Build the apps that make the world run. Join us at SpringOne in San Francisco, Dec 6–8. >
close

Spring Blog

Reactor Core 3.0 becomes a unified Reactive Foundation on Java 8

Releases
Stephane Maldini
0 Comment

[Update]
Following our Reactor 2.5 introduction, we shifted the version to 3.0 to reflect better the major redesign effort overtaken.

A meaningful composition API

The former reactor-stream module has been merged into Reactor Core 3.0. Flux and Mono respectively representing 0..N and 0..1 sequence types now cover a solid range of operations for the following categories:

  • Cold-to-Hot or Multicasting : publish, publishNext, cache, multicast…
  • Aggregating/Reducing (Transforming) : buffer, reduce, scan, window, sample…
  • Filtering : filter, exists, single…
  • Conditioning : timeout, take, takeUntil, skip, skipUntil…
  • Combining : withLatestFrom, combineLatest
  • Backpressuring : onBackpressureDrop, onBackpressureLatest…

While the capabilities existed, we are now using the hyper efficient Reactive Streams Commons operator implementations.

Designed for Java 8 and beyond

All Reactor 3.0 projects have been upgraded to Java 8. While the JVM market is getting standardized on Java 8, we see no reason to delay further our transition. Let’s enumerate the benefits :

  • All “backported” functional callbacks in reactor.fn have been replaced by java.util.function. As a result Reactor offers a standard interactive contract and a competitive surface API.
  • Convert Flux to/from java.util.stream.Stream
  • Convert Mono to/from CompletableFuture
  • Safely create Mono from Optional
  • Sane time period support with Duration
  • Dropped shadowed JSR 166 backport for ConcurrentHashMap use specially affecting Reactor Addons
  • A massive step closer to the programming experience delivered by Spring Framework 5, itself building on Java 8.

Some Java 8 extractors and generators in action:

java.util.stream.Stream<String> stream = 
    Mono.fromFuture(someCompletableFuture)
        .timeout(Duration.ofSeconds(30))
        .log("hello")
        .flatMap( pojo -> Flux.just(pojo.getId(), pojo.getName())
        .toStream();

Flux.fromStream(stream)
    .delayMillis(1000L)
    .subscribe(System.out::println);

We could envision a backport for Android compatibility later, still our focus for now is on Java 8 applications and onward.

Highlights

  • Updated Javadoc with more Marble Diagram illustrations.
  • 100% Flux and Mono operational design based on the collaboration effort Reactive Streams Commons
    – Build on years of experience from Reactive4Java, RxJava and Reactor
    – Efficient, really efficient, supports even further efficiency with operation fusion
    – More Performance benchmarks from Reactive Streams Commons
  • Available concurrency and prefetch arguments for operators with queue
  • Mono publish-subscribe : MonoProcessor<T>
  • More tests : combining current Reactive Streams Commons (879) and Core (1153) plus our internal early adopters such as Spring Framework 5 and Cloud Foundry Java Client.

Refer to the issues tracker for more details.

What’s (on)Next ?

With this release we have a foundation we think will be useful for library and application developers alike. We are ready to collect more feedbacks while our next stop will be on Reactor IPC : Reactor Aeron and Reactor Netty. There are great plans stirring for months now for these two Reactor IPC modules to help you cross network boundaries with backpressure factored-in ! The two inspiring traits are the same as in Reactor Core : Focus on API and Efficiency, no extra ceremony or academic background required.

In parallel we are doubling our effort on creating various starting experiences, expect some blogging, guides and close collaboration with the Spring Boot team to get this awesomeness right at your fingertip with little fuss.

comments powered by Disqus