You can not achieve parallel via combineLast
or zip
, rxjava
will execute and emit your items in sequence in my testing.
If one of your task fail, your Func2#call
will not get called and onError
will submitted instead. You even can not get the results of other successful tasks in this way.
The solution is flatMap
, it's the traditional way to achieve concurrent in rxjava
. It also meet your other requirements.
Here is a small but completed example.
I use a simple website service to test.
I use a Semaphore
to wait for all task done, you can completely ignore it. And I add logging to the http request for better understanding, you can complete ignore it also.
public interface WebsiteService {
@GET
Observable<ResponseBody> website(@Url String url);
}
Then I use the following to test the result with rxjava
.
HttpLoggingInterceptor loggingInterceptor = new HttpLoggingInterceptor();
loggingInterceptor.setLevel(HttpLoggingInterceptor.Level.BASIC);
Retrofit retrofit = new Retrofit.Builder().baseUrl("https://www.google.com")
.addCallAdapterFactory(RxJava2CallAdapterFactory.create())
.client(new OkHttpClient.Builder().addInterceptor(loggingInterceptor).build())
.build();
WebsiteService websiteService = retrofit.create(WebsiteService.class);
final Semaphore s = new Semaphore(1);
try {
s.acquire();
} catch (InterruptedException e) {
e.printStackTrace();
}
Observable<ResponseBody> first = websiteService.website("http://github.com");
Observable<ResponseBody> second = websiteService.website("http://stackoverflow.com");
Observable<ResponseBody> third = websiteService.website("http://notexisting.com");
final int numberOfCalls = 3; // testing for three calls
Observable.just(first, second, third)
.flatMap(new Function<Observable<ResponseBody>, ObservableSource<ResponseBody>>() {
@Override
public ObservableSource<ResponseBody> apply(@NonNull Observable<ResponseBody> responseBodyObservable) throws Exception {
return responseBodyObservable.subscribeOn(Schedulers.computation());
}
})
.subscribeOn(Schedulers.computation())
.subscribe(new Observer<ResponseBody>() {
private int currentDoneCalls = 0;
private void checkShouldReleaseSemaphore() {
if (currentDoneCalls >= numberOfCalls) {
s.release();
}
}
@Override
public void onSubscribe(@NonNull Disposable d) {
}
@Override
public void onNext(@NonNull ResponseBody responseBody) {
System.out.println("Retrofit call success " + responseBody.contentType());
synchronized (this) {
currentDoneCalls++;
}
checkShouldReleaseSemaphore();
}
@Override
public void onError(@NonNull Throwable e) {
System.out.println("Retrofit call failed " + e.getMessage());
synchronized (this) {
currentDoneCalls++;
}
checkShouldReleaseSemaphore();
}
@Override
public void onComplete() {
System.out.println("onComplete, All request success");
checkShouldReleaseSemaphore();
}
});
try {
s.acquire();
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
System.out.println("All request done");
s.release();
}
I use rxjava2
and retrofit adapter-rxjava2
for testing.
compile 'io.reactivex.rxjava2:rxandroid:2.0.1'
compile 'io.reactivex.rxjava2:rxjava:2.1.0'
compile 'com.squareup.retrofit2:retrofit:2.3.0'
compile 'com.squareup.retrofit2:adapter-rxjava2:2.3.0'
compile 'com.squareup.okhttp3:logging-interceptor:3.8.1'
Updated
The introduction page of RxJava2 from github has pointed out the practical way to implement paralellism.
Practically, paralellism in RxJava means running independent flows and merging their results back into a single flow. The operator flatMap
does this...
Although this example is based on RxJava2
, the operation flatMap
is
already existing in RxJava
.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…