Files
HMCL/HMCLAPI/src/main/java/rx/operators/OperationZip.java
2016-01-01 11:03:09 +08:00

285 lines
11 KiB
Java
Executable File

/**
* Copyright 2013 Netflix, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package rx.operators;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentLinkedQueue;
import java.util.concurrent.atomic.AtomicBoolean;
import rx.Observable;
import rx.Observer;
import rx.Subscription;
import rx.util.AtomicObservableSubscription;
import rx.util.SynchronizedObserver;
import rx.util.functions.Func1;
import rx.util.functions.Func2;
import rx.util.functions.Func3;
import rx.util.functions.Func4;
import rx.util.functions.FuncN;
import rx.util.functions.Functions;
public final class OperationZip {
public static <T0, T1, R> Func1<Observer<R>, Subscription> zip(Observable<T0> w0, Observable<T1> w1, Func2<T0, T1, R> zipFunction) {
Aggregator<R> a = new Aggregator<>(Functions.fromFunc(zipFunction));
a.addObserver(new ZipObserver<>(a, w0));
a.addObserver(new ZipObserver<>(a, w1));
return a;
}
public static <T0, T1, T2, R> Func1<Observer<R>, Subscription> zip(Observable<T0> w0, Observable<T1> w1, Observable<T2> w2, Func3<T0, T1, T2, R> zipFunction) {
Aggregator<R> a = new Aggregator<>(Functions.fromFunc(zipFunction));
a.addObserver(new ZipObserver<>(a, w0));
a.addObserver(new ZipObserver<>(a, w1));
a.addObserver(new ZipObserver<>(a, w2));
return a;
}
public static <T0, T1, T2, T3, R> Func1<Observer<R>, Subscription> zip(Observable<T0> w0, Observable<T1> w1, Observable<T2> w2, Observable<T3> w3, Func4<T0, T1, T2, T3, R> zipFunction) {
Aggregator<R> a = new Aggregator<>(Functions.fromFunc(zipFunction));
a.addObserver(new ZipObserver<>(a, w0));
a.addObserver(new ZipObserver<>(a, w1));
a.addObserver(new ZipObserver<>(a, w2));
a.addObserver(new ZipObserver<>(a, w3));
return a;
}
/*
* ThreadSafe
*/
private static class ZipObserver<R, T> implements Observer<T> {
final Observable<T> w;
final Aggregator<R> a;
private final AtomicObservableSubscription subscription = new AtomicObservableSubscription();
private final AtomicBoolean subscribed = new AtomicBoolean(false);
public ZipObserver(Aggregator<R> a, Observable<T> w) {
this.a = a;
this.w = w;
}
public void startWatching() {
if (subscribed.compareAndSet(false, true))
// only subscribe once even if called more than once
subscription.wrap(w.subscribe(this));
}
@Override
public void onCompleted() {
a.complete(this);
}
@Override
public void onError(Exception e) {
a.error(this, e);
}
@Override
public void onNext(T args) {
try {
a.next(this, args);
} catch (Exception e) {
onError(e);
}
}
}
/**
* Receive notifications from each of the Observables we are reducing and
* execute the zipFunction whenever we have received events from all
* Observables.
*
* This class is thread-safe.
*
* @param <T>
*/
private static class Aggregator<T> implements Func1<Observer<T>, Subscription> {
private volatile SynchronizedObserver<T> observer;
private final FuncN<T> zipFunction;
private final AtomicBoolean started = new AtomicBoolean(false);
private final AtomicBoolean running = new AtomicBoolean(true);
private final ConcurrentHashMap<ZipObserver<T, ?>, Boolean> completed = new ConcurrentHashMap<>();
/*
* we use ConcurrentHashMap despite synchronization of methods because
* stop() does NOT use synchronization and this map is used by it and
* can be called by other threads
*/
private final ConcurrentHashMap<ZipObserver<T, ?>, ConcurrentLinkedQueue<Object>> receivedValuesPerObserver = new ConcurrentHashMap<>();
/*
* we use a ConcurrentLinkedQueue to retain ordering (I'd like to just
* use a ConcurrentLinkedHashMap for 'receivedValuesPerObserver' but
* that doesn't exist in standard java
*/
private final ConcurrentLinkedQueue<ZipObserver<T, ?>> observers = new ConcurrentLinkedQueue<>();
public Aggregator(FuncN<T> zipFunction) {
this.zipFunction = zipFunction;
}
/**
* Receive notification of a Observer starting (meaning we should
* require it for aggregation)
*
* Thread Safety => Invoke ONLY from the static factory methods at top
* of this class which are always an atomic execution by a single
* thread.
*
* @param w
*/
private void addObserver(ZipObserver<T, ?> w) {
// initialize this ZipObserver
observers.add(w);
receivedValuesPerObserver.put(w, new ConcurrentLinkedQueue<>());
}
/**
* Receive notification of a Observer completing its iterations.
*
* @param w
*/
void complete(ZipObserver<T, ?> w) {
// store that this ZipObserver is completed
completed.put(w, Boolean.TRUE);
// if all ZipObservers are completed, we mark the whole thing as completed
if (completed.size() == observers.size())
if (running.compareAndSet(true, false))
// this thread succeeded in setting running=false so let's propagate the completion
// mark ourselves as done
observer.onCompleted();
}
/**
* Receive error for a Observer. Throw the error up the chain and stop
* processing.
*
* @param w
*/
void error(ZipObserver<T, ?> w, Exception e) {
if (running.compareAndSet(true, false)) {
// this thread succeeded in setting running=false so let's propagate the error
observer.onError(e);
/*
* since we receive an error we want to tell everyone to stop
*/
stop();
}
}
/**
* Receive the next value from a Observer.
* <p>
* If we have received values from all Observers, trigger the zip
* function, otherwise store the value and keep waiting.
*
* @param w
* @param arg
*/
void next(ZipObserver<T, ?> w, Object arg) {
if (observer == null)
throw new RuntimeException("This shouldn't be running if a Observer isn't registered");
/*
* if we've been 'unsubscribed' don't process anything further even
* if the things we're watching keep sending (likely because they
* are not responding to the unsubscribe call)
*/
if (!running.get())
return;
// store the value we received and below we'll decide if we are to send it to the Observer
receivedValuesPerObserver.get(w).add(arg);
// define here so the variable is out of the synchronized scope
Object[] argsToZip = new Object[observers.size()];
/*
* we have to synchronize here despite using concurrent data
* structures because the compound logic here must all be done
* atomically
*/
synchronized (this) {
// if all ZipObservers in 'receivedValues' map have a value, invoke the zipFunction
for (ZipObserver<T, ?> rw : receivedValuesPerObserver.keySet())
if (receivedValuesPerObserver.get(rw).peek() == null)
// we have a null meaning the queues aren't all populated so won't do anything
return;
// if we get to here this means all the queues have data
int i = 0;
for (ZipObserver<T, ?> rw : observers)
argsToZip[i++] = receivedValuesPerObserver.get(rw).remove();
}
// if we did not return above from the synchronized block we can now invoke the zipFunction with all of the args
// we do this outside the synchronized block as it is now safe to call this concurrently and don't need to block other threads from calling
// this 'next' method while another thread finishes calling this zipFunction
observer.onNext(zipFunction.call(argsToZip));
}
@Override
public Subscription call(Observer<T> observer) {
if (started.compareAndSet(false, true)) {
AtomicObservableSubscription subscription = new AtomicObservableSubscription();
this.observer = new SynchronizedObserver<>(observer, subscription);
/*
* start the Observers
*/
for (ZipObserver<T, ?> rw : observers)
rw.startWatching();
return subscription.wrap(this::stop);
} else
/*
* a Observer already has subscribed so blow up
*/
throw new IllegalStateException("Only one Observer can subscribe to this Observable.");
}
/*
* Do NOT synchronize this because it gets called via unsubscribe which
* can occur on other threads
* and result in deadlocks. (http://jira/browse/API-4060)
*
* AtomicObservableSubscription uses compareAndSet instead of locking to
* avoid deadlocks but ensure single-execution.
*
* We do the same in the implementation of this method.
*
* ThreadSafety of this method is provided by:
* - AtomicBoolean[running].compareAndSet
* - ConcurrentLinkedQueue[Observers]
* - ZipObserver.subscription being an AtomicObservableSubscription
*/
private void stop() {
/*
* tell ourselves to stop processing onNext events by setting
* running=false
*/
if (running.compareAndSet(true, false))
/*
* propogate to all Observers to unsubscribe if this thread
* succeeded in setting running=false
*/
for (ZipObserver<T, ?> o : observers)
if (o.subscription != null)
o.subscription.unsubscribe();
}
}
}