CopyOnWriteArrayList简单使用及性能测试

CopyOnWriteArrayList用于替代同步List,在某些情况下有更好的并发性能，并且在迭代时不需要加锁或复制。在每次修改的时候都会创建并重新发布一个新的容器副本。

下面是源码：

    //添加元素，会加锁，创建一个新的对象，原引用指向新对象
    public boolean add(E e) {
        final ReentrantLock lock = this.lock;
        lock.lock();
        try {
            Object[] elements = getArray();
            int len = elements.length;
            Object[] newElements = Arrays.copyOf(elements, len + 1);
            newElements[len] = e;
            setArray(newElements);
            return true;
        } finally {
            lock.unlock();
        }
    }

    //获取元素，不加锁
    private E get(Object[] a, int index) {
        return (E) a[index];
    }

    public E get(int index) {
        return get(getArray(), index);
    }

优点：多个线程可以同时对这个容器进行迭代，而不会相互干扰或者与修改容器的线程互相干扰，即不会抛出ConcurrentModificationException，并且返回的元素与迭代器创建时的元素完全一致，而不必考虑之后修改带来的影响。

缺点：

1.在添加元素时会创建副本，如果副本占用内较大，会增加gc，严重可能会影响相应时间。

2.数据无法保证实时性，可能读到的还是老的数据。

 

下面贴上测试代码：

import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.Random;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

/**
 * @author wanting
 *
 */
public class ListPerformanceTest {
	// 线程数
	private static int threadNum;
	// 初始容量
	private static int initCapacity;
	// 循环次数
	private static int loop;

	private static CountDownLatch startLatch;
	private static CountDownLatch endLatch;

	private static List<String> cowList;
	// 耗时
	private volatile static double cowListTotalCost;

	private static List<String> syncList;
	// 耗时
	private volatile static double syncListTotalCost;

	public ListPerformanceTest(int threadNum, int initCapacity, int loop) {
		ListPerformanceTest.initCapacity = initCapacity;
		ListPerformanceTest.threadNum = threadNum;
		ListPerformanceTest.loop = loop;
		startLatch = new CountDownLatch(threadNum);
		endLatch = new CountDownLatch(threadNum);
		cowList = new CopyOnWriteArrayList<>(new ArrayList<>(initCapacity));
		syncList = Collections.synchronizedList(new ArrayList<>(initCapacity));
		initList();
	}

	public void initList() {
		for (int i = 0; i < initCapacity; i++) {
			cowList.add(i + "");
			syncList.add(i + "");
		}
	}

	public static void testPerformance(List<String> list) {
		long t1 = System.nanoTime();
		Random random = new Random();
		for (int i = 0; i < loop; i++) {
			int index = random.nextInt(initCapacity);
			list.get(index);
		}
		long t2 = System.nanoTime();
		double costTime = (t2 - t1) / 1000000d;
		if (list instanceof CopyOnWriteArrayList) {
			cowListTotalCost = cowListTotalCost + costTime;
		} else {
			syncListTotalCost = syncListTotalCost + costTime;
		}
		// System.out.println(list.getClass() + "耗时：" + costTime + "秒");
	}

	public static void addEle(String element) {
		cowList.add(element);
		syncList.add(element);
	}

	public static void main(String[] args) {
		ListPerformanceTest listPerformanceTest = new ListPerformanceTest(500, 100000, 10000);
		ExecutorService pool = Executors.newFixedThreadPool(threadNum);
		for (int i = 1; i <= threadNum; i++) {
			final int flag = i;
			pool.submit(new Runnable() {

				@Override
				public void run() {
					try {
						startLatch.await();
						 //if (flag % 2 == 0) {
						 //addEle(flag + "");
						// }
						testPerformance(cowList);
						testPerformance(syncList);
					} catch (Exception e) {
						e.printStackTrace();
					}
					endLatch.countDown();
				}
			});
			startLatch.countDown();
		}
		try {
			endLatch.await();
		} catch (InterruptedException e) {
			e.printStackTrace();
		}
		System.out.println("cowz总耗时：" + cowListTotalCost + "微秒，平均耗时：" + (cowListTotalCost / (threadNum * 1.0) + "微秒"));
		System.out
				.println("sync总耗时：" + syncListTotalCost + "微秒，平均耗时：" + (syncListTotalCost / (threadNum * 1.0) + "微秒"));
		pool.shutdown();
	}

总结：适合用于迭代操作远远多于修改操作时。随着并发量增加或者迭代次数增加，CopyOnWriteArrayList性能要优于同步List。