如何在 Reactor 中将 Context 与 flatMap() 一起使用？

Question

I have a problem understanding Context . So documentations says that Context is:

A key/value store that is propagated between components such as operators via the context protocol. Contexts are ideal to transport orthogonal information such as tracing or security tokens.

Great.

Now let's suppose we want to propagate something using Context to have it everywhere. To call another async code we simply use the flatMap() method.

Problem : how to access Context inside called method?

Sample (simple) code:

public class TestFlatMap {
    public static void main(final String ...args) {
        final Flux<String> greetings = Flux.just("Hubert", "Sharon")
            .flatMap(TestFlatMap::nameToGreeting)
            .subscriberContext(context ->
                Context.of("greetingWord", "Hello")  // context initialized
            );
        greetings.subscribe(System.out::println);
    }

    private static Mono<String> nameToGreeting(final String name) {
        return Mono.just("Hello " + name + " !!!");  // ALERT: we don't have Context here
    }
}

The called method can be (and most likely will be) in another class.

Thanks for help in advance !

Edit : removed some code to make the question more concise and straight to the point.

Answer 1

Chain your Publisher s and may the Context be with you

In the case, you connected all your Publisher s (and this includes connections within the flatMap / concatMap and similar operators) you will have Context correctly propagated among the whole stream runtime.

To access Context in the nameToGreeting method, you may call Mono.subscribeContext and retrieve stored info event if it seems that methods are not related. The following shows the mentioned concept:

public class TestFlatMap {
    public static void main(final String ...args) {
        final Flux<String> greetings = Flux.just("Hubert", "Sharon")
                                           .flatMap(TestFlatMap::nameToGreeting)
                                           .subscriberContext(context ->
                                                   Context.of("greetingWord", "Hello")  // context initialized
                                           );
        greetings.subscribe(System.out::println);
    }

    private static Mono<String> nameToGreeting(final String name) {
        return Mono.subscriberContext()
                   .filter(c -> c.hasKey("greetingWord"))
                   .map(c -> c.get("greetingWord"))
                   .flatMap(greetingWord -> Mono.just(greetingWord + " " + name + " " + "!!!"));// ALERT: we have Context here !!!
    }
}

Also, you can do the same in the following way, using zip operator, to combine results later:

public class TestFlatMap {
    public static void main(final String ...args) {
        final Flux<String> greetings = Flux.just("Hubert", "Sharon")
                                           .flatMap(TestFlatMap::nameToGreeting)
                                           .subscriberContext(context ->
                                                   Context.of("greetingWord", "Hello")  // context initialized
                                           );
        greetings.subscribe(System.out::println);
    }

    private static Mono<String> nameToGreeting(final String name) {
        return Mono.zip(
            Mono.subscriberContext()
                .filter(c -> c.hasKey("greetingWord"))
                .map(c -> c.get("greetingWord")), // ALERT: we have Context here !!!
            Mono.just(name),
            (greetingWord, receivedName) -> greetingWord + " " + receivedName + " " + "!!!"
        );
    }
}

So, why it works?

As we can see from the sample above, the nameToGreeting is called within the context of the main Flux . Under the hood -> (Here Some FluxFlatMap internals) , each mapped Publisher is subscribed by FlatMapInner . If we look at the FlatMapInner and look for the currentContext override we will see, that FlatMapInner uses parent Context , which means if the parent has a Reactor Context - then this context will be propagated to each inner Publisher .

Therefore, returned by the nameToGreeting method Mono will have the same Context as its parent

Answer 2

Reactor-Core v3.4 introduced Mono.deferContextual and Flux.deferContextual , which supersede Mono.deferWithContext and Flux.deferWithContext introduced in v3.3.

Using these methods, Oleh Dokukas zip example can be simplified to

public class TestFlatMap {
    public static void main(final String ...args) {
        final Flux<String> greetings = Flux.just("Hubert", "Sharon")
                .flatMap(TestFlatMap::nameToGreeting)
                .subscriberContext(context ->
                        Context.of("greetingWord", "Hello"));  // context initialized
        greetings.subscribe(System.out::println);
    }

    private static Mono<String> nameToGreeting(final String name) {
        return Mono.deferContextual(c -> Mono.just(name)
                .filter(x -> c.hasKey("greetingWord"))
                .map(n -> c.get("greetingWord") + " " + n + " " + "!!!"));
    }
}