Overview

Let’s say we develop a service that has to interact with other components. Unfortunately, those components are slow and blocking.

It may be a legacy service that is very slow or some blocking API that we must use regardless we have no control over it. In this post, we will call 2 APIs. One of them will block for 2 seconds and another for 5 seconds.

We also need to print response status codes once both responses are available. If we do it in the old fashion, non-reactive way we would block a calling thread for five seconds. Holding thread for five seconds is not efficient, is it?

event

Services

I used httpstat.us as a web service. This is a simple service for generating different HTTP codes to test web clients. It is possible to provide extra parameters, in this case, sleep that blocks HTTP requests for a provided amount of time.

Let’s use httpie to test both services

Service 1 will block for 5 seconds and return a response with status code 200


http://httpstat.us/200?sleep=5000
_____________________________________________

HTTP/1.1 200 OK
Content-Length: 6
Content-Type: text/plain
Date: Tue, 08 Mar 2022 17:05:08 GMT
Request-Context: appId=cid-v1:1e93d241-20e4-4513-bbd7-f452a16a5d69
Server: Kestrel
Set-Cookie: ARRAffinity=e2c17206c539113795daf64bd958d003f2b29b9f62da53617beea05468875ba5;Path=/;HttpOnly;Domain=httpstat.us

200 OK

Service 2 is identical to the previous one except that it blocks for 2 seconds instead of 5.


http://httpstat.us/200?sleep=2000
_____________________________________________

HTTP/1.1 200 OK
Content-Length: 6
Content-Type: text/plain
Date: Tue, 08 Mar 2022 17:11:53 GMT
Request-Context: appId=cid-v1:1e93d241-20e4-4513-bbd7-f452a16a5d69
Server: Kestrel
Set-Cookie: ARRAffinity=e2c17206c539113795daf64bd958d003f2b29b9f62da53617beea05468875ba5;Path=/;HttpOnly;Domain=httpstat.us

200 OK

Web Client

We have learned about services let’s discuss web client. In this post, I used Vert.x Web Client. It is asynchronous, easy to use HTTP and HTTP/2 client that supports RxJava too.


private static Single<Integer> service1(WebClient webClient) {
        return webClient.getAbs("http://httpstat.us/200?sleep=5000")
                .rxSend()
                .doOnSuccess(response -> out.println("[" + Thread.currentThread().getName() + "] service 1: response received"))
                .map(HttpResponse::statusCode);
    }
}

private static Single<Integer> service2(WebClient webClient) {
        return webClient.getAbs("http://httpstat.us/200?sleep=2000")
                .rxSend()
                .doOnSuccess(response -> out.println("[" + Thread.currentThread().getName() + "] service 2 response received"))
                .map(HttpResponse::statusCode);
    }

Both methods are very similar. They take WebClient as parameter and send HTTP request returning Single<Integer>. Where integer is an HTTP response code. Returning RxJava Single assures us that result is asynchronous. Status code will be accessible later when it is available. This also gives us lazy evaluation, services will get invoked only if an active subscription is present.

Consuming Single sources

There are two sources that we will need to subscribe to. RxJava has a convenient method to combine Single sources together. We can invoke method .zipWith on the first source and supply two parameters. The first is the source to zip with and the second one is a function to consume both results, process them and return something else.

In this case, the return type is AbstractMap.SimpleEntry<Integer, Integer> that is a simple tuple of two integers. Looks verbose, doesn’t it? Unfortunately, there are no better tuple or pair implementations in core Java libraries.

Thanks to Java lambdas we can pass behaviour as a parameter.


Single<Integer> service1Code = service1(webClient);
Single<Integer> service2Code = service2(webClient);

Single<AbstractMap.SimpleEntry<Integer, Integer>> tupleSource = 
            service1Code.zipWith(service2Code, (s1, s2) -> new AbstractMap.SimpleEntry<>(s1, s2));

You may implement your own tuple or pair if AbstractMap.SimpleEntry feels too verbose

All together

Finally, we can put all bits and peaces together

// Vertx instance and web client
Vertx vertx = Vertx.vertx();
WebClient webClient = WebClient.create(vertx);

// single sources. Lazy evaluation, no invocation at this point
Single<Integer> service1Code = service1(webClient);
Single<Integer> service2Code = service2(webClient);

// combine results together and create tuple
Single<AbstractMap.SimpleEntry<Integer, Integer>> tupleSource =
            service1Code.zipWith(service2Code, (s1, s2) -> new AbstractMap.SimpleEntry<>(s1, s2));

// subscribe and invoke services
tupleSource
    .doFinally(countDownLatch::countDown)
    .subscribe(Services::printResult);

Here are the results printed on the console after running the code. Both requests were dispatched from the same vertx event loop thread. The program also prints messages that the thread is not blocked every second. Finally, it prints both status codes as a final result. As you can see everything happened on the same thread.

[vert.x-eventloop-thread-1] is released
[vert.x-eventloop-thread-1] is released
[vert.x-eventloop-thread-1] service 2 response received
[vert.x-eventloop-thread-1] is released
[vert.x-eventloop-thread-1] is released
[vert.x-eventloop-thread-1] is released
[vert.x-eventloop-thread-1] service 1: response received
[vert.x-eventloop-thread-1] Result: service1:200 service2:200

That was all for this time. The complete code is on gist

You can also run the code directly with jbang

jbang https://gist.github.com/1da6638f5df21fe83c13d055dd9ee679

Happy reactive coding!