zoukankan      html  css  js  c++  java
  • java高并发系列

    这是java高并发系列第16篇文章。

    本篇内容

    1. 介绍CountDownLatch及使用场景
    2. 提供几个示例介绍CountDownLatch的使用
    3. 手写一个并行处理任务的工具类

    假如有这样一个需求,当我们需要解析一个Excel里多个sheet的数据时,可以考虑使用多线程,每个线程解析一个sheet里的数据,等到所有的sheet都解析完之后,程序需要统计解析总耗时。分析一下:解析每个sheet耗时可能不一样,总耗时就是最长耗时的那个操作。

    我们能够想到的最简单的做法是使用join,代码如下:

    package com.itsoku.chat13;
    
    import java.util.concurrent.TimeUnit;
    
    /**
     * 微信公众号:javacode2018,获取年薪50万课程
     */
    public class Demo1 {
    
        public static class T extends Thread {
            //休眠时间(秒)
            int sleepSeconds;
    
            public T(String name, int sleepSeconds) {
                super(name);
                this.sleepSeconds = sleepSeconds;
            }
    
            @Override
            public void run() {
                Thread ct = Thread.currentThread();
                long startTime = System.currentTimeMillis();
                System.out.println(startTime + "," + ct.getName() + ",开始处理!");
                try {
                    //模拟耗时操作,休眠sleepSeconds秒
                    TimeUnit.SECONDS.sleep(this.sleepSeconds);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
                long endTime = System.currentTimeMillis();
                System.out.println(endTime + "," + ct.getName() + ",处理完毕,耗时:" + (endTime - startTime));
            }
        }
    
        public static void main(String[] args) throws InterruptedException {
            long starTime = System.currentTimeMillis();
            T t1 = new T("解析sheet1线程", 2);
            t1.start();
    
            T t2 = new T("解析sheet2线程", 5);
            t2.start();
    
            t1.join();
            t2.join();
            long endTime = System.currentTimeMillis();
            System.out.println("总耗时:" + (endTime - starTime));
    
        }
    }
    

    输出:

    1563767560271,解析sheet1线程,开始处理!
    1563767560272,解析sheet2线程,开始处理!
    1563767562273,解析sheet1线程,处理完毕,耗时:2002
    1563767565274,解析sheet2线程,处理完毕,耗时:5002
    总耗时:5005
    

    代码中启动了2个解析sheet的线程,第一个耗时2秒,第二个耗时5秒,最终结果中总耗时:5秒。上面的关键技术点是线程的join()方法,此方法会让当前线程等待被调用的线程完成之后才能继续。可以看一下join的源码,内部其实是在synchronized方法中调用了线程的wait方法,最后被调用的线程执行完毕之后,由jvm自动调用其notifyAll()方法,唤醒所有等待中的线程。这个notifyAll()方法是由jvm内部自动调用的,jdk源码中是看不到的,需要看jvm源码,有兴趣的同学可以去查一下。所以JDK不推荐在线程上调用wait、notify、notifyAll方法。

    而在JDK1.5之后的并发包中提供的CountDownLatch也可以实现join的这个功能。

    CountDownLatch介绍

    CountDownLatch称之为闭锁,它可以使一个或一批线程在闭锁上等待,等到其他线程执行完相应操作后,闭锁打开,这些等待的线程才可以继续执行。确切的说,闭锁在内部维护了一个倒计数器。通过该计数器的值来决定闭锁的状态,从而决定是否允许等待的线程继续执行。

    常用方法:

    public CountDownLatch(int count):构造方法,count表示计数器的值,不能小于0,否者会报异常。

    public void await() throws InterruptedException:调用await()会让当前线程等待,直到计数器为0的时候,方法才会返回,此方法会响应线程中断操作。

    public boolean await(long timeout, TimeUnit unit) throws InterruptedException:限时等待,在超时之前,计数器变为了0,方法返回true,否者直到超时,返回false,此方法会响应线程中断操作。

    public void countDown():让计数器减1

    CountDownLatch使用步骤:

    1. 创建CountDownLatch对象
    2. 调用其实例方法await(),让当前线程等待
    3. 调用countDown()方法,让计数器减1
    4. 当计数器变为0的时候,await()方法会返回

    示例1:一个简单的示例

    我们使用CountDownLatch来完成上面示例中使用join实现的功能,代码如下:

    package com.itsoku.chat13;
    
    import java.util.concurrent.CountDownLatch;
    import java.util.concurrent.TimeUnit;
    
    /**
     * 微信公众号:javacode2018,获取年薪50万课程
     */
    public class Demo2 {
    
        public static class T extends Thread {
            //休眠时间(秒)
            int sleepSeconds;
            CountDownLatch countDownLatch;
    
            public T(String name, int sleepSeconds, CountDownLatch countDownLatch) {
                super(name);
                this.sleepSeconds = sleepSeconds;
                this.countDownLatch = countDownLatch;
            }
    
            @Override
            public void run() {
                Thread ct = Thread.currentThread();
                long startTime = System.currentTimeMillis();
                System.out.println(startTime + "," + ct.getName() + ",开始处理!");
                try {
                    //模拟耗时操作,休眠sleepSeconds秒
                    TimeUnit.SECONDS.sleep(this.sleepSeconds);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                } finally {
                    countDownLatch.countDown();
                }
                long endTime = System.currentTimeMillis();
                System.out.println(endTime + "," + ct.getName() + ",处理完毕,耗时:" + (endTime - startTime));
            }
        }
    
        public static void main(String[] args) throws InterruptedException {
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "线程 start!");
            CountDownLatch countDownLatch = new CountDownLatch(2);
    
            long starTime = System.currentTimeMillis();
            T t1 = new T("解析sheet1线程", 2, countDownLatch);
            t1.start();
    
            T t2 = new T("解析sheet2线程", 5, countDownLatch);
            t2.start();
    
            countDownLatch.await();
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "线程 end!");
            long endTime = System.currentTimeMillis();
            System.out.println("总耗时:" + (endTime - starTime));
    
        }
    }
    

    输出:

    1563767580511,main线程 start!
    1563767580513,解析sheet1线程,开始处理!
    1563767580513,解析sheet2线程,开始处理!
    1563767582515,解析sheet1线程,处理完毕,耗时:2002
    1563767585515,解析sheet2线程,处理完毕,耗时:5002
    1563767585515,main线程 end!
    总耗时:5003
    

    从结果中看出,效果和join实现的效果一样,代码中创建了计数器为2的CountDownLatch,主线程中调用countDownLatch.await();会让主线程等待,t1、t2线程中模拟执行耗时操作,最终在finally中调用了countDownLatch.countDown();,此方法每调用一次,CountDownLatch内部计数器会减1,当计数器变为0的时候,主线程中的await()会返回,然后继续执行。注意:上面的countDown()这个是必须要执行的方法,所以放在finally中执行。

    示例2:等待指定的时间

    还是上面的示例,2个线程解析2个sheet,主线程等待2个sheet解析完成。主线程说,我等待2秒,你们还是无法处理完成,就不等待了,直接返回。如下代码:

    package com.itsoku.chat13;
    
    import java.util.concurrent.CountDownLatch;
    import java.util.concurrent.TimeUnit;
    
    /**
     * 微信公众号:javacode2018,获取年薪50万课程
     */
    public class Demo3 {
    
        public static class T extends Thread {
            //休眠时间(秒)
            int sleepSeconds;
            CountDownLatch countDownLatch;
    
            public T(String name, int sleepSeconds, CountDownLatch countDownLatch) {
                super(name);
                this.sleepSeconds = sleepSeconds;
                this.countDownLatch = countDownLatch;
            }
    
            @Override
            public void run() {
                Thread ct = Thread.currentThread();
                long startTime = System.currentTimeMillis();
                System.out.println(startTime + "," + ct.getName() + ",开始处理!");
                try {
                    //模拟耗时操作,休眠sleepSeconds秒
                    TimeUnit.SECONDS.sleep(this.sleepSeconds);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                } finally {
                    countDownLatch.countDown();
                }
                long endTime = System.currentTimeMillis();
                System.out.println(endTime + "," + ct.getName() + ",处理完毕,耗时:" + (endTime - startTime));
            }
        }
    
        public static void main(String[] args) throws InterruptedException {
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "线程 start!");
            CountDownLatch countDownLatch = new CountDownLatch(2);
    
            long starTime = System.currentTimeMillis();
            T t1 = new T("解析sheet1线程", 2, countDownLatch);
            t1.start();
    
            T t2 = new T("解析sheet2线程", 5, countDownLatch);
            t2.start();
    
            boolean result = countDownLatch.await(2, TimeUnit.SECONDS);
    
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "线程 end!");
            long endTime = System.currentTimeMillis();
            System.out.println("主线程耗时:" + (endTime - starTime) + ",result:" + result);
    
        }
    }
    

    输出:

    1563767637316,main线程 start!
    1563767637320,解析sheet1线程,开始处理!
    1563767637320,解析sheet2线程,开始处理!
    1563767639321,解析sheet1线程,处理完毕,耗时:2001
    1563767639322,main线程 end!
    主线程耗时:2004,result:false
    1563767642322,解析sheet2线程,处理完毕,耗时:5002
    

    从输出结果中可以看出,线程2耗时了5秒,主线程耗时了2秒,主线程中调用countDownLatch.await(2, TimeUnit.SECONDS);,表示最多等2秒,不管计数器是否为0,await方法都会返回,若等待时间内,计数器变为0了,立即返回true,否则超时后返回false。

    示例3:2个CountDown结合使用的示例

    有3个人参见跑步比赛,需要先等指令员发指令枪后才能开跑,所有人都跑完之后,指令员喊一声,大家跑完了。

    示例代码:

    package com.itsoku.chat13;
    
    import java.util.concurrent.CountDownLatch;
    import java.util.concurrent.TimeUnit;
    
    /**
     * 微信公众号:javacode2018,获取年薪50万课程
     */
    public class Demo4 {
    
        public static class T extends Thread {
            //跑步耗时(秒)
            int runCostSeconds;
            CountDownLatch commanderCd;
            CountDownLatch countDown;
    
            public T(String name, int runCostSeconds, CountDownLatch commanderCd, CountDownLatch countDown) {
                super(name);
                this.runCostSeconds = runCostSeconds;
                this.commanderCd = commanderCd;
                this.countDown = countDown;
            }
    
            @Override
            public void run() {
                //等待指令员枪响
                try {
                    commanderCd.await();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
                Thread ct = Thread.currentThread();
                long startTime = System.currentTimeMillis();
                System.out.println(startTime + "," + ct.getName() + ",开始跑!");
                try {
                    //模拟耗时操作,休眠runCostSeconds秒
                    TimeUnit.SECONDS.sleep(this.runCostSeconds);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                } finally {
                    countDown.countDown();
                }
                long endTime = System.currentTimeMillis();
                System.out.println(endTime + "," + ct.getName() + ",跑步结束,耗时:" + (endTime - startTime));
            }
        }
    
        public static void main(String[] args) throws InterruptedException {
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "线程 start!");
            CountDownLatch commanderCd = new CountDownLatch(1);
            CountDownLatch countDownLatch = new CountDownLatch(3);
    
            long starTime = System.currentTimeMillis();
            T t1 = new T("小张", 2, commanderCd, countDownLatch);
            t1.start();
    
            T t2 = new T("小李", 5, commanderCd, countDownLatch);
            t2.start();
    
            T t3 = new T("路人甲", 10, commanderCd, countDownLatch);
            t3.start();
    
            //主线程休眠5秒,模拟指令员准备发枪耗时操作
            TimeUnit.SECONDS.sleep(5);
            System.out.println(System.currentTimeMillis() + ",枪响了,大家开始跑");
            commanderCd.countDown();
    
            countDownLatch.await();
            long endTime = System.currentTimeMillis();
            System.out.println(System.currentTimeMillis() + "," + Thread.currentThread().getName() + "所有人跑完了,主线程耗时:" + (endTime - starTime));
    
        }
    }
    

    输出:

    1563767691087,main线程 start!
    1563767696092,枪响了,大家开始跑
    1563767696092,小张,开始跑!
    1563767696092,小李,开始跑!
    1563767696092,路人甲,开始跑!
    1563767698093,小张,跑步结束,耗时:2001
    1563767701093,小李,跑步结束,耗时:5001
    1563767706093,路人甲,跑步结束,耗时:10001
    1563767706093,main所有人跑完了,主线程耗时:15004
    

    代码中,t1、t2、t3启动之后,都阻塞在commanderCd.await();,主线程模拟发枪准备操作耗时5秒,然后调用commanderCd.countDown();模拟发枪操作,此方法被调用以后,阻塞在commanderCd.await();的3个线程会向下执行。主线程调用countDownLatch.await();之后进行等待,每个人跑完之后,调用countDown.countDown();通知一下countDownLatch让计数器减1,最后3个人都跑完了,主线程从countDownLatch.await();返回继续向下执行。

    手写一个并行处理任务的工具类

    package com.itsoku.chat13;
    
    import org.springframework.util.CollectionUtils;
    
    import java.util.List;
    import java.util.concurrent.CountDownLatch;
    import java.util.concurrent.ExecutorService;
    import java.util.concurrent.Executors;
    import java.util.concurrent.TimeUnit;
    import java.util.function.Consumer;
    import java.util.stream.Collectors;
    import java.util.stream.Stream;
    
    /**
     * 微信公众号:javacode2018,获取年薪50万课程
     */
    public class TaskDisposeUtils {
        //并行线程数
        public static final int POOL_SIZE;
    
        static {
            POOL_SIZE = Integer.max(Runtime.getRuntime().availableProcessors(), 5);
        }
    
        /**
         * 并行处理,并等待结束
         *
         * @param taskList 任务列表
         * @param consumer 消费者
         * @param <T>
         * @throws InterruptedException
         */
        public static <T> void dispose(List<T> taskList, Consumer<T> consumer) throws InterruptedException {
            dispose(true, POOL_SIZE, taskList, consumer);
        }
    
        /**
         * 并行处理,并等待结束
         *
         * @param moreThread 是否多线程执行
         * @param poolSize   线程池大小
         * @param taskList   任务列表
         * @param consumer   消费者
         * @param <T>
         * @throws InterruptedException
         */
        public static <T> void dispose(boolean moreThread, int poolSize, List<T> taskList, Consumer<T> consumer) throws InterruptedException {
            if (CollectionUtils.isEmpty(taskList)) {
                return;
            }
            if (moreThread && poolSize > 1) {
                poolSize = Math.min(poolSize, taskList.size());
                ExecutorService executorService = null;
                try {
                    executorService = Executors.newFixedThreadPool(poolSize);
                    CountDownLatch countDownLatch = new CountDownLatch(taskList.size());
                    for (T item : taskList) {
                        executorService.execute(() -> {
                            try {
                                consumer.accept(item);
                            } finally {
                                countDownLatch.countDown();
                            }
                        });
                    }
                    countDownLatch.await();
                } finally {
                    if (executorService != null) {
                        executorService.shutdown();
                    }
                }
            } else {
                for (T item : taskList) {
                    consumer.accept(item);
                }
            }
        }
    
        public static void main(String[] args) throws InterruptedException {
            //生成1-10的10个数字,放在list中,相当于10个任务
            List<Integer> list = Stream.iterate(1, a -> a + 1).limit(10).collect(Collectors.toList());
            //启动多线程处理list中的数据,每个任务休眠时间为list中的数值
            TaskDisposeUtils.dispose(list, item -> {
                try {
                    long startTime = System.currentTimeMillis();
                    TimeUnit.SECONDS.sleep(item);
                    long endTime = System.currentTimeMillis();
    
                    System.out.println(System.currentTimeMillis() + ",任务" + item + "执行完毕,耗时:" + (endTime - startTime));
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            });
            //上面所有任务处理完毕完毕之后,程序才能继续
            System.out.println(list + "中的任务都处理完毕!");
        }
    }
    

    运行代码输出:

    1563769828130,任务1执行完毕,耗时:1000
    1563769829130,任务2执行完毕,耗时:2000
    1563769830131,任务3执行完毕,耗时:3001
    1563769831131,任务4执行完毕,耗时:4001
    1563769832131,任务5执行完毕,耗时:5001
    1563769833130,任务6执行完毕,耗时:6000
    1563769834131,任务7执行完毕,耗时:7001
    1563769835131,任务8执行完毕,耗时:8001
    1563769837131,任务9执行完毕,耗时:9001
    1563769839131,任务10执行完毕,耗时:10001
    [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]中的任务都处理完毕!
    

    TaskDisposeUtils是一个并行处理的工具类,可以传入n个任务内部使用线程池进行处理,等待所有任务都处理完成之后,方法才会返回。比如我们发送短信,系统中有1万条短信,我们使用上面的工具,每次取100条并行发送,待100个都处理完毕之后,再取一批按照同样的逻辑发送。

    java高并发系列

    java高并发系列连载中,总计估计会有四五十篇文章,可以关注公众号:javacode2018,送年薪50万课程,获取最新文章。

  • 相关阅读:
    linux多线程下载工具mwget
    mysql性能优化学习笔记-参数介绍及优化建议
    mongodb 基础知识
    cas 单点登录出现org.jasig.cas.client.util.CommonUtils.getResponseFromServer
    mysql性能优化学习笔记-存储引擎
    mysql性能优化学习笔记
    (Code) Python implementation of phrase extraction from sentence
    论文笔记:Siamese Cascaded Region Proposal Networks for Real-Time Visual Tracking
    论文笔记:ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware
    论文笔记:DARTS: Differentiable Architecture Search
  • 原文地址:https://www.cnblogs.com/itsoku123/p/11225963.html
Copyright © 2011-2022 走看看