RxJava, RxJava 2, Reactor État de l’art des Reactive Streams sur la JVM

David Wursteisen

No content

Écrire du code asynchrone : 
 ça craint

Future ExecutorService ex = Executors.newCachedThreadPool();
 Future future = ex.submit(() -> longProcessing());
 String result = future.get(); Appel bloquant

Future Future> future1 = /* ... */
 Future> future2 = /* ... */
 Future> future3 = /* ... */
 Future> future4 = /* ... */
 Future> future5 = /* ... */ Orchestration optimale ?

Callback RemoteService service = buildRemoteService();
 service.getUser(id -> {
 service.getData(id, data -> {
 service.getSomething(data, whut -> { service.neverEndingCallBack(whut, () -> { }); });
 });
 }); Callback Hell

Relationship Status:
 it’s complicated

Le problème du code synchrone

Synchrone Attente de la réponse Appelant Appelé

Impacte le temps de réponse Serveur lent

Il y a toujours un truc qui ne marche pas ça m’énerve !

No content

Ça ne marche pas ! ça a planté ? dans le doute, reboot

Asynchrone Appelant Appelé Traitement continu

L’asynchrone permet de profiter du temps d’attente

Impact minimum sur le temps de réponse

No content

Écrire du code asynchrone facilement ?

Émergence 
 de différentes approches

2 Reactive Streams

Reactive Streams API Reactive Streams RxJava 2 Reactor Interface Implémentation

L’API Reactive Streams est 
 un pont 
 entre les implémentations

Contrat Reactive Streams onSubscribe onNext onError onComplete

RxJava n’est pas compatible 
 avec Reactive Streams (il faut alors utiliser un adapteur : RxJavaReactiveStreams) https://github.com/ReactiveX/RxJavaReactiveStreams

onNext * ( onError | onComplete ) 
 onNext * ( onError | onCompleted ) RxJava Reactive Streams

onNext * ( onError | onComplete ) 
 onNext * ( onError | onCompleted ) RxJava Reactive Streams Nom différent

Reactor peut utiliser RxJava 2 et réciproquement

Approche commune

API pour manipuler 
 des événements de manière synchrone ou asynchrone à travers un flux d’événements

No content

Map ( ⃝ → ⬛)

No content

No content

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println);

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println);

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Push du résultat

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Récupération des véhicules de Luke Récupération des vaisseaux de Luke

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Flux 
 d’événements Flux 
 d’événements

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Fusion des 2 flux

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Déclenche l’exécution du code

remoteApi.people(1).flatMap(luke -> { 
 Observable vehicles = Observable.from(luke.getVehiclesIds())
 .flatMap(remoteApi::vehicle)
 .map(vehicle -> luke.getName() + " can drive " + vehicle.getName());
 
 Observable starships = Observable.from(luke.getStarshipsIds())
 .flatMap(remoteApi::starship)
 .map(starship -> luke.getName() + " can fly with " + starship.getName());
 
 return Observable.merge(vehicles, starships); 
 }).subscribe(System.out::println); Flux 
 d’événements

No content

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) 
 .observeOn(Schedulers.io()) 
 .switchMap(evt -> remoteApi.getData())
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe();

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData())
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Écoute des clicks

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData())
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Changement du contexte d’exécution

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData()) // Observable
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Appel asynchrone d’un web service

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData()) // Observable
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Changement du contexte d’exécution

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData()) // Observable
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Changement au niveau de l’UI

Button btn = new Button();
 btn.setText("Click Me");
 
 JavaFx.fromClick(btn) // Observable
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData()) // Observable
 .observeOn(javaFx()) 
 .doOnNext(value -> btn.setText("Data: " + value))
 .subscribe(); Flux 
 d’événements

Button btn = new Button();
 btn.setText("Click Me");
 
 ExecutorService io = Executors.newSingleThreadExecutor();
 btn.setOnAction(evt -> {
 io.submit(() -> {
 remoteApi.getData(data -> {
 Platform.runLater(() -> {
 btn.setText(data);
 });
 });
 });
 }); Sac de noeud Structure pyramidale

Grâce à RxJava et Reactor …

Écrire du code asynchrone : 
 ça craint

Il était une fois…

Création des 
 Reactive Extensions Création de RxJava Participation active Reprise de 
 RxJava & RxJava 2 Création de Reactor Participation à Reactor

RxJava 
 est une technologie 
 éprouvée

Reactor
 profite de l’expérience de 
 RxJava (et vice versa)

Types d’objet

Observable

No content

No content

No content

No content

No content

Single

No content

Completable

No content

RxJava Contrat Backpressure Observable [N] Oui Single [1] Non Completable [0] Non Ajouté après coup Appel Web service Processus d’arrière plan

Écouter une websocket, pour c h a q u e c o m m a n d e r e ç u e , composer une réponse à partir de 3 webservices différents, puis exécuter 2 jobs séquentiellement ?

Écouter une websocket, pour c h a q u e c o m m a n d e r e ç u e , composer une réponse à partir de 3 webservices différents, puis exécuter 2 jobs séquentiellement ?

Écouter une websocket, pour c h a q u e c o m m a n d e r e ç u e , composer une réponse à partir de 3 webservices différents, puis exécuter 2 jobs séquentiellement ?

Écouter une websocket, pour c h a q u e c o m m a n d e r e ç u e , composer une réponse à partir de 3 webservices différents, puis exécuter 2 jobs séquentiellement ?

websocket("/topics/cmd")
 .observeOn(Schedulers.io())
 .switchMap(cmd ->
 Single.zip(
 api.getActions(),
 api.getScore(),
 api.getUserData(),
 this::composeResult).toObservable())
 .observeOn(Schedulers.computation())
 .concatMap(result -> updateDb(result).andThen(getLastResults()))
 .subscribe(last -> System.out.println("last results -> " + last));

websocket("/topics/cmd")
 .observeOn(Schedulers.io())
 .switchMap(cmd ->
 Single.zip(
 api.getActions(),
 api.getScore(),
 api.getUserData(),
 this::composeResult).toObservable())
 .observeOn(Schedulers.computation())
 .concatMap(result -> updateDb(result).andThen(getLastResults()))
 .subscribe(last -> System.out.println("last results -> " + last)); Écoute de la websocket Observable

websocket("/topics/cmd")
 .observeOn(Schedulers.io())
 .switchMap(cmd ->
 Single.zip(
 api.getActions(),
 api.getScore(),
 api.getUserData(),
 this::composeResult).toObservable())
 .observeOn(Schedulers.computation())
 .concatMap(result -> updateDb(result).andThen(getLastResults()))
 .subscribe(last -> System.out.println("last results -> " + last)); Composition d’une réponse à partir de 3 webservices Single

websocket("/topics/cmd")
 .observeOn(Schedulers.io())
 .switchMap(cmd ->
 Single.zip(
 api.getActions(),
 api.getScore(),
 api.getUserData(),
 this::composeResult).toObservable())
 .observeOn(Schedulers.computation())
 .concatMap(result -> updateDb(result).andThen(getLastResults()))
 .subscribe(last -> System.out.println("last results -> " + last)); Completable Exécuter 2 jobs

websocket("/topics/cmd")
 .observeOn(Schedulers.io())
 .switchMap(cmd ->
 Single.zip(
 api.getActions(),
 api.getScore(),
 api.getUserData(),
 this::composeResult).toObservable())
 .observeOn(Schedulers.computation())
 .concatMap(result -> updateDb(result).andThen(getLastResults()))
 .subscribe(last -> System.out.println("last results -> " + last)); Écoute de la websocket Observable Composition d’une réponse à partir de 3 webservices Single Completable Exécuter 2 jobs

RxJava 2 Contrat Backpressure Observable [N] Non Single [1] Non Completable [0] Non Maybe [0|1] Non Nouveau ! Concept « similaire » à Optional

Maybe

No content

Backpressure avec RxJava 2

Consommateur

Consommateur ça déborde !

backpressure

MissingBackpressureException ?

acceptedIntent
 .filter(intent -> !intent.getBooleanExtra("UpdatePhoneMode", false))
 .concatMap(intent -> approximatedEngine.detectCurrentPlace())
 .doOnNext(score -> Log.info(TAG, "Scan completed with result " + score))
 .concatMap(this::detectSleepMode)
 .concatMap((score) -> isNewPlace(score.getScore().getPlace()).map(p -> score))
 .doOnNext((p) -> Log.info(TAG, "Current place found is : " + p))
 .subscribe()

acceptedIntent
 .filter(intent -> !intent.getBooleanExtra("UpdatePhoneMode", false)) .onBackpressureDrop()
 .concatMap(intent -> approximatedEngine.detectCurrentPlace())
 .doOnNext(score -> Log.info(TAG, "Scan completed with result " + score))
 .onBackpressureDrop() .concatMap(this::detectSleepMode)
 .onBackpressureDrop()
 .concatMap((score) -> isNewPlace(score.getScore().getPlace()).map(p -> score))
 .doOnNext((p) -> Log.info(TAG, "Current place found is : " + p))
 .subscribe() Ajouté dans un moment de panique

RxJava 2 Contrat Backpressure Observable [N] Non Single [1] Non Completable [0] Non Maybe [0|1] Non Nouveau ! Concept « similaire » à Optional

RxJava 2 Contrat Backpressure Observable [N] Non Single [1] Non Completable [0] Non Maybe [0|1] Non Flowable [N] Oui Nouveau ! Observable avec 
 back pressure Concept « similaire » à Optional

Observable → moins de 1000 éléments → Gestion d’élément d’interface graphique → Remplacement de Stream Java Flowable → plus de 10 000 éléments → contrôle de l’émission de données → Flux réseau avec contrôle de débit

Que propose Reactor ?

Flux Mono Maximum 1 élément

Reactor Contrat Backpressure Flux [N] Oui Mono [0|1] Oui Identique à Flowable Flux avec 
 au maximum 
 1 élément

Types d’objets et
 Reactive Streams

Publisher Flux Flowable Reactive Streams

Flux.range(1, 10)
 .flatMap(i -> Flux.just(1))
 .subscribe();

Flux.range(1, 10)
 .flatMap(i -> Flux.just(1))
 .subscribe(); Publisher

Flux.range(1, 10)
 .flatMap(i -> Flowable.just(1))
 .subscribe(); Publisher RxJava 2

Flowable.defer(() -> Flux.range(1, 10))
 .subscribe(System.out::println); RxJava 2 Reactor

Flux.defer(() -> Flowable.range(1, 10))
 .subscribe(System.out::println); RxJava 2 Reactor

Encore plus de possibilités avec reactor-addons https://github.com/reactor/reactor-addons

Observable.range(1, 10)
 .to(obs -> RxJava2Adapter.observableToFlux(obs, BackpressureStrategy.BUFFER))
 .checkpoint()
 .subscribe(); Maybe.empty()
 .to(maybe -> RxJava2Adapter.maybeToMono(maybe))
 .checkpoint()
 .subscribe(); Single.just(1)
 .to(single -> RxJava2Adapter.singleToMono(single))
 .checkpoint()
 .subscribe(); Observable → Flux Single → Mono Maybe → Mono

RxJava2Adapter.monoToFlowable(Mono.just(1))
 .timestamp()
 .subscribe();
 
 RxJava2Adapter.fluxToFlowable(Flux.range(1, 10))
 .timestamp()
 .subscribe(); Mono → Flowable Flux → Flowable

P o u v o i r u t i l i s e r u n e biblio t hèq ue utilisant RxJava 2 dans son projet Reactor (et vice versa)

Opérateurs

Catalogue conséquent et homogène

all amb ambArray ambWith any as awaitOnSubscribe blockFirst blockFirstMillis blockLast blockLastMillis blockingFirst blockingForEach blockingIterable blockingLast blockingLatest blockingMostRecent blockingNext blockingSingle blockingSubscribe buffer bufferMillis bufferSize bufferTimeout bufferTimeoutMillis bufferUntil bufferWhile cache cacheWithInitialCapacity cancelOn cast checkpoint collect collectInto collectList collectMap collectMultimap collectSortedList combineLatest combineLatestDelayError compose concat concatArray concatArrayDelayError concatArrayEager concatDelayError concatEager concatMap concatMapDelayError concatMapEager concatMapEagerDelayError concatMapIterable concatWith contains count create debounce defaultIfEmpty defer delay delayElements delayElementsMillis delayMillis delaySubscription delaySubscriptionMillis dematerialize distinct distinctUntilChanged doAfterNext doAfterTerminate doFinally doOnCancel doOnComplete doOnEach doOnError doOnLifecycle doOnNext doOnRequest doOnSubscribe doOnTerminate elapsed elementAt elementAtOrError empty equals error filter first firstElement firstEmitting firstEmittingWith firstOrError flatMap flatMapCompletable flatMapIterable flatMapMaybe flatMapSequential flatMapSingle forEach forEachWhile from fromArray fromCallable fromFuture fromIterable fromPublisher fromStream generate getClass getPrefetch groupBy groupJoin handle hasElement hasElements hashCode hide ignoreElements interval intervalMillis intervalRange isEmpty join just last lastElement lastOrError lift limitRate log map mapError materialize merge mergeArray mergeArrayDelayError mergeDelayError mergeSequential mergeWith never next notify notifyAll observeOn ofType onBackpressureBuffer onBackpressureDrop onBackpressureError onBackpressureLatest onErrorResumeNext onErrorResumeWith onErrorReturn onErrorReturnItem onExceptionResumeNext onTerminateDetach parallel publish publishNext publishOn range rangeLong rebatchRequests reduce reduceWith repeat repeatUntil repeatWhen replay replayMillis retry retryUntil retryWhen safeSubscribe sample sampleFirst sampleFirstMillis sampleMillis sampleTimeout scan scanWith sequenceEqual serialize share single singleElement singleOrEmpty singleOrError skip skipLast skipMillis skipUntil skipUntilOther skipWhile sort sorted startWith startWithArray strict subscribe subscribeOn subscribeWith switchIfEmpty switchMap switchMapDelayError switchOnError switchOnNext switchOnNextDelayError take takeLast takeMillis takeUntil takeUntilOther takeWhile test then thenEmpty thenMany throttleFirst throttleLast throttleWithTimeout timeInterval timeout timeoutMillis timer timestamp to toFuture toIterable toList toMap toMultimap toObservable toSortedList toStream toString transform unsafeCreate unsubscribeOn using wait window windowMillis windowTimeout windowTimeoutMillis windowUntil windowWhile withLatestFrom zip zipArray zipIterable zipWith zipWithIterable

RxJava RxJava 2 Reactor flatMap flatMap flatMap Émission d’aucune, une ou plusieurs valeurs amb amb firstEmitting Émission des valeurs du flux répondant en premier … … … … debounce debounce N/A Ignore les événements pendant un laps de temps

RxJava RxJava 2 Reactor flatMap flatMap flatMap Émission d’aucune, une ou plusieurs valeurs amb amb firstEmitting Émission des valeurs du flux répondant en premier … … … … debounce debounce N/A Ignore les événements pendant un laps de temps Renommage

L’ensemble des opérateurs couvre de nombreuses situations

Nota bene

https://github.com/ReactiveX/RxJava/wiki/Implementing-custom-operators-(draft) writing operators is hard when one writes an operator, the Observable protocol, unsubscription, backpressure and concurrency have to be taken into account and adhered to the letter

Écrire un opérateur avec RxJava 2 est plus complexe qu’avec RxJava

Rendre son application Reactive

-Jake Wharton « Unless you can model your entire system synchronously, a single asynchronous source breaks imperative programming »

-Jake Wharton « Unless you can model your entire system synchronously, a single asynchronous source breaks imperative programming »

Reactive Reactive API Synchrone Callback Reactive Reactive

Utilisation de Factory

RxJava
 RxJava 2 Reactor Flowable.just Flux.just Émission d’une valeur existante Flowable.defer Flux.defer Émission tardive d’un Observable Flowable.fromCallable Mono.fromCallable Émission à partir d’un appel de fonction Flowable.create Flux.create Contrôle manuel de l’émission des événements Flowable.using Flux.using Gestion d’une ressource Flowable.fromPublisher Flux.from Utilisation d’un Publisher (Reactive Streams) Flowable.generate Flux.generate Utilisation d’un générateur de valeurs

RxJava
 RxJava 2 Reactor Flowable.just Flux.just Émission d’une valeur existante Flowable.defer Flux.defer Émission tardive d’un Observable Flowable.fromCallable Mono.fromCallable Émission à partir d’un appel de fonction Flowable.create Flux.create Contrôle manuel de l’émission des événements Flowable.using Flux.using Gestion d’une ressource Flowable.fromPublisher Flux.from Utilisation d’un Publisher (Reactive Streams) Flowable.generate Flux.generate Utilisation d’un générateur de valeurs

RxJava
 RxJava 2 Reactor Flowable.just Flux.just Émission d’une valeur existante Flowable.defer Flux.defer Émission tardive d’un Observable Flowable.fromCallable Mono.fromCallable Émission à partir d’un appel de fonction Flowable.create Flux.create Contrôle manuel de l’émission des événements Flowable.using Flux.using Gestion d’une ressource Flowable.fromPublisher Flux.from Utilisation d’un Publisher (Reactive Streams) Flowable.generate Flux.generate Utilisation d’un générateur de valeurs

RxJava
 RxJava 2 Reactor Flowable.just Flux.just Émission d’une valeur existante Flowable.defer Flux.defer Émission tardive d’un Observable Flowable.fromCallable Mono.fromCallable Émission à partir d’un appel de fonction Flowable.create Flux.create Contrôle manuel de l’émission des événements Flowable.using Flux.using Gestion d’une ressource Flowable.fromPublisher Flux.from Utilisation d’un Publisher (Reactive Streams) Flowable.generate Flux.generate Utilisation d’un générateur de valeurs

RxJava
 RxJava 2 Reactor Flowable.just Flux.just Émission d’une valeur existante Flowable.defer Flux.defer Émission tardive d’un Observable Flowable.fromCallable Mono.fromCallable Émission à partir d’un appel de fonction Flowable.create Flux.create Contrôle manuel de l’émission des événements Flowable.using Flux.using Gestion d’une ressource Flowable.fromPublisher Flux.from Utilisation d’un Publisher (Reactive Streams) Flowable.generate Flux.generate Utilisation d’un générateur de valeurs

exemple d’encapsulation

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private String redis() throws InterruptedException {
 CountDownLatch latch = new CountDownLatch(1);
 AtomicReference result = new AtomicReference<>();
 this.connection.subscribe((message, pattern) -> {
 result.set(message.toString());
 latch.countDown();
 }, TOPIC_NAME);
 latch.await();
 return result.get();
 } }

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private String redis() throws InterruptedException {
 CountDownLatch latch = new CountDownLatch(1);
 AtomicReference result = new AtomicReference<>();
 this.connection.subscribe((message, pattern) -> {
 result.set(message.toString());
 latch.countDown();
 }, TOPIC_NAME);
 latch.await();
 return result.get();
 } } Écoute de Redis

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private String redis() throws InterruptedException {
 CountDownLatch latch = new CountDownLatch(1);
 AtomicReference result = new AtomicReference<>();
 this.connection.subscribe((message, pattern) -> {
 result.set(message.toString());
 latch.countDown();
 }, TOPIC_NAME);
 latch.await();
 return result.get();
 } } Code de synchronisation Code de synchronisation

Étape 1 Encapsuler

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private String redis() throws InterruptedException {
 
 String result = Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> {
 sub.onNext(message.toString());
 sub.onComplete();
 }, TOPIC_NAME);
 }, BackpressureStrategy.BUFFER)
 .blockingFirst();
 return result;
 } }

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private String redis() throws InterruptedException {
 
 String result = Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> {
 sub.onNext(message.toString());
 sub.onComplete();
 }, TOPIC_NAME);
 }, BackpressureStrategy.BUFFER)
 .blockingFirst();
 return result;
 } } Encapsulation Synchronisation Contract Reactive

Étape 2 Asynchrone

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private DeferredResult redis() throws InterruptedException {
 
 DeferredResult result = new DeferredResult<>(10_000l);
 
 Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> {
 sub.onNext(message.toString());
 sub.onComplete();
 }, TOPIC_NAME);
 }, BackpressureStrategy.BUFFER)
 .subscribe(result::setResult);
 
 return result;
 }

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis")
 private DeferredResult redis() throws InterruptedException {
 
 DeferredResult result = new DeferredResult<>(10_000l);
 
 Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> {
 sub.onNext(message.toString());
 sub.onComplete();
 }, TOPIC_NAME);
 }, BackpressureStrategy.BUFFER)
 .subscribe(result::setResult);
 
 return result;
 } Résultat tardif Utilisation de DeferredResult

Étape 3 Reactive Streams

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
 private Flux redis() throws InterruptedException {
 
 Flowable rxjava = Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> sub.onNext(message.toString()), TOPIC_NAME);
 }, BackpressureStrategy.BUFFER);
 
 return Flux.defer(() -> rxjava);
 } }

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
 private Flux redis() throws InterruptedException {
 
 Flowable rxjava = Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> sub.onNext(message.toString()), TOPIC_NAME);
 }, BackpressureStrategy.BUFFER);
 
 return Flux.defer(() -> rxjava);
 } } RxJava 2 → Flux Flux → SSE

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
 private Publisher redis() throws InterruptedException {
 return Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> sub.onNext(message.toString()), TOPIC_NAME);
 }, BackpressureStrategy.BUFFER);
 
 }

@RestController
 public class HelloController {
 
 private static final byte[] TOPIC_NAME = "topic".getBytes();
 
 @RequestMapping(value = "/redis", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
 private Publisher redis() throws InterruptedException {
 return Flowable.create(sub -> {
 this.connection.subscribe((message, pattern) -> sub.onNext(message.toString()), TOPIC_NAME);
 }, BackpressureStrategy.BUFFER);
 
 } Publisher

No content

Migrer son application Reactive depuis RxJava

rx.Observable io.reactivex.Observable rx.Single rx.Completable io.reactivex.Completable io.reactivex.Single RxJava RxJava 2 package différent

RxJavaInterop.toV2Observable(rx.Observable.just(1, 2, 3))
 .firstElement()
 .subscribe();
 
 RxJavaInterop.toV1Observable(io.reactivex.Observable.just(1, 2, 3), 
 BackpressureStrategy.DROP)
 .first()
 .subscribe(); https://github.com/akarnokd/RxJava2Interop RxJava → RxJava 2 RxJava 2 → RxJava

Reactive Streams n’ acceptent pas de valeur null (sauf Maybe.fromCallable)

Observable.just(null);
 
 Single.just(null);
 
 Observable.fromCallable(() -> null)
 .subscribe(System.out::println, Throwable::printStackTrace);
 
 Observable.just(1).map(v -> null)
 .subscribe(System.out::println, Throwable::printStackTrace); NullPointerException NullPointerException NullPointerException N ullPointerException

Reactor se base sur Java 8 tandis que RxJava 2 utilise Java 6

Reactor se base sur Java 8 tandis que RxJava 2 utilise Java 6

Flux.interval(Duration.of(1, ChronoUnit.SECONDS))
 .map(i -> String.valueOf(i))
 .subscribe();

Flux.interval(Duration.of(1, ChronoUnit.SECONDS))
 .map(i -> String.valueOf(i))
 .subscribe(); java.time.Duration java.util.Function

Flux.fromStream(stream)
 .map(i -> "value -> " + String.valueOf(i))
 .toStream()
 .forEach(System.out::println);

Flux.fromStream(stream)
 .map(i -> "value -> " + String.valueOf(i))
 .toStream()
 .forEach(System.out::println); Conversion Stream Java 8 en Flux Conversion Flux en Stream Java 8

Reactor se base sur Java 8 tandis que RxJava 2 utilise Java 6

Reactor se base sur Java 8 tandis que RxJava 2 utilise Java 6

rx.Action1 rx.Func1 RxJava RxJava 2 rx.Func2 rx.Func3 rx.Action2 io.reactivex.functions.Consumer io.reactivex.functions.Function io.reactivex.functions.BiFunction io.reactivex.functions.Function3 io.reactivex.functions.BiConsumer Suivre les conventions Java 8

Observable.interval(1, TimeUnit.SECONDS)
 .map(i -> String.valueOf(i))
 .subscribe();

Observable.interval(1, TimeUnit.SECONDS)
 .map(i -> String.valueOf(i))
 .subscribe(); java.util.concurrent.TimeUnit

Observable.interval(1, TimeUnit.SECONDS)
 .map(i -> String.valueOf(i))
 .subscribe(); io.reactivex.functions.Function

Asynchrone gestion de contexte d’exécution

Schedulers Rethink cette partie là

Scheduler 
 (Pool de Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Tâche 1 Tâche 2 Tâche 3 Tâche 4 Scheduler 
 (Pool de Thread)

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

Scheduler 
 (Single Thread) Tâche 1 Tâche 2 Tâche 3 Tâche 4

java.lang.IllegalStateException: 
 Not on the main thread NetworkOnMainThreadException

JavaFx.fromClick(btn) 
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData())
 .observeOn(Schedulers.computation())
 .flatMap(data -> intensiveComputation(data))
 .observeOn(javaFx())
 .doOnNext(value -> btn.setText("Data: " + value)) 
 .subscribe();

JavaFx.fromClick(btn) 
 .observeOn(Schedulers.io())
 .switchMap(evt -> remoteApi.getData())
 .observeOn(Schedulers.computation())
 .flatMap(data -> intensiveComputation(data))
 .observeOn(javaFx())
 .doOnNext(value -> btn.setText("Data: " + value)) 
 .subscribe(); entrée/sortie calculs Thread graphique

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer Absent pour votre bien

RxJava RxJava 2 Reactor Description io() io() elastic() Pool de Thread qui grossit au besoin computation() computation() parallel() Pool fini de Thread single() single() single() Pool de 1 Thread immediate() immediate() Execute la tâche immédiatement trampoline() trampoline() Ajoute la tâche dans une file d’attente avant de l’executer

Reactor nommage technique

RxJava nommage fonctionnel

Backpressure

Consommateur ça déborde !

Flowable.range(1, 100000)
 .subscribe(new Subscriber() {
 
 private Subscription source;
 
 @Override
 public void onSubscribe(Subscription s) {
 this.source = s;
 this.source.request(1);
 }
 
 @Override
 public void onNext(Integer o) {
 longComputation(o);
 this.source.request(1);
 }
 });

Flowable.range(1, 100000)
 .subscribe(new Subscriber() {
 
 private Subscription source;
 
 @Override
 public void onSubscribe(Subscription s) {
 this.source = s;
 this.source.request(1);
 }
 
 @Override
 public void onNext(Integer o) {
 longComputation(o);
 this.source.request(1);
 }
 }); Garder la référence de l’abonnement Demander l’émission des éléments suivants Demande l’émission des premiers éléments

Flowable.range(1, 100000)
 .subscribe(new Subscriber() {
 
 private Subscription source;
 
 @Override
 public void onSubscribe(Subscription s) {
 this.source = s;
 this.source.request(1);
 }
 
 @Override
 public void onNext(Integer o) {
 longComputation(o);
 this.source.request(1);
 }
 }); Contrôle de l’émission

source()
 .subscribe(new Subscriber() {
 
 private Subscription source;
 
 @Override
 public void onSubscribe(Subscription s) {
 this.source = s;
 this.source.request(1);
 }
 
 @Override
 public void onNext(Integer o) {
 longComputation(o);
 this.source.request(1);
 }
 }); Kafka JDBC Fichier

La mécanique est identique avec Reactor (merci les Reactive Streams)

Debug Rajouter stack insignifiant

rx.exceptions.OnErrorNotImplementedException: Sequence contains too many elements at rx.internal.util.InternalObservableUtils$ErrorNotImplementedAction.call(InternalObservableUtils.java: 386) at rx.internal.util.InternalObservableUtils$ErrorNotImplementedAction.call(InternalObservableUtils.java: 383) at rx.internal.util.ActionSubscriber.onError(ActionSubscriber.java:44) at rx.observers.SafeSubscriber._onError(SafeSubscriber.java:153) at rx.observers.SafeSubscriber.onError(SafeSubscriber.java:115) at rx.internal.operators.OperatorSingle$ParentSubscriber.onNext(OperatorSingle.java:97) at rx.internal.operators.OnSubscribeRange$RangeProducer.slowPath(OnSubscribeRange.java:90) at rx.internal.operators.OnSubscribeRange$RangeProducer.request(OnSubscribeRange.java:68) at rx.Subscriber.setProducer(Subscriber.java:211) at rx.internal.operators.OnSubscribeRange.call(OnSubscribeRange.java:38) Stacktrace inexploitable

RxJavaHooks.enableAssemblyTracking(); […] Caused by: rx.exceptions.AssemblyStackTraceException: Assembly trace: at rx.Observable.create(Observable.java:100) at rx.Observable.lift(Observable.java:237) at rx.Observable.single(Observable.java:9330) at rxjava2.BasicExampleTest.stacktrace(BasicExampleTest.java:225) rx.Observable.range(1, 100)
 .single()
 .subscribe();

enableTracing n’est pas encore supporté par RxJava 2 (pour cela, il faut utiliser RxJava2Extensions) https://github.com/akarnokd/RxJava2Extensions

Hooks.onOperator(providedHook -> providedHook.operatorStacktrace());
 
 Flux.just(1, 2, 3)
 .single()
 .subscribe(System.out::println, Throwable::printStackTrace);

Hooks.onOperator(providedHook -> providedHook.operatorStacktrace());
 
 Flux.just(1, 2, 3)
 .single()
 .subscribe(System.out::println, Throwable::printStackTrace);

Flux.range(1, 100)
 .single()
 .checkpoint("label")
 .subscribe(…);

Flux.range(1, 100)
 .single()
 .checkpoint("label")
 .subscribe(…);

Flux.range(1, 100)
 .single()
 .checkpoint("label")
 .subscribe(…); Assembly trace from producer [reactor.core.publisher.MonoSingle], described as [label] : reactor.core.publisher.Mono.checkpoint(Mono.java:1304) rxjava2.BasicExampleTest.stacktraceReactor(BasicExampleTest.java:236) Error has been observed by the following operator(s): |_ Mono.checkpoint(BasicExampleTest.java:236)

Attention dégradation des performances !

Performance

Génération 3 4 5+ Fusion RxJava 2 Reactor

Attention Slides conceptuels

Sans fusion

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Avec fusion

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

Transform ( ⬛ → ⃝ ) Transform ( ⃝ → ⬛)

La fusion diminue la consommation mémoire et augmente les performances

for (int x = 0; x < 10_000; x++) {
 
 Observable.interval(10, TimeUnit.MILLISECONDS)
 .takeWhile(i -> take.get())
 .flatMap(i -> Observable.range(1, 100))
 .subscribe();
 }

RxJava 2 RxJava Reactor

http://akarnokd.blogspot.fr/2016/12/the-reactive-scrabble-benchmarks.html Stream Java 8 RxJava 2 / Reactor RxJava Décembre 2016

Stream Java 8 RxJava 2 / Reactor RxJava https://twitter.com/akarnokd/status/844555409012740096 Mars 2017

RxJava 2 : 
 streams performants pour Android

Reactor & RxJava 2 : 
 streams performants pour serveur web

Ecosystème

Communauté Sponsor RxJava Grande Anciennement Netflix
 Maintenant Community RxJava 2 Grande Community Reactor En croissance Pivotal

RxJava RxJava 2 Reactor Retrofit Oui Oui Non RxAndroid Oui Oui Non Realm Oui Non Non Hystrix Oui Non Non Couchbase Oui Non Non MongoDB Oui Non Non Spring Data 2.0 Oui Non Oui Reactor IPC Non Non Oui WebFlux Non Oui Oui Android Spring

RxJava RxJava 2 Reactor Retrofit Oui Oui Non RxAndroid Oui Oui Non Realm Oui Non Non Hystrix Oui Non Non Couchbase Oui Non Non MongoDB Oui Non Non Spring Data 2.0 Oui Non Oui Reactor IPC Non Non Oui WebFlux Non Oui Oui Android Spring

We are aggressively 
 migrating our internal code to RxJava 2 https://github.com/uber/AutoDispose

RxJava RxJava 2 Reactor Retrofit Oui Oui Non RxAndroid Oui Oui Non Realm Oui Non Non Hystrix Oui Non Non Couchbase Oui Non Non MongoDB Oui Non Non Spring Data 2.0 Oui Non Oui Reactor IPC Non Non Oui WebFlux Non Oui Oui Android Spring genre pourquoi si tu as une stack spring se forcer a faire du rx2 si le support reactor est meilleur)?

Forte inertie pour migrer

Migrer vers RxJava 2 ou Reactor

RxJava RxJava 2 Reactor Retrofit Oui Oui Non RxAndroid Oui Oui Non Realm Oui Non Non Hystrix Oui Non Non Couchbase Oui Non Non MongoDB Oui Non Non Spring Data 2.0 Oui Non Oui Reactor IPC Non Non Oui WebFlux Non Oui Oui Android Spring

Spring 5 va accelerer l’adoption de Reactor

No content

No content

Android RxJava 2

Backend RxJava 2

Spring Reactor

2017 l’année des Reactive Streams

Merci pour votre attention On reste en contact ? [email protected] @dwursteisen http://blog.soat.fr RefCard RxJava 2 bientôt disponible !