.util.Try[Int]] = Some(Success(2000))
```
👌

---

## Further reading and links

* [FS2](https://github.com/functional-streams-for-scala/fs2) by [@mpilquist](https://github.com/mpilquist)
* [cats-effect](https://github.com/typelevel/cats-effect)

## Thanks!

* [mdoc](https://github.com/scalameta/mdoc) used to evaluate scala examples

## Questions?