在Android单元测试中模拟改造Observable <T>响应

Question

我有一个API接口,我正在测试View涉及网络调用.

@Config(emulateSdk = 18)
public class SampleViewTest extends RobolectricTestBase {

    ServiceApi apiMock;

    @Inject
    SampleView fixture;

    @Override
    public void setUp() {
        super.setUp(); //injection is performed in super
        apiMock = mock(ServiceApi.class);
        fixture = new SampleView(activity);
        fixture.setApi(apiMock);
    }

    @Test
    public void testSampleViewCallback() {
        when(apiMock.requestA()).thenReturn(Observable.from(new ResponseA());
        when(apiMock.requestB()).thenReturn(Observable.from(new ResponseB());

        AtomicReference<Object> testResult = new AtomicReference<>();
        fixture.updateView(new Callback() {

            @Override
            public void onSuccess(Object result) {
                testResult.set(result);
            }

            @Override
            public void onError(Throwable error) {
                throw new RuntimeException(error);
            }
        });

        verify(apiMock, times(1)).requestA();
        verify(apiMock, times(1)).requestB();

        assertNotNull(testResult.get());
    }
}

由于某种原因apiMock,从不调用方法,验证总是失败.

在我看来,我这样称呼我的api

apiV2.requestA()
    .subscribeOn(Schedulers.io())
    .observeOn(AndroidSchedulers.mainThread())
    .subscribe(new Observer());

我在这里错过了什么？

更新#1:
经过一些调查后,似乎在我的实施中(上面的示例)我observeOn(AndroidSchedulers.mainThread())没有调用订户.还是不知道为什么.

更新#2:
当订阅就像这样apiV2.requestA().subscribe(new Observer());一切正常 - 模拟api被调用并且测试通过.
前进ShadowLooper.idleMainLooper(5000)没有做任何事.甚至从处理程序中抓住了looper HandlerThreadScheduler并将其推进.结果相同.

更新#3: 添加使用API​​的实际代码.

public void updateView(final Callback) {
    Observable.zip(wrapObservable(api.requestA()), wrapObservable(api.requestB()),
        new Func2<ResponseA, ResponseB, Object>() {
            @Override
            public Object call(ResponseA responseA, ResponseB responseB) {
                return mergeBothResponses(responseA, responseB);
            }
        }
    ).subscribe(new EndlessObserver<Object>() {

        @Override
        public void onError(Throwable e) {
            Log.e(e);
            listener.onError(e);
        }

        @Override
        public void onNext(Object config) {
            Log.d("Configuration updated [%s]", config.toString());
            listener.onSuccess(config);
        }
    });
}

protected <T> Observable<T> wrapObservable(Observable<T> observable) {
    return observable.subscribeOn(Schedulers.io()).observeOn(AndroidSchedulers.mainThread());
}

Answer 1

我仍在围绕如何正确使用rxjava,但我会尝试修改你的代码,以便你只在最终的压缩Observable上观察On(mainThread),而不是在原始请求响应的Observable上进行.然后我会验证这是否会影响你必须推进两个Loopers的事实.

为了简单地进行测试并消除对Looper空闲的需要,我会将线程从等式中取出,因为在运行测试时不需要后台处理.您可以通过注入Scheduler而不是静态创建它们来实现.在运行生产代码时,您需要注入AndroidSchedulers.mainThread和Schedulers.io,并且在运行测试代码时,您可以在适用的情况下注入Schedulers.immediate.

@Inject
@UIScheduler /* Inject Schedulers.immediate for tests and AndroidSchedulers.mainThread for production code */
private Scheduler mainThreadSched;

@Inject
@IOScheduler /* Inject Scheduler.immediate for tests and Schedulers.io for production code */
private Scheduler ioSched;

public void updateView(final Callback) {
    Observable.zip(wrapObservable(api.requestA()), wrapObservable(api.requestB()),
        new Func2<ResponseA, ResponseB, Object>() {
            @Override
            public Object call(ResponseA responseA, ResponseB responseB) {
                return mergeBothResponses(responseA, responseB);
            }
        }
    ).observeOn(mainThreadSched)
    .subscribe(new EndlessObserver<Object>() {

        @Override
        public void onError(Throwable e) {
            Log.e(e);
            listener.onError(e);
        }

        @Override
        public void onNext(Object config) {
            Log.d("Configuration updated [%s]", config.toString());
            listener.onSuccess(config);
        }
    });
}

protected <T> Observable<T> wrapObservable(Observable<T> observable) {
    return observable.subscribeOn(ioSched);
}