Rx Observable定期发射值

时间:2014-07-03 15:00:04

标签: android rx-java

我必须定期轮询一些RESTful端点以刷新我的Android应用程序的数据。我还必须根据连接暂停和恢复它(如果手机处于离线状态,则无需尝试)。我目前的解决方案是有效的,但它使用标准的Java ScheduledExecutorService来执行定期任务,但我想留在Rx范例中。

这是我当前的代码,为简洁起见,跳过部分代码。

userProfileObservable = Observable.create(new Observable.OnSubscribe<UserProfile>() {
    @Override
    public void call(final Subscriber<? super UserProfile> subscriber) {
        final ScheduledExecutorService scheduledExecutorService = Executors.newSingleThreadScheduledExecutor();
        final Runnable runnable = new Runnable() {
            @Override
            public void run() {
                // making http request here
            }
        };
        final List<ScheduledFuture<?>> futures = new ArrayList<ScheduledFuture<?>>(1);
        networkStatusObservable.subscribe(new Action1<Boolean>() {
            @Override
            public void call(Boolean networkAvailable) {
                if (!networkAvailable) {
                    pause();
                } else {
                    pause();                        
                    futures.add(scheduledExecutorService.scheduleWithFixedDelay(runnable, 0, SECOND_IN_MILLIS * SECONDS_TO_EXPIRE, TimeUnit.MILLISECONDS));
                }
            }

            private void pause() {
                for (ScheduledFuture<?> future : futures) {
                    future.cancel(true);
                }
                futures.clear();
            }
        });

        final Subscription subscription = new Subscription() {
            private boolean isUnsubscribed = false;

            @Override
            public void unsubscribe() {
                scheduledExecutorService.shutdownNow();
                isUnsubscribed = true;
            }

            @Override
            public boolean isUnsubscribed() {
                return isUnsubscribed;
            }
        };
        subscriber.add(subscription);
    }
}).multicast(BehaviorSubject.create()).refCount();

networkStatusObservable基本上是一个包含在Observable<Boolean>中的广播接收器,表示手机已连接到网络。

正如我所说,这个解决方案正在运行,但我想使用Rx方法进行定期轮询并发出新的UserProfile,因为手动调度事件存在很多问题,我想避免这些问题。我知道Observable.timerObservable.interval,但无法弄清楚如何将它们应用于此任务(我不确定是否需要使用它们)。

5 个答案:

答案 0 :(得分:26)

在这个GitHub问题上有一些方法可能会对您有所帮助。

https://github.com/ReactiveX/RxJava/issues/448

这三个实现是:


Observable.interval

Observable.interval(delay, TimeUnit.SECONDS).timeInterval()
        .flatMap(new Func1<Long, Observable<Notification<AppState>>>() {
            public Observable<Notification<AppState>> call(Long seconds) {
                return lyftApi.updateAppState(params).materialize(); } });

Scheduler.schedulePeriodically

Observable.create({ observer ->
        Schedulers.newThread().schedulePeriodically({
            observer.onNext("application-state-from-network");
        }, 0, 1000, TimeUnit.MILLISECONDS);
    }).take(10).subscribe({ v -> println(v) });

Manual Recursion

Observable.create(new OnSubscribeFunc<String>() {
        @Override
        public Subscription onSubscribe(final Observer<? super String> o) {
            return Schedulers.newThread().schedule(0L, new Func2<Scheduler, Long, Subscription>() {
                @Override
                public Subscription call(Scheduler inner, Long t2) {
                    o.onNext("data-from-polling");
                    return inner.schedule(t2, this, 1000, TimeUnit.MILLISECONDS);
                }
            });
        }
    }).toBlockingObservable().forEach(new Action1<String>() {
        @Override
        public void call(String v) {
            System.out.println("output: " + v);
        }
    });

结论是手动递归是要走的路,因为它等到操作完成后再安排下一次执行。

答案 1 :(得分:4)

其中一个选项是使用Observable.interval并在发出间隔时检查用户状态:

     Observable<Long> interval = Observable.interval(1, TimeUnit.SECONDS);

    //pulling the user data
    Observable<Observable<String>> userObservable = interval.map(new Func1<Long, Observable<String>>() {
        Random random = new Random();
        @Override
        public Observable<String> call(Long tick) {
            //here you are pulling user data; you should do it asynchronously - rx way - because the interval is using Schedulers.computation which is not best suited for doing io operations
            switch(random.nextInt(10)){
                case 0://suppose this is for cases when network in  not available or exception happens
                    return Observable.<String>just(null);
                case 1:
                case 2:
                    return Observable.just("Alice");
                default:
                    return Observable.just("Bob");
            }
        }
    });

    Observable<String> flatUsers = userObservable.flatMap(new Func1<Observable<String>, Observable<? extends String>>() {
        @Override
        public Observable<? extends String> call(Observable<String> stringObservable) {
            return stringObservable;
        }
    });

    //filter valid data
    Observable<String> usersWithoutErrors = flatUsers.filter(new Func1<String, Boolean>() {
        @Override
        public Boolean call(String s) {
            return s != null;
        }
    });

    //publish only changes
    Observable<String> uniqueUsers = usersWithoutErrors.distinctUntilChanged();

如果您的networkStatusObservable发送至少与检查用户数据一样频繁的事件,您可以更简单地执行此操作

 networkStatusObservable.sample(1,TimeUnit.Seconds).filter(/*the best is to filter only connected state */).map(/*now start pulling the user data*/)

最后,您可以创建使用调度程序定期发出用户状态的observable - 请参阅Schedulers documentation以了解哪种调度程序最适合您:

public abstract class ScheduledOnSubscribe<T> implements Observable.OnSubscribe<T>{
    private final Scheduler scheduler;
    private final long initialDelay;
    private final long period;
    private final TimeUnit unit;

    public ScheduledOnSubscribe(Scheduler scheduler, long initialDelay, long period, TimeUnit unit) {
        this.scheduler = scheduler;
        this.initialDelay = initialDelay;
        this.period = period;
        this.unit = unit;
    }

    abstract T next() throws Exception;


    @Override
    public void call(final Subscriber<? super T> subscriber) {
        final Scheduler.Worker worker = scheduler.createWorker();
        subscriber.add(worker);
        worker.schedulePeriodically(new Action0() {
            @Override
            public void call() {
                try {
                    subscriber.onNext(next());
                } catch (Throwable e) {
                    try {
                        subscriber.onError(e);
                    } finally {
                        worker.unsubscribe();
                    }
                }
            }

        }, initialDelay, period, unit);
    }
}

//And here is the sample usage
 Observable<String> usersObservable = Observable.create(new ScheduledOnSubscribe(Schedulers.io(), 1, 1, TimeUnit.SECONDS ){
        Random random = new Random();
        @Override
        String next() throws Exception {
            //if you use Schedulers.io, you can call the remote service synchronously
            switch(random.nextInt(10)){
                case 0:
                    return null;
                case 1:
                case 2:
                    return "Alice";
                default:
                    return "Bob";
            }
        }
    });

答案 2 :(得分:3)

使用interval()有一种更简单的方法。我测试了这段代码,但它确实有效。 但首先,您应该封装要在Action1的子类中定期执行的作业。

class Act<T> implements Action1<T> {
     public Service service;
     public String data;
     public void call(T t){
         service.log(data); //the periodic job
     }
}

(为了简洁起见,我公开了这些字段,但这并不可取)。现在您可以通过以下方式安排它:

Act<Long> act=new Act<>();
act.data="dummy data";
act.service=this;
Observable.interval(0l, period, TimeUnit.SECONDS).subscribeOn(Schedulers.from(Executors.newFixedThreadPool(10))).subscribe((Action1<Long>)act);

这不会阻止你的线程,不像其他答案给出的方法。这种方法允许我们将变量作为Action中的一种可变存储传递,这在后续调用中很方便。此外,这样您就可以在自己的线程池上订阅您的调用。

答案 3 :(得分:3)

简短回答。 RxJava2:

Observable.interval(initialDelay, unitAmount, timeUnit)
            .subscribe(value -> {
                // code for periodic execution
            });

根据您的需要选择initialDelay,unitAmount和TimeUnit。

示例:0,1,TimeUnit.MINUTES。

答案 4 :(得分:1)

好的,我会发布自己的解决方案,也许有人会从中受益。我只会发布与问题相关的部分,省略HTTP和缓存内容。以下是我的表现方式:

private ConnectableObservable<Long> createNetworkBoundHeartbeatObservable(final Observable<Boolean> networkStatusObservable,
                                                                          final Observable<Boolean> pauseResumeObservable) {

    final Observable<Boolean> pausableHeartbeatObservable = Observable.combineLatest(networkStatusObservable, pauseResumeObservable,
            new Func2<Boolean, Boolean, Boolean>() {
                @Override
                public Boolean call(Boolean networkAvailable, Boolean mustPause) {
                    return mustPause && networkAvailable;
                }
            }
    ).distinctUntilChanged();

    final Observable<Boolean> hasToResumeObservable = pausableHeartbeatObservable.filter(new Func1<Boolean, Boolean>() {
        @Override
        public Boolean call(Boolean networkAvailable) {
            return networkAvailable;
        }
    });
    final Observable<Boolean> hasToStopObservable = pausableHeartbeatObservable.filter(new Func1<Boolean, Boolean>() {
        @Override
        public Boolean call(Boolean networkAvailable) {
            return !networkAvailable;
        }
    });


    return pausableHeartbeatObservable.concatMap(new Func1<Boolean, Observable<Long>>() {
        @Override
        public Observable<Long> call(Boolean shouldResumeRequests) {
            if (shouldResumeRequests) {
                long timeToUpdate;
                final Date oldestModifiedExpiresAt = cache.oldestModifiedExpiresAt();
                timeToUpdate = Math.max(0, oldestModifiedExpiresAt.getTime() - System.currentTimeMillis());
                Log.d(TAG, String.format("Have to restart updates, %d seconds till next update", timeToUpdate / SECOND_IN_MILLIS));
                return Observable
                        .timer(timeToUpdate, SECONDS_TO_EXPIRE * SECOND_IN_MILLIS, TimeUnit.MILLISECONDS)
                        .takeUntil(hasToStopObservable);
            } else {
                Log.d(TAG, "Have to pause updates");
                return Observable.<Long>never().takeUntil(hasToResumeObservable);
            }
        }
    }).multicast(PublishSubject.<Long>create());
}

正如您所看到的,暂停或恢复更新的条件变得有点复杂,添加了一个新的Observable以支持在应用程序转到后台时暂停。

然后解决方案的核心是concatMap操作,它按顺序发出Observables(因此concatMap,而不是flatMap,请参阅此问题:What is the difference between concatMap and flatMap in RxJava)。它会发出intervalnever Observables,具体取决于是应该继续更新还是暂停更新。然后,每个发出的Observable takenUntil都是Observable&#39; ConnectableObservable会发出新的价值。

返回

Observable是因为创建的connect很热,并且所有预期订阅者在开始发送内容之前都必须订阅它,否则初始事件可能会丢失。我稍后会打电话给{{1}}。

如果有的话,我会根据投票接受我或其他答案。