Developing Reactive applications with Reactive Streams and Java 8

@bclozel @sdeleuze #Devoxx #reactive Developing Reactive applications with  Reactive Streams
and Java 8 Brian Clozel / Sébastien Deleuze Pivotal

@bclozel @sdeleuze #Devoxx #reactive Brian Clozel • Lyon, France + 
• Spring Framework  • Spring Boot  • @bclozel This guy

@bclozel @sdeleuze #Devoxx #reactive Sébastien Deleuze • Remote worker at
Pivota from Reactor Spring Framework 5 Kotlin support • conference staff member

@bclozel @sdeleuze #Devoxx #reactive Agenda • Introduction to Reactive •
Live coding with Reactor Core 3 • Coffee break (20 minutes) • Building a Reactive application with Spring Boot • Live coding

@bclozel @sdeleuze #Devoxx #reactive Why going Reactive? More for scalability
and stability than for speed

@bclozel @sdeleuze #Devoxx #reactive Reactive, what is it? Reactive is
used to broadly deﬁne event-driven systems

@bclozel @sdeleuze #Devoxx #reactive Reactive programming Moving imperative logic to
async, non-blocking, functional-style code, in particular when interacting with external resources

@bclozel @sdeleuze #Devoxx #reactive Use case: remote call with latency

@bclozel @sdeleuze #Devoxx #reactive Use case: serve a lot of
slow clients

@bclozel @sdeleuze #Devoxx #reactive Use case: push events to the
client Server-Sent Events or Websocket RabbitMQ broker Messages

@bclozel @sdeleuze #Devoxx #reactive Other use cases • Live database
queries • Mobile (Android) • Big Data • Real time analytics • HTTP/2 • …

@bclozel @sdeleuze #Devoxx #reactive Can’t we just use Java 8
lambdas? public class CallbackHell {    public void callbackHell() {    asyncMethod(a ->  asyncMethod(b ->  asyncMethod(c ->  asyncMethod(  d -> System.out.println("Values received: " + a + "," + b + "," + c + "," + d),  dEx -> System.err.println("An error occurred: " + dEx)  )  , cEx -> System.err.println("An error occurred: " + cEx))  ,bEx -> System.err.println("An error occurred: " + bEx)),  aEx -> System.err.println("An error occurred: " + aEx));  }    private void asyncMethod(Consumer<Object> success, Consumer<? super Throwable> failure) {  // ....  }    }

@bclozel @sdeleuze #Devoxx #reactive Can’t we just use Java 8
types? Type Non-blocking Streaming Future<T> CompletableFuture<T> X Stream<T> X InputStream / OutputStream X

@bclozel @sdeleuze #Devoxx #reactive We need tools Reactor Core RxJava
Reactive Streams Akka Streams Reactive APIs &

@bclozel @sdeleuze #Devoxx #reactive Reactive Streams • Reactive Streams is
a contract for asynchronous stream processing with non-blocking back pressure handling • De-facto standard for the behaviour of reactive libraries and for interoperability • Co-designed by Netﬂix, Lightbend, Pivotal, RedHat, Kaazing,  Twitter, and many others • Implemented by RxJava 2, Reactor, Akka Stream …

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber subscribe demand: 0

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber request(1) demand: 1

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber onNext(element1) demand: 0

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber Publisher is not
allowed to send new elements, even if new ones are ready. demand: 0

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber request(3) demand: 3

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber onNext(element2) demand: 2

@bclozel @sdeleuze #Devoxx #reactive onNext(element3) Back-pressure Publisher Subscriber demand: 1

@bclozel @sdeleuze #Devoxx #reactive Back-pressure Publisher Subscriber onComplete( ) demand:
1

@bclozel @sdeleuze #Devoxx #reactive Back-pressure • Allows to control the
amount of inﬂight data • Regulate the transfer between • Slow publisher and fast consumer • Fast publisher and slow consumer

@bclozel @sdeleuze #Devoxx #reactive Reactive Streams is 4 interfaces (and
a TCK) public interface Publisher<T> { void subscribe(Subscriber<? super T> s); } public interface Subscriber<T> { void onSubscribe(Subscription s); void onNext(T t); void onError(Throwable t); void onComplete(); } public interface Subscription { void request(long n); void cancel(); } public interface Processor<T, R> extends Subscriber<T>, Publisher<R> { }

@bclozel @sdeleuze #Devoxx #reactive Available in 2 distinct packages Same
interfaces are included in the upcoming Java 9 in java.util.concurrent: • Flow.Publisher • Flow.Subscriber • Flow.Subscription • Flow.Processor As a standalone JAR with org.reactivestreams package: • Publisher • Subscriber • Subscription • Processor

@bclozel @sdeleuze #Devoxx #reactive Reactive APIs on the JVM Apply
a wide range of transformations to your data with operators: merge, buffer, split, transform, delay … Reactor Core RxJava Akka Streams

@bclozel @sdeleuze #Devoxx #reactive Reactive APIs on the JVM (1/3)
Reactive API RxJava 1 Akka Stream 2 Reactor Core 3 RxJava 2 Types for  0..n elements Types for  0..1 elements Observable Single (1)  Completable (0) Source Sink Flow Flux Mono (0..1) Flowable Observable Single (1) Maybe (0..1)  Completable (0)

Reactive API RxJava 1 Akka Stream 2 Reactor Core 3 RxJava 2 Reactive Streams types Non Reactive  Streams types Observable Single  Completable Source Sink Flow Flux Mono Flowable Observable Single Maybe  Completable Limited back-pressure support

Reactive API RxJava 1 Akka Stream 2 Reactor Core 3 RxJava 2 Generation 2nd 3rd 4th 4th Support Limited back-pressure Reactive Streams +  actor fusion Reactive Streams +  operator fusion Reactive Streams +  operator fusion

Reactive API RxJava 1 Akka Stream 2 Reactor Core 3 RxJava 2 Generation 2nd 3rd 4th 4th Support Limited back-pressure Reactive Streams +  actor fusion Reactive Streams +  operator fusion Reactive Streams +  operator fusion Check out  http://akarnokd.blogspot.fr/2016/03/operator-fusion-part-1.html

@bclozel @sdeleuze #Devoxx #reactive RxJava 2 or Reactor Core 3?

@bclozel @sdeleuze #Devoxx #reactive Focus on Reactor Core 3 •
Natively designed on top of Reactive Streams • Lightweight API with 2 types: Mono and Flux • Native Java 8 support and optimisations • Single 1 Mbytes JAR • Focus on performance • Reactive foundation of Spring Framework 5

@bclozel @sdeleuze #Devoxx #reactive • Implements Reactive Streams Publisher •
0 to n elements • Operators: flux.map(…).zip(…).flatMap(…) Flux<T>

@bclozel @sdeleuze #Devoxx #reactive • Implements Reactive Streams Publisher •
0 to 1 element • Operators: mono.then(…).otherwise(…) Mono<T>

@bclozel @sdeleuze #Devoxx #reactive • Designed to test easily Reactive
Streams Publishers • Carefully designed after writing thousands of Reactor and Spring reactive tests StepVerifier StepVerifier.create(flux)  .expectNext("foo", "bar")  .expectComplete()  .verify();

@bclozel @sdeleuze #Devoxx #reactive Reactor 3 ecosystem

@bclozel @sdeleuze #Devoxx #reactive Upcoming high-level features like …

@bclozel @sdeleuze #Devoxx #reactive Live coding session • Creating Mono
and Flux • StepVeriﬁer • Transform: map() + ﬂatMap() • Merge • Request and Back pressure handling • Error handling • Convert & adapt: RxJava 2, List • Reactive to blocking and other way around

@bclozel @sdeleuze #Devoxx #reactive Flux.just(,☕); see you in 20 minutes!

@bclozel @sdeleuze #Devoxx #reactive Spring Reactive Web +

@bclozel @sdeleuze #Devoxx #reactive Building a Reactive application Building a
web application with: • Spring Boot Reactive starter (Experimental ⚠) • Spring Boot 2.0 (SNAPSHOT ) • Spring Framework 5.0 (MILESTONE ) • Reactor 3.0 (RELEASE )

@bclozel @sdeleuze #Devoxx #reactive

@bclozel @sdeleuze #Devoxx #reactive Framework Multiple runtimes

@bclozel @sdeleuze #Devoxx #reactive Live coding session • Calling remote
REST APIs • Dealing with datastores • Error handling • Rendering HTML views • Serving static resources • Server Sent Events • And more!

@bclozel @sdeleuze #Devoxx #reactive Our experience with Reactor 3

@bclozel @sdeleuze #Devoxx #reactive Changing your mindset Non-blocking when Thread/IO
boundary  (Servlet async IO, Java NIO Channels, …)

@bclozel @sdeleuze #Devoxx #reactive Changing your mindset Revisit/reconsider some concepts 
(ThreadLocal, Transactions, shared state, …)

@bclozel @sdeleuze #Devoxx #reactive Changing your mindset Common language with
your peers  (async, non-blocking, back-pressure, reactivex, reactive streams, operators…)

@bclozel @sdeleuze #Devoxx #reactive Reactive Streams 1. Don’t implement Reactive
Streams interfaces yourself  (or do it "for fun") 2. Reactive Streams by itself is not enough 3. Nothing happens until something subscribes to the Publisher

@bclozel @sdeleuze #Devoxx #reactive Concurrency 1. Debugging concurrency issues is
hard 2. concurrency Debugging is issues hard

@bclozel @sdeleuze #Devoxx #reactive Reactive ecosystem 1. Carefully select your
libraries / API / drivers 2. Tooling is important 3. Deﬁne additional code style rules in your team

@bclozel @sdeleuze #Devoxx #reactive Links • @ProjectReactor & https://projectreactor.io •
https://github.com/reactor/lite-rx-api-hands-on  • https://github.com/bclozel/spring-reactive-university • https://spring.io • https://spring.io/blog/2016/04/19/understanding-reactive-types

@bclozel @sdeleuze #Devoxx #reactive Reactive Web Applications with Spring 5
Rossen Stoyanchev Friday 9:30, Room 8 +

@bclozel @sdeleuze #Devoxx #reactive The Spring BOF Spring team Wednesday
7:00 PM, BOF2

@bclozel @sdeleuze #Devoxx #reactive Questions?

Developing Reactive applications with Reactive ...

Developing Reactive applications with Reactive Streams and Java 8

More Decks by Sébastien Deleuze

Other Decks in Programming

Featured

Transcript