如何组合 Observables 以避免给定的嵌套和依赖回调？

Question

在this blog, he gives this（copy/pasted下面的代码）例子中为回调地狱。但是，没有提到如何使用 Reactive Extensions 来消除这个问题。

所以这里的F3依赖于F1的完成，F4和F5依赖于F2的完成。

想知道 Rx 中的功能等价物是什么。
如何在Rx中表示F1,F2,F3,F4,F5都应该异步拉取？

注意：我目前正在尝试围绕 Rx 思考，所以在问这个问题之前我没有尝试解决这个例子。

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicReference;

public class CallbackB {

    /**
     * Demonstration of nested callbacks which then need to composes their responses together.
     * <p>
     * Various different approaches for composition can be done but eventually they end up relying upon
     * synchronization techniques such as the CountDownLatch used here or converge on callback design
     * changes similar to <a href="https://github.com/Netflix/RxJava">Rx</a>.
     */
    public static void run() throws Exception {
        final ExecutorService executor = new ThreadPoolExecutor(4, 4, 1, TimeUnit.MINUTES, new LinkedBlockingQueue<Runnable>());
        /* the following are used to synchronize and compose the asynchronous callbacks */
        final CountDownLatch latch = new CountDownLatch(3);
        final AtomicReference<String> f3Value = new AtomicReference<String>();
        final AtomicReference<Integer> f4Value = new AtomicReference<Integer>();
        final AtomicReference<Integer> f5Value = new AtomicReference<Integer>();

        try {
            // get f3 with dependent result from f1
            executor.execute(new CallToRemoteServiceA(new Callback<String>() {

                @Override
                public void call(String f1) {
                    executor.execute(new CallToRemoteServiceC(new Callback<String>() {

                        @Override
                        public void call(String f3) {
                            // we have f1 and f3 now need to compose with others
                            System.out.println("intermediate callback: " + f3 + " => " + ("f4 * f5"));
                            // set to thread-safe variable accessible by external scope 
                            f3Value.set(f3);
                            latch.countDown();
                        }

                    }, f1));
                }

            }));

            // get f4/f5 after dependency f2 completes 
            executor.execute(new CallToRemoteServiceB(new Callback<Integer>() {

                @Override
                public void call(Integer f2) {
                    executor.execute(new CallToRemoteServiceD(new Callback<Integer>() {

                        @Override
                        public void call(Integer f4) {
                            // we have f2 and f4 now need to compose with others
                            System.out.println("intermediate callback: f3" + " => " + (f4 + " * f5"));
                            // set to thread-safe variable accessible by external scope 
                            f4Value.set(f4);
                            latch.countDown();
                        }

                    }, f2));
                    executor.execute(new CallToRemoteServiceE(new Callback<Integer>() {

                        @Override
                        public void call(Integer f5) {
                            // we have f2 and f5 now need to compose with others
                            System.out.println("intermediate callback: f3" + " => " + ("f4 * " + f5));
                            // set to thread-safe variable accessible by external scope 
                            f5Value.set(f5);
                            latch.countDown();
                        }

                    }, f2));
                }

            }));

            /* we must wait for all callbacks to complete */
            latch.await();
            System.out.println(f3Value.get() + " => " + (f4Value.get() * f5Value.get()));
        } finally {
            executor.shutdownNow();
        }
    }

    public static void main(String[] args) {
        try {
            run();
        } catch (Exception e) {
            e.printStackTrace();
        }
    }

    private static final class CallToRemoteServiceA implements Runnable {

        private final Callback<String> callback;

        private CallToRemoteServiceA(Callback<String> callback) {
            this.callback = callback;
        }

        @Override
        public void run() {
            // simulate fetching data from remote service
            try {
                Thread.sleep(100);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            callback.call("responseA");
        }
    }

    private static final class CallToRemoteServiceB implements Runnable {

        private final Callback<Integer> callback;

        private CallToRemoteServiceB(Callback<Integer> callback) {
            this.callback = callback;
        }

        @Override
        public void run() {
            // simulate fetching data from remote service
            try {
                Thread.sleep(40);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            callback.call(100);
        }
    }

    private static final class CallToRemoteServiceC implements Runnable {

        private final Callback<String> callback;
        private final String dependencyFromA;

        private CallToRemoteServiceC(Callback<String> callback, String dependencyFromA) {
            this.callback = callback;
            this.dependencyFromA = dependencyFromA;
        }

        @Override
        public void run() {
            // simulate fetching data from remote service
            try {
                Thread.sleep(60);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            callback.call("responseB_" + dependencyFromA);
        }
    }

    private static final class CallToRemoteServiceD implements Runnable {

        private final Callback<Integer> callback;
        private final Integer dependencyFromB;

        private CallToRemoteServiceD(Callback<Integer> callback, Integer dependencyFromB) {
            this.callback = callback;
            this.dependencyFromB = dependencyFromB;
        }

        @Override
        public void run() {
            // simulate fetching data from remote service
            try {
                Thread.sleep(140);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            callback.call(40 + dependencyFromB);
        }
    }

    private static final class CallToRemoteServiceE implements Runnable {

        private final Callback<Integer> callback;
        private final Integer dependencyFromB;

        private CallToRemoteServiceE(Callback<Integer> callback, Integer dependencyFromB) {
            this.callback = callback;
            this.dependencyFromB = dependencyFromB;
        }

        @Override
        public void run() {
            // simulate fetching data from remote service
            try {
                Thread.sleep(55);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            callback.call(5000 + dependencyFromB);
        }
    }

    private static interface Callback<T> {
        public void call(T value);
    }
}

Answer 1

根据你的代码。假设使用 Observable.

完成远程调用

 Observable<Integer>  callRemoveServiceA()  { /* async call */  }

/* .... */

Observable<Integer>  callRemoveServiceE(Integer f2) { /* async call */  }

你想要什么：

调用 serviceA 然后调用 serviceB 得到 serviceA
调用 serviceC 然后调用 serviceD 和 serviceE 得到 serviceC
利用serviceE和serviceD的结果，构建一个新值
用serviceB

使用 RxJava，您将使用以下代码实现此目的：

Observable<Integer> f3 = callRemoveServiceA() // call serviceA
            // call serviceB with the result of serviceA
            .flatMap((f1) -> callRemoveServiceB(f1)); 


Observable<Integer> f4Andf5 = callRemoveServiceC() // call serviceC
                    // call serviceD and serviceE then build a new value
                    .flatMap((f2) -> callRemoveServiceD(f2).zipWith(callRemoveServiceE(f2), (f4, f5) -> f4 * f5));

// compute the string to display from f3, and the f4, f5 pair
f3.zipWith(f4Andf5, (childF3, childF4Andf5) -> childF3 + " => " + childF4Andf5)
            // display the value
            .subscribe(System.out::println);

这里重要的部分是使用 flapMap 和 zip（或 zipWith）

flapMap 会将一个值转换为另一个 Observable。这里的 Observable 将是您的新异步调用。 ( http://reactivex.io/documentation/operators/flatmap.html )
zip 将从两个不同的 Observable 组成一个新值。因此，您可以使用两个（或更多）Obsevable ( http://reactivex.io/documentation/operators/zip.html )

您可以在此处获取有关 flapMap 的更多信息：When do you use map vs flatMap in RxJava?

Answer 2

我是有关回调和 Java 期货的参考博客 post 的原作者。下面是一个使用 flatMap、zip 和 merge 异步进行服务组合的例子。

它获取一个 User 对象，然后同时获取 Social 和 PersonalizedCatalog 数据，然后从 PersonalizedCatalog 中为每个视频同时获取一个 Bookmark、Rating 和 Metadata，将它们压缩在一起，并将所有响应合并到一个渐进式流输出中作为服务器发送的事件。

return getUser(userId).flatMap(user -> {
    Observable<Map<String, Object>> catalog = getPersonalizedCatalog(user)
            .flatMap(catalogList -> catalogList.videos().<Map<String, Object>> flatMap(
                    video -> {
                        Observable<Bookmark> bookmark = getBookmark(video);
                        Observable<Rating> rating = getRatings(video);
                        Observable<VideoMetadata> metadata = getVideoMetadata(video);
                        return Observable.zip(bookmark, rating, metadata, (b, r, m) -> combineVideoData(video, b, r, m));
                    }));

    Observable<Map<String, Object>> social = getSocial(user).map(s -> {
        return s.getDataAsMap();
    });

    return Observable.merge(catalog, social);
}).flatMap(data -> {
    String json = SimpleJson.mapToJson(data);
    return response.writeStringAndFlush("data: " + json + "\n");
});

可以在 https://github.com/Netflix/ReactiveLab/blob/952362b89a4d4115ae0eecf0e73f273ecb27ba98/reactive-lab-gateway/src/main/java/io/reactivex/lab/gateway/routes/RouteForDeviceHome.java#L33

上运行的应用程序上下文中查看此示例

由于我不可能在此处提供所有信息，您还可以在 https://speakerdeck.com/benjchristensen/reactive-streams-with-rx-at-javaone-2014?slide=32.

找到演示文稿形式的解释（带有 link 视频）

如何组合 Observables 以避免给定的嵌套和依赖回调？

How to compose Observables to avoid the given nested and dependent callbacks?

java

future

rx-java