Fresco Fbcore源码分析_executor(一）

Fbcore简介

Fbcore包含两大模块，一个是common 另一个是datasource ，它主要提供了Fresco 框架通用的接口和相关的API。这些API被Imagepipeline 和 drawee大量的使用

Executor功能分析

Executor类简介

1.CallerThreadExecutor.java

 CallerThreadExecutor是一个单例类，它将线程的执行过程进行了封装，所有的线程的运行都可以通过下面的方法来运行，简单来说他是一个线程的执行者

public void execute(Runnable command) {    command.run();  }

2.ConstrainedExecutorService.java

   ConstrainedExecutorService主要是用来处理多线程的并发操作，它同时也支持自定义并发线程的个数，接下来上代码

构造方法分析

  public ConstrainedExecutorService(      String name,      int maxConcurrency,      Executor executor,      BlockingQueue<Runnable> workQueue) {//这里使用BlockingQueue是为了保证线程不会被阻塞，它基于生产者消费者模型来实现，当队列里线程满的时候他会阻止线程的进入，当队列空的时候，他会阻止从队列里面取出线程    if (maxConcurrency <= 0) {      throw new IllegalArgumentException("max concurrency must be > 0");    }    mName = name;    mExecutor = executor;    mMaxConcurrency = maxConcurrency;//代表最大同时运行线程的个数    mWorkQueue = workQueue;    mTaskRunner = new Worker();//Workr是一个线程，它实现了从队列里拿出一个线程并执行它，同时会实时调整mPendingWorkers 的个数    mPendingWorkers = new AtomicInteger(0);//Pending work 代表正在从队列里出来的正在执行的线程的个数    mMaxQueueSize = new AtomicInteger(0);//代表线成队列的最大值  }

接下来看看这个类主要做了什么？
这个类主要通过下面两个方法来工作：execute 和 startWorkerIfNeeded，下面是具体的分析：

public void execute(Runnable runnable) {    if (runnable == null) {      throw new NullPointerException("runnable parameter is null");    }    if (!mWorkQueue.offer(runnable)) {//将这个线程添加到队列里，如果失败则抛出异常      throw new RejectedExecutionException(          mName + " queue is full, size=" + mWorkQueue.size());    }    final int queueSize = mWorkQueue.size();//获取队列的大小    final int maxSize = mMaxQueueSize.get();//获取自定义的队列的最大值    if ((queueSize > maxSize) && mMaxQueueSize.compareAndSet(maxSize, queueSize)) {      FLog.v(TAG, "%s: max pending work in queue = %d", mName, queueSize);    } // else, there was a race and another thread updated and logged the max queue size    startWorkerIfNeeded();  }

 private void startWorkerIfNeeded() {    // Perform a compare-and-swap retry loop for synchronization to make sure we don't start more    // workers than desired.    int currentCount = mPendingWorkers.get();//获取当前正在运行的线程的个数    while (currentCount < mMaxConcurrency) {//当运行的线程个数小于最大并发线程数的时候，循环从队列里面获取线程并运行      int updatedCount = currentCount + 1;      if (mPendingWorkers.compareAndSet(currentCount, updatedCount)) {//更新正在运行的线程的个数        // Start a new worker.        FLog.v(TAG, "%s: starting worker %d of %d", mName, updatedCount, mMaxConcurrency);        mExecutor.execute(mTaskRunner);//开始运行从队列里取出来的线程并实时更新正在运行的线程的个数        break;      }      // else: compareAndSet failed due to race; snapshot the new count and try again      FLog.v(TAG, "%s: race in startWorkerIfNeeded; retrying", mName);      currentCount = mPendingWorkers.get();//获取当前正在运行的线程的个数    }  }

接下来我们看看mTaskRunner都干了什么？

  private class Worker implements Runnable {    @Override    public void run() {      try {        Runnable runnable = mWorkQueue.poll();//从队列中获取并删除第一个元素        if (runnable != null) {          runnable.run();//获取到的线程如果不为空则执行该线程        } else {          // It is possible that a new worker was started before a previously started worker          // de-queued its work item.          FLog.v(TAG, "%s: Worker has nothing to run", mName);        }      } finally {        int workers = mPendingWorkers.decrementAndGet();//当线程运行完了之后，正在运行的线程数减一        if (!mWorkQueue.isEmpty()) {          startWorkerIfNeeded();//如果队列不为空的话，判断是否需要从队列中获取线程并执行        } else {          FLog.v(TAG, "%s: worker finished; %d workers left", mName, workers);        }      }    }  }

接下来我们以流程图来看下整个逻辑的处理过程

3.DefaultSerialExecutorService.java

DefaultSerialExecutorService默认实现了SerialExecutorService接口并继承ConstrainedExecutorService，他将会按先进先出的顺序执行队列中的任务