如何与Project Reactor并行调用两个或多个Web服务或REST并加入答案

时间:2018-02-13 08:46:33

标签: java spring spring-boot project-reactor

您好我想知道如何在pararelo中调用两个或更多Web服务或Rest服务并撰写调用的响应。

我在网上发现了一些使用其他技术的例子,但我不能让它与反应堆一起工作

// start task A asynchronously
CompletableFuture<ResponseA> futureA = asyncServiceA.someMethod(someParam);
// start task B asynchronously
CompletableFuture<ResponseB> futureB = asyncServiceB.someMethod(someParam);

CompletableFuture<String> combinedFuture = futureA
        .thenCombine(futureB, (a, b) -> a.toString() + b.toString());

// wait till both A and B complete
String finalValue = combinedFuture.join();

/////////////////////////////////////////////// /////////////////////////////////

static void Run()
{
    //Follow steps at this link for addding a reference to the necessary .NET library:
    //http://stackoverflow.com/questions/9611316/system-net-http-missing-from-    
    //namespace-using-net-4-5

    //Create an HTTP Client
    var client = new HttpClient();

    //Call first service
    var task1 = client.GetAsync("http://www.cnn.com");

    //Call second service
    var task2 = client.GetAsync("http://www.google.com");

    //Create list of all returned async tasks
    var allTasks = new List<Task<HttpResponseMessage>> { task1, task2 };

    //Wait for all calls to return before proceeding
    Task.WaitAll(allTasks.ToArray());

}

3 个答案:

答案 0 :(得分:2)

让我们假设你需要点击2个服务,所以你需要2个基础 <my-typeahead-control formControlName="item" [items]="allItemsArray" itemLabelKey="name"></my-typeahead-control> (每个都配置了正确的基本URL,例如一个身份验证方案):

WebClient

从那里开始,假设您在控制器中注入了这两个Web客户端(作为限定的 bean)。这是使用Reactor联合调用两个代码的代码:

@Bean
public WebClient serviceAClient(String authToken) {
    return WebClient.builder()
            .baseUrl("http://serviceA.com/api/v2/")
            .defaultHeader(HttpHeaders.AUTHORIZATION, "Basic " + authToken)
            .defaultHeader(HttpHeaders.CONTENT_TYPE, MediaType.APPLICATION_JSON_VALUE)
            .build();
}

@Bean
public WebClient serviceBClient(String authToken): WebClient {
    return WebClient.builder()
            .baseUrl("https://api.serviceB.com/")
            .defaultHeader(HttpHeaders.AUTHORIZATION, "token " + authToken)
            .defaultHeader(HttpHeaders.CONTENT_TYPE, MediaType.APPLICATION_JSON_VALUE)
            .build();
}

请注意,zip函数可以生成更有意义的内容,就像2个响应中的业务对象一样。结果Mono<ResponseA> respA = webclientA.get() .uri("/sub/path/" + foo) .retrieve() .bodyToMono(ResponseA.class); Mono<ResponseB> respB = webclientB.get() .uri("/path/for/b") .retrieve() .bodyToMono(ResponseB.class); Mono<String> join = respA.zipWith(respB, (a, b) -> a.toString + b.toString); return join; 只会触发2个请求,如果有东西订阅它(在Spring WebFlux的情况下,如果从控制器方法返回它,框架会这样做)。

答案 1 :(得分:1)

如果您正在使用Spring reactor,那么您需要的是运营商Zip,以便运行您的流程并将其压缩一次。

/**
 * Zip operator execute the N number of Flux independently, and once all them are finished, results
 * are combined in TupleN object.
 */
@Test
public void zip() {
    Flux<String> flux1 = Flux.just("hello ");
    Flux<String> flux2 = Flux.just("reactive");
    Flux<String> flux3 = Flux.just(" world");
    Flux.zip(flux1, flux2, flux3)
            .map(tuple3 -> tuple3.getT1().concat(tuple3.getT2()).concat(tuple3.getT3()))
            .map(String::toUpperCase)
            .subscribe(value -> System.out.println("zip result:" + value));
}

您可以在此处查看有关反应技术的更多信息https://github.com/politrons/reactive

答案 2 :(得分:0)

如果您已经有一个同步实现,则可以借助Mono.fromCallable()方法轻松添加一些反应堆功能使其并行运行。

Mono<ResponseA> responseA = Mono
.fromCallable(() -> blockingserviceA.getResponseA())
.subscribeOn(Schedulers.elastic()); // will execute on a separate thread when called
Mono<ResponseB> responseB = Mono
.fromCallable(() -> blockingserviceB.getResponseB())
.subscribeOn(Schedulers.elastic());
// At that point nothing has been called yet, responseA and responseB are empty Mono

AggregatedStuff aggregatedStuff = Mono.zip(responseA, responseB) // zip as many Mono as you want
.flatMap(r -> doStuff(r.getT1(), r.getT2())) // do whatever needed with the results
.block(); // execute the async calls, and then execute flatMap transformation

fromCallable()just()之间需要注意的重要一点是,just()将直接在主线程中执行,但是fromCallable()懒惰的表示仅在需要时才执行,例如:当您调用block(),collect()(对于Flux),...等...

Mono<ResponseA> responseA = Mono
.just(blockingserviceA.getResponseA()) // execute directly
Mono<ResponseB> responseB = Mono
.just(blockingserviceB.getResponseB()) // execute directly
// Above code have been executed sequentially, you can access response result 
// => yes it is kind of useless, and yes it is exactly how I did it the first time!

因此,避免将just()用于要并行运行的繁重任务。使用just()进行实例化是完全正确的,因为您不想在每次实例化String或任何其他对象时都创建新线程并获得随之而来的开销。

PS:正如SimonBaslé所指出的,您可以使用WebClient直接返回Mono和Flux并进行异步调用,但是如果您已经实现了api客户端,并且无法选择重构整个应用程序,{{1 }}是设置异步过程而无需重构大量代码的简单方法。