Kotlin Coroutines and Rx

Kotlin coroutines and Rx
Shaul Rosenzweig
Fullstack as a
service

- I am part of Tikal's Android group. We
meet, share, contribute and code
together on a monthly basis… We have
Tech radar, FuseDays, weekly lightning
talks, monthly meetups...
WHO AM I?

- Coding since 1980es
WHO AM I ?

Professional coding since late 1990es,
mobile since 2001, PalmOS, WinCE,
Symbian, Brew, J2ME, and Android
since 2009

Currently working in Cellebrite,
working on an application 100%
written in Kotlin, using Reactive,
Dagger, etc.

We have always done it this way

James Gosling has a nicer beard than
Jake Wharton

Nobody died from boilerplate, you
spoilt brat!

Regular synchronous query
fun query(): List<CallLogDao> {
val cursor = context.contentResolver.query(
CallLog.Calls.CONTENT_URI
null, null, null, null)
return queryAllLogs(cursor)
}

Reactive with SQL Brite
fun queryRxSqbrite(): Observable<List<CallLogDao>> {
return sqBriteResolver.createQuery(
CallLog.Calls.CONTENT_URI,
null, null, null, null, false)
.mapToList({
c -> CallLogDao(c)
}).firstElement().toObservable()
}

Consuming rx Observable
@Test
fun testRxSqBrite() {
val latch = CountDownLatch(1)
var gotList = ArrayList<CallLogDao> ()
manager.queryRxSqbrite().subscribeOn(Schedulers.newThread())
.observeOn(AndroidSchedulers.mainThread())
.subscribe({
list: List<CallLogDao> ->
gotList.addAll( list)
latch.countDown()
})
latch.await()
assertTrue("got some logs", gotList.size > 0)
}

What are coroutines?
Light weight way of asynchronously execute
computations that can be suspended without
blocking the thread

How are they different than
threads?
● Much lighter than threads
● Suspending is not blocking
● Millions of coroutines can run in one thread
● Less context switch overhead
● Less memory overhead

Trivial example
suspend fun doALongQuery() { … }
suspend fun processResults() { … }
suspend fun uploadResultsToServer() { … }
fun saveSessionStatus(status) { … }
suspend fun doALotOfWork() {
val queryResult = doALongQuery()
val processedResults = processResults(queryResult)
val status = uploadResultsToServer(processedResults)
saveSessionStatus(status)
}

Coroutine builder
fun doALotOfWork() {
launch(UI) {
val queryResult = doALongQuery()
val processedResults = processResults(queryResult)
val status = uploadResultsToServer(processedResults)
saveSessionStatus(status)
}
}

Async with Anko coroutines
fun queryAsyncAnko() : Deferred<List<CallLogDao>>? {
return bg {
queryAllLogs(cursor)
}
} else {
return null
}
}

Consuming Anko deferred
@Test
fun testAsyncAnko() {
val deferredLogs = manager.queryAsyncAnko()
assertNotNull("is it null", deferredLogs?.await())
assertTrue("got some?", actualLogs!!.size > 0)
}

How do coroutines work?
Not a new JVM feature, implemented via compile
time bytecode changes.

Code turned into a state machine
When compiling, suspension points are inserted
into the code.
When suspended, state and variables are saved
and serialized.
Suspended coroutine is called continuation.

In Kotlin 1.2 they are opt-in
alpha feature

Async with Kotlin coroutines
suspend fun queryAsyncCoroutines(): Job? {
try {
job = launch(CommonPool) {
queryAllLogs(cursor, selection, selectionArgs)
}
} finally {
return job
}
}

What about Rx?
● Experimental lib, kotlinx coroutines reactive
● Wrappers for Reactive Streams, Reactor,
RxJava 1.x and RxJava 2.x using coroutines
● Under heavy development, and very
experimental
● At the time of writing, version is 0.18

Current status
IMPORTANT NOTE: We advise library authors to follow the
same convention: add the "experimental" (e.g.
com.example.experimental) suffix to your packages exposing
coroutine-based APIs so that your library remains binary
compatible. When the final API is released, follow these steps:

Rx2 example: flowable
return rxFlowable(coroutineContext) {
if (cursor.moveToFirst()) {
do {
send(CallLogDao(cursor))
} while (cursor.moveToNext())
}
}

Consuming rx2 Flowable
@Test
fun testRx2Coroutines() = runBlocking {
val source = manager.queryRxCoroutines(coroutineContext)
if (source != null) {
var success = false
source.observeOn(Schedulers.io(), false, 10)
.doOnComplete {
success = true
}
.subscribe {
log -> Log.d(TAG, "got " + log.toString())
} }
source.awaitLast()
assertTrue("got published via RX", success)
} else {
assertTrue("failed to generate RX stream", false)
}
}

But what is this coroutineContext?
Most important part is that they include a
coroutine dispatcher which determines on what
threads can the coroutine be executed.
They can be specified, or parent context can be
used.

Coroutine children
When a croutine is launched with context of
another coroutine, it becomes its child.
When a parent is cancelled, all of its children are
cancelled.

Waiting in coroutine context
If Thread.sleep(), CountdownLatch or similar
mechanism is used, it blocks the thread,
therefore coroutine does not get executed.
Instead use suspending, such as await, delay, or
similar.

Backend, Frontend, Mobile
and DevOps: Kotlin is fullstack

Links
● https://github.com/MaTriXy/mobileEvent-rx-kotlin
● https://github.com/Kotlin/kotlin-coroutines/
● https://github.com/Kotlin/kotlinx.coroutines/
● https://kotlin.github.io/kotlinx.coroutines/
● https://kotlinlang.org/docs/reference/coroutines.html
● https://kotlinlang.org/docs/tutorials/coroutines-basic-jvm.html
● https://github.com/Kotlin/anko/wiki/Anko-Coroutines
● Talk by Andrey Breslav
● Talk by Roman Elizarov
● Talk by Svetlana Isakova

THANK YOU
THANK YOU
Shaul Rosenzweig
Email: shaulr@tikalk.com
Tel: +972-52-644-5240

Kotlin Coroutines and Rx

More Related Content

What's hot

Similar to Kotlin Coroutines and Rx

More from Shaul Rosenzwieg

Recently uploaded

Kotlin Coroutines and Rx

Editor's Notes