线程池的学习和使用

什么是线程池
创建线程池
使用线程池

什么是线程池

线程池的作用是初始化一些线程，当有任务的时候，就从中启动一个来执行相关任务，执行完后，线程资源重新回收到线程池中，达到复用的效果，从而减少资源的开销

创建线程池

在JDK中，Executors类已经帮我们封装了创建线程池的方法。

    Executors.newFixedThreadPool();
    Executors.newCachedThreadPool();
    Executors.newScheduledThreadPool();

但是点进去看的话，

public static ExecutorService newFixedThreadPool(int nThreads) {
    return new ThreadPoolExecutor(nThreads, nThreads,
                                  0L, TimeUnit.MILLISECONDS,
                                  new LinkedBlockingQueue<Runnable>());
}

它的内部实现还是基于ThreadPoolExecutor来实现的。通过阿里代码规范插件扫描会提示我们用ThreadPoolExecutor去实现线程池。通过查看ThreadPoolExecutor的构造方法

public ThreadPoolExecutor(int corePoolSize,
                              int maximumPoolSize,
                              long keepAliveTime,
                              TimeUnit unit,
                              BlockingQueue<Runnable> workQueue,
                              ThreadFactory threadFactory,
                              RejectedExecutionHandler handler) {
                                    ...
                                  do something
                                  ...
                              }

我觉得有以下几方面的原因。

可以灵活设置keepAliveTime(当线程池中线程数大于corePoolSize的数m, 为这m个线程设置的最长等待时间 )，节约系统资源。
workQueue：线程等待队列，在Executors中默认的是LinkedBlockingQueue。可以理解是一种无界的数组，当有不断有线程来的时候，可能会撑爆机器内存。
可以设线程工厂，里面添加自己想要的一些元素，只需要实现JDK的ThreadFactory类。
按照自己的业务设置合适的拒绝策略。策略有以下几种
1. AbortPolicy：直接抛出拒绝异常（继承自RuntimeException），会中断调用者的处理过程，所以除非有明确需求，一般不推荐
2. DiscardPolicy：默默丢弃无法加载的任务。
3. DiscardOldestPolicy：丢弃队列中最老的，然后再次尝试提交新任务。
4. CallerRunsPolicy：在调用者线程中（也就是说谁把 r 这个任务甩来的），运行当前被丢弃的任务。只会用调用者所在线程来运行任务，也就是说任务不会进入线程池。如果线程池已经被关闭，则直接丢弃该任务。

使用线程池

声明`ThreadFactory`

public class NacosSyncThreadFactory implements ThreadFactory {
    private final AtomicInteger threadNum = new AtomicInteger(1);
    private String threadPrefix = null;
    private ThreadGroup threadGroup;

    public NacosSyncThreadFactory(String prefix) {
        this.threadPrefix = "thread" + "-" + prefix + "-" ;
        threadGroup = Thread.currentThread().getThreadGroup();

    }

    public NacosSyncThreadFactory() {
        this("pool");
    }

    @Override
    public Thread newThread(Runnable r) {
        String name = threadPrefix + threadNum.incrementAndGet();
        Thread thread = new Thread(threadGroup, r, name);
        return thread;
    }
}

创建线程池类

public class MyThreadPool {
    private ThreadFactory threadFactory;
    private int threadNum;
    private BlockingQueue blockingQueue;
    private RejectedExecutionHandler handler;

    public MyThreadPool(ThreadFactory threadFactory, int threadNum,
                        BlockingQueue blockingQueue,
                        RejectedExecutionHandler handler ) {
        this.threadFactory = threadFactory;
        this.threadNum = threadNum;
        this.blockingQueue = blockingQueue;
        this.handler = handler;
    }

    public MyThreadPool() {
        this(Executors.defaultThreadFactory(), 10,
                new ArrayBlockingQueue(10), new ThreadPoolExecutor.AbortPolicy());
    }

    public ThreadPoolExecutor initThreadPool(ThreadFactory threadFactory, int threadNum, BlockingQueue blockingQueue, RejectedExecutionHandler handler) {
        if (handler == null) {
            handler = new ThreadPoolExecutor.AbortPolicy();
        }


        return new ThreadPoolExecutor(1, threadNum, 5, TimeUnit.SECONDS, blockingQueue, threadFactory, handler);
    }

}

调用线程池

初始化线程池类

  MyThreadPool myThreadPool = new MyThreadPool();

    threadPoolExecutor = myThreadPool.initThreadPool(
            new NacosSyncThreadFactory("nacos-sync"),
            threadNum,
            new ArrayBlockingQueue(10),
            new ThreadPoolExecutor.DiscardPolicy()

    );

}

创建Callable(FutureTask)

/**
 * 分页获取task信息
 * @return
 */
private List<Task> getTask(int pageNum) {
   IPage<Task> page = new Page(pageNum, 25);
   IPage<Task> taskIPage = this.taskService.page(page);
   if (null == taskIPage || CollectionUtils.isEmpty(taskIPage.getRecords())) {
       return null;
   }

   return taskIPage.getRecords();

}
// 执行任务
private FutureTask<String> assembleTaskFuture(Task task) {
    FutureTask<String> futureTask = new FutureTask(() -> {

        // 执行任务
        this.doSyncWork(task);
        return "success";
    });

    return futureTask;
}

执行任务（FutureTask）

public void zkSync() {
    // 获取数据总数，得到线程数
    int count = this.taskService.count();
    int pageSize = 25;
    int num = count / pageSize;
    int pageTotal = count % pageSize == 0 ? num : num + 1;
    log.info("========总记录数：{}=====总页数：{}", count, pageTotal);

    for (int i = 1; i <=  pageTotal; i++) {
        List<Task> taskList = this.getTask(i);
        if (CollectionUtils.isEmpty(taskList)) {
            break;
        }
        List<Integer> collect = taskList.stream().map(task -> task.getId()).collect(Collectors.toList());
        taskList.forEach(task -> {
            FutureTask<String> futureTask = this.assembleTaskFuture(task);
            threadPoolExecutor.execute(futureTask);
        });

    }

    threadPoolExecutor.shutdown();

}

来源：https://www.cnblogs.com/KevinStark/p/12430091.html

标签

线程池

handler