disruptor-3.3.2源码解析(5)-框架支持

阅读更多

disruptor-3.3.2源码解析(5)-框架支持

作者：大飞

 

• 更方便的使用Disruptor：

       前面几篇看了Disruptor中的一些重要组件和组件的运行方式，也通过手动组合这些组件的方式给出了一些基本的用例。框架也提供了一个DSL-style API，来帮助我们更容易的使用框架，屏蔽掉一些细节(比如怎么构建RingBuffer、怎么关联追踪序列等)，相当于Builder模式。

       在看Disruptor之前，先看一些辅助类，首先看下ConsumerRepository：

class ConsumerRepository implements Iterable{

    private final Map, EventProcessorInfo> eventProcessorInfoByEventHandler = new IdentityHashMap, EventProcessorInfo>();
    private final Map eventProcessorInfoBySequence = new IdentityHashMap();
    private final Collection consumerInfos = new ArrayList();

       可见ConsumerRepository内部存储着事件处理者(消费者)的信息，相当于事件处理者的仓库。

       看一下里面的方法： 

    public void add(final EventProcessor eventprocessor,
                    final EventHandler handler,
                    final SequenceBarrier barrier){
        final EventProcessorInfo consumerInfo = new EventProcessorInfo(eventprocessor, handler, barrier);
        eventProcessorInfoByEventHandler.put(handler, consumerInfo);
        eventProcessorInfoBySequence.put(eventprocessor.getSequence(), consumerInfo);
        consumerInfos.add(consumerInfo);
    }

       添加事件处理者(Event模式)、事件处理器和序列栅栏到仓库中。 

 

    public void add(final EventProcessor processor){
        final EventProcessorInfo consumerInfo = new EventProcessorInfo(processor, null, null);
        eventProcessorInfoBySequence.put(processor.getSequence(), consumerInfo);
        consumerInfos.add(consumerInfo);
    }

       添加事件处理者(Event模式)到仓库中。 

 

    public void add(final WorkerPool workerPool, final SequenceBarrier sequenceBarrier){
        final WorkerPoolInfo workerPoolInfo = new WorkerPoolInfo(workerPool, sequenceBarrier);
        consumerInfos.add(workerPoolInfo);
        for (Sequence sequence : workerPool.getWorkerSequences()){
            eventProcessorInfoBySequence.put(sequence, workerPoolInfo);
        }
    }

       添加事件处理者(Work模式)和序列栅栏到仓库中。 

 

    public Sequence[] getLastSequenceInChain(boolean includeStopped){
        List lastSequence = new ArrayList();
        for (ConsumerInfo consumerInfo : consumerInfos){
            if ((includeStopped || consumerInfo.isRunning()) && consumerInfo.isEndOfChain()){
                final Sequence[] sequences = consumerInfo.getSequences();
                Collections.addAll(lastSequence, sequences);
            }
        }
        return lastSequence.toArray(new Sequence[lastSequence.size()]);
    }

       获取当前已经消费到RingBuffer上事件队列末尾的事件处理者的序列，可通过参数指定是否要包含已经停止的事件处理者。 

 

    public void unMarkEventProcessorsAsEndOfChain(final Sequence... barrierEventProcessors){
        for (Sequence barrierEventProcessor : barrierEventProcessors){
            getEventProcessorInfo(barrierEventProcessor).markAsUsedInBarrier();
        }
    }

       重置已经处理到事件队列末尾的事件处理者的状态。

 

       其他方法就不看了。

 

       上面代码中出现的ConsumerInfo就相当于事件处理者信息和序列栅栏的包装类，ConsumerInfo本身是一个接口，针对Event模式和Work模式提供了两种实现：EventProcessorInfo和WorkerPoolInfo，代码都很容易理解，这里就不贴了。

 

        现在来看下Disruptor类，先看结构： 

public class Disruptor{

    //事件队列。
    private final RingBuffer ringBuffer;
    //用于执行事件处理的执行器。
    private final Executor executor;
    //事件处理信息仓库。
    private final ConsumerRepository consumerRepository = new ConsumerRepository();
    //运行状态。
    private final AtomicBoolean started = new AtomicBoolean(false);
    //异常处理器。
    private ExceptionHandler exceptionHandler;

       可见，Disruptor内部包含了我们之前写用例使用到的所有组件。

 

        再看下构造方法： 

    public Disruptor(final EventFactory eventFactory, final int ringBufferSize, final Executor executor){
        this(RingBuffer.createMultiProducer(eventFactory, ringBufferSize), executor);
    }

    public Disruptor(final EventFactory eventFactory,
                     final int ringBufferSize,
                     final Executor executor,
                     final ProducerType producerType,
                     final WaitStrategy waitStrategy){
        this(RingBuffer.create(producerType, eventFactory, ringBufferSize, waitStrategy),
             executor);
    }

    private Disruptor(final RingBuffer ringBuffer, final Executor executor)
    {
        this.ringBuffer = ringBuffer;
        this.executor = executor;
    }

       可见，通过构造方法，可以对内部的RingBuffer和执行器进行初始化。

 

        有了事件队列(RingBuffer)，接下来看看怎么构建事件处理者： 

    @SuppressWarnings("varargs")
    public EventHandlerGroup handleEventsWith(final EventHandler... handlers){
        return createEventProcessors(new Sequence[0], handlers);
    }
    EventHandlerGroup createEventProcessors(final Sequence[] barrierSequences,
                                               final EventHandler[] eventHandlers){
        checkNotStarted();
        final Sequence[] processorSequences = new Sequence[eventHandlers.length];
        final SequenceBarrier barrier = ringBuffer.newBarrier(barrierSequences);
        for (int i = 0, eventHandlersLength = eventHandlers.length; i < eventHandlersLength; i++){
            final EventHandler eventHandler = eventHandlers[i];
            final BatchEventProcessor batchEventProcessor = new BatchEventProcessor(ringBuffer, barrier, eventHandler);
            if (exceptionHandler != null){
                batchEventProcessor.setExceptionHandler(exceptionHandler);
            }
            consumerRepository.add(batchEventProcessor, eventHandler, barrier);
            processorSequences[i] = batchEventProcessor.getSequence();
        }
        if (processorSequences.length > 0){
            consumerRepository.unMarkEventProcessorsAsEndOfChain(barrierSequences);
        }
        return new EventHandlerGroup(this, consumerRepository, processorSequences);
    }

       可见，handleEventsWith方法内部会创建BatchEventProcessor。

       当然，对于Event模式，还有一些玩法，其实之前几篇就看到过，我们可以设置两个EventHandler，然后事件会依次被这两个handler处理。Disruptor类中提供了更明确的定义(事实是结合了EventHandlerGroup的一些方法)，比如我想让事件先被处理器a处理，然后在被处理器b处理，就可以这么写： 

    EventHandler a = new EventHandler() { ... };
    EventHandler b = new EventHandler() { ... };
    disruptor.handleEventsWith(a); //语句1
    disruptor.after(a).handleEventsWith(b);

       注意上面必须先写语句1，然后才能针对a调用after，否则after找不到处理器a，会报错。

       上面的例子也可以这么写： 

    EventHandler a = new EventHandler() { ... };
    EventHandler b = new EventHandler() { ... };
    disruptor.handleEventsWith(a).then(b);

       效果是一样的。

 

       Disruptor还允许我们定制事件处理者： 

    public EventHandlerGroup handleEventsWith(final EventProcessorFactory... eventProcessorFactories){
        final Sequence[] barrierSequences = new Sequence[0];
        return createEventProcessors(barrierSequences, eventProcessorFactories);
    }
public interface EventProcessorFactory{
    EventProcessor createEventProcessor(RingBuffer ringBuffer, Sequence[] barrierSequences);
}

 

       handleEventsWith方法内部创建的Event模式的事件处理者，有没有Work模式的呢？

    @SuppressWarnings("varargs")
    public EventHandlerGroup handleEventsWithWorkerPool(final WorkHandler... workHandlers){
        return createWorkerPool(new Sequence[0], workHandlers);
    }
    EventHandlerGroup createWorkerPool(final Sequence[] barrierSequences, final WorkHandler[] workHandlers){
        final SequenceBarrier sequenceBarrier = ringBuffer.newBarrier(barrierSequences);
        final WorkerPool workerPool = new WorkerPool(ringBuffer, sequenceBarrier, exceptionHandler, workHandlers);
        consumerRepository.add(workerPool, sequenceBarrier);
        return new EventHandlerGroup(this, consumerRepository, workerPool.getWorkerSequences());
    }

        handleEventsWithWorkerPool内部会创建WorkerPool。

 

         事件处理者也构建好了，接下来看看怎么启动它们： 

    public RingBuffer start(){
        final Sequence[] gatingSequences = consumerRepository.getLastSequenceInChain(true);
        ringBuffer.addGatingSequences(gatingSequences);
        checkOnlyStartedOnce();
        for (final ConsumerInfo consumerInfo : consumerRepository){
            consumerInfo.start(executor);
        }
        return ringBuffer;
    }

        可见，启动过程中会将事件处理者的序列设置为RingBuffer的追踪序列。最后会启动事件处理者，并利用执行器来执行事件处理线程。

 

         事件处理者构建好了，也启动了，看看怎么发布事件： 

    public void publishEvent(final EventTranslator eventTranslator){
        ringBuffer.publishEvent(eventTranslator);
    }

        很简单，里面就是直接调用了RingBuffer来发布事件，之前几篇都分析过了。

 

        再看看其他的方法：

    public void halt(){
        for (final ConsumerInfo consumerInfo : consumerRepository){
            consumerInfo.halt();
        }
    }

        停止事件处理者。  

 

    public void shutdown(){
        try{
            shutdown(-1, TimeUnit.MILLISECONDS);
        }catch (final TimeoutException e){
            exceptionHandler.handleOnShutdownException(e);
        }
    }
    public void shutdown(final long timeout, final TimeUnit timeUnit) throws TimeoutException{
        final long timeOutAt = System.currentTimeMillis() + timeUnit.toMillis(timeout);
        while (hasBacklog()){
            if (timeout >= 0 && System.currentTimeMillis() > timeOutAt){
                throw TimeoutException.INSTANCE;
            }
            // Busy spin
        }
        halt();
    }
    private boolean hasBacklog(){
        final long cursor = ringBuffer.getCursor();
        for (final Sequence consumer : consumerRepository.getLastSequenceInChain(false)){
            if (cursor > consumer.get()){
                return true;
            }
        }
        return false;
    }

        等待所有能处理的事件都处理完了，再定制事件处理者，有超时选项。

 

        好了，其他方法都比较简单，不看了。最后来使用Disruptor写一个生产者消费者模式吧： 

	public static void main(String[] args) {
		//创建一个执行器(线程池)。
		Executor executor = Executors.newFixedThreadPool(4);
		//创建一个Disruptor。
		Disruptor disruptor = 
				new Disruptor(new MyDataEventFactory(), 4, executor);
		//创建两个事件处理器。
		MyDataEventHandler handler1 = new MyDataEventHandler();
		KickAssEventHandler handler2 = new KickAssEventHandler();
		//同一个事件，先用handler1处理再用handler2处理。
		disruptor.handleEventsWith(handler1).then(handler2);
		//启动Disruptor。
		disruptor.start();
		//发布10个事件。
		for(int i=0;i<10;i++){
			disruptor.publishEvent(new MyDataEventTranslator());
			System.out.println("发布事件["+i+"]");
			try {
				TimeUnit.SECONDS.sleep(3);
			} catch (InterruptedException e) {
				e.printStackTrace();
			}
		}
		try {
			System.in.read();
		} catch (IOException e) {
			e.printStackTrace();
		}
	}

        看下输出： 

    发布事件[0]
    handle event's data:MyData [id=0, value=holy shit!]isEndOfBatch:true
    kick your ass 0 times!!!!
    发布事件[1]
    handle event's data:MyData [id=1, value=holy shit!]isEndOfBatch:true
    kick your ass 1 times!!!!
    发布事件[2]
    handle event's data:MyData [id=2, value=holy shit!]isEndOfBatch:true
    kick your ass 2 times!!!!
    发布事件[3]
    handle event's data:MyData [id=3, value=holy shit!]isEndOfBatch:true
    kick your ass 3 times!!!!
    发布事件[4]
    handle event's data:MyData [id=4, value=holy shit!]isEndOfBatch:true
    kick your ass 4 times!!!!
    发布事件[5]
    handle event's data:MyData [id=5, value=holy shit!]isEndOfBatch:true
    kick your ass 5 times!!!!
    发布事件[6]
    handle event's data:MyData [id=6, value=holy shit!]isEndOfBatch:true
    kick your ass 6 times!!!!
    发布事件[7]
    handle event's data:MyData [id=7, value=holy shit!]isEndOfBatch:true
    kick your ass 7 times!!!!
    发布事件[8]
    handle event's data:MyData [id=8, value=holy shit!]isEndOfBatch:true
    kick your ass 8 times!!!!
    发布事件[9]
    handle event's data:MyData [id=9, value=holy shit!]isEndOfBatch:true
    kick your ass 9 times!!!!

 

   

• 最后总结：

       1.使用时可以直接使用Disruptor这个类来更方便的完成代码编写，注意灵活使用。

       2.最后别忘了单线程/多线程生产者、Event/Work处理模式等等。