我正在尝试使用rx Java中的backpressure在我的Android应用程序中创建无限滚动。我希望它仅在请求的次数上调用外部服务(在调用request(1)
之后)。但在使用flatmap后,每subscribe
次加载16页。
在我的代码下面,带有预期的结果。几乎每个测试都因第一次请求(n = 16)而失败
import org.junit.Before;
import org.junit.Test;
import org.mockito.Mockito;
import rx.Observable;
import rx.observers.TestSubscriber;
import java.util.Arrays;
import java.util.List;
import java.util.concurrent.atomic.AtomicInteger;
import static java.util.Collections.emptyList;
import static org.mockito.Mockito.*;
import static rx.internal.util.UtilityFunctions.identity;
public class ServiceObservablesTest {
public static <T> Observable<List<T>> create(DataProvider<T> dataProvider) {
Observable<Observable<List<T>>> metaObservalble = Observable.create(subscriber -> {
AtomicInteger pageNumber = new AtomicInteger();
subscriber.setProducer(n -> {
// at subscribe rxJava makes request for 16 elements - probably because of flatMap
// after first request with 16 elements everything seems to work fine even if i ignore the 'n' param
Observable<List<T>> page = dataProvider.requestPage(pageNumber.getAndIncrement());
subscriber.onNext(page);
});
});
return metaObservalble.flatMap(identity()).takeWhile(page -> !page.isEmpty());
}
public interface DataProvider<T> {
Observable<List<T>> requestPage(int page);
}
private DataProvider provider;
@Before
public void setUp() throws Exception {
provider = Mockito.mock(DataProvider.class);
List<Object> list = Arrays.asList(new Object());
when(provider.requestPage(anyInt())).thenReturn(Observable.just(list));
}
@Test
public void shouldRequestOnlyFirstPageOnSubscribe() {
//given
TestSubscriber<List<Object>> subscriber = new TestSubscriber<>(1);
Observable<List<Object>> flightsObservable = create(provider);
//when
flightsObservable.subscribe(subscriber);
//then
subscriber.assertValueCount(1);
subscriber.assertNotCompleted();
verify(provider, times(1)).requestPage(0);
verify(provider, never()).requestPage(1);
}
@Test
public void shouldRequestNumberOfPagesSpecified() {
//given
int requested_pages = 5;
TestSubscriber<List<Object>> subscriber = new TestSubscriber<>(0);
Observable<List<Object>> flightsObservable = create(provider);
//when
flightsObservable.subscribe(subscriber);
subscriber.requestMore(requested_pages);
//then
subscriber.assertValueCount(requested_pages);
subscriber.assertNotCompleted();
for (int i = 0; i < requested_pages; i++) {
verify(provider, times(1)).requestPage(i);
}
verify(provider, never()).requestPage(requested_pages);
}
@Test
public void shouldCompleteAfterRetrievingEmptyResult() {
//given
int emptyPage = 2;
when(provider.requestPage(emptyPage)).thenReturn(Observable.just(emptyList()));
TestSubscriber<List<Object>> subscriber = new TestSubscriber<>(100);
Observable<List<Object>> flightsObservable = create(provider);
//when
flightsObservable.subscribe(subscriber);
//then
subscriber.assertValueCount(emptyPage);
subscriber.assertCompleted();
verify(provider, times(1)).requestPage(0); //requested at subscribe
for (int i = 1; i <= emptyPage; i++) {
verify(provider, times(1)).requestPage(i);
}
verify(provider, never()).requestPage(emptyPage + 1);
}
@Test
public void shouldRequestNextPageWhenRequestedMore() {
//given
TestSubscriber<List<Object>> subscriber = new TestSubscriber<>(1);
Observable<List<Object>> flightsObservable = create(provider);
//when
flightsObservable.subscribe(subscriber);
subscriber.requestMore(1);
//then
subscriber.assertValueCount(2);
verify(provider, times(1)).requestPage(0);
verify(provider, times(1)).requestPage(1);
verify(provider, never()).requestPage(2);
//when
subscriber.requestMore(1);
//then
subscriber.assertValueCount(3);
subscriber.assertNotCompleted();
verify(provider, times(1)).requestPage(0);
verify(provider, times(1)).requestPage(1);
verify(provider, times(1)).requestPage(2);
verify(provider, never()).requestPage(3);
}
@Test
public void shouldWorkWithMultipleSubscribers() {
//given
TestSubscriber<List<Object>> subscriber1 = new TestSubscriber<>(1);
TestSubscriber<List<Object>> subscriber2 = new TestSubscriber<>(1);
Observable<List<Object>> flightsObservable = create(provider);
//when
flightsObservable.subscribe(subscriber1);
flightsObservable.subscribe(subscriber2);
//then
subscriber1.assertValueCount(1);
subscriber2.assertValueCount(1);
verify(provider, times(2)).requestPage(0);
verify(provider, never()).requestPage(1);
//when
subscriber1.requestMore(1);
//then
subscriber1.assertValueCount(2);
subscriber2.assertValueCount(1);
verify(provider, times(2)).requestPage(0);
verify(provider, times(1)).requestPage(1);
verify(provider, never()).requestPage(2);
//when
subscriber2.requestMore(1);
//then
subscriber1.assertValueCount(2);
subscriber2.assertValueCount(2);
verify(provider, times(2)).requestPage(0);
verify(provider, times(2)).requestPage(1);
verify(provider, never()).requestPage(2);
}
}
答案 0 :(得分:3)
背压旨在协商并发的消费者生产者行为,并允许程序作者设定策略,以便在产生的数据速率超过所消耗的数据速率时解决该做什么。
也就是说,您会看到组合merge
等可观察对象的运算符会为您提供与您所需数据量不对应的请求数量。外部可观察对象(Observable of Observables)在合并时将始终在RxAndroid(RxJava中为128)上收到16的请求。然后,当它接收内部可观察列表时,每个内部可观察量将接收基于来自下游订户的请求量的请求。如果您尝试编写Observable<Observable<T>>
,您将被迫编写一个OnSubscribe<Observable<List<T>>>
函数,该函数在内部管理合并行为,使其为Observable<List<T>>
而不是Observable<Observable<List<T>>
。写这将强制您订阅数据提供程序返回的observable以展开{on List<T>
。
我建议您将屏幕y位置映射到End-Of-Page事件,然后使用扫描将其转换为单调递增的数字,然后将该数字concatMap映射到DataProvider.requestPage()
的调用中。
screenYPositions
.map(this::isUninitializedOrNearEndOfPage)
.scan(1, (event, pageNumber) -> pageNumber + 1 )
.concatMap(dataProvider::requestPage)
.subscribe(testSubscriber);