转载

集合在耗时程序中流遍历、迭代遍历的效率测量对比

测试环境

操作系统：WIN10

逻辑处理器：4核

运行内存：16G

测试程序：

CloudLogVo.java

import lombok.Data;

/**
 * @className: CloudLogVo
 * @description: 日志信息Vo
 * @author: niaonao
 **/
@Data
public class CloudLogVo {

    private String level;
    private String location;
    private String message;
    private String time;

    public CloudLogVo(String level, String location, String message, String time) {
        this.level = level;
        this.location = location;
        this.message = message;
        this.time = time;
    }
}

TraversalTimeConsume.java

import java.util.ArrayList;
import java.util.Collections;
import java.util.List;

/**
 * @className: TraversalStatisticsConsumeTime
 * @description: 耗时操作的测试类
 * 此处使用 System.out.print() 模拟耗时操作；为什么耗时？点进去看内部实现，write()方法就很重量级
 * @author: niaonao
 **/
public class TraversalTimeConsume {

    // 遍历测试集合
    private static List<CloudLogVo> voList = null;

    /**
     * @description: 内部类：for-i 线程类
     */
    class ForIThread extends Thread {
        @Override
        public void run() {
            long start = System.currentTimeMillis();
            for (int i = 0; i < voList.size(); i++) {
                voList.get(i).getLevel();
                System.out.print("");
            }
            long end = System.currentTimeMillis();
            System.out.println("         for-i: " + (end - start));
        }
    }

    /**
     * @description: 内部类：foreach 线程类
     */
    class ForeachThread extends Thread {
        @Override
        public void run() {
            long start = System.currentTimeMillis();
            for (CloudLogVo item : voList) {
                item.getLevel();
                System.out.print("");
            }
            long end = System.currentTimeMillis();
            System.out.println("       foreach: " + (end - start));
        }
    }

    /**
     * @description: 内部类：stream 线程类
     */
    class StreamThread extends Thread {
        @Override
        public void run() {
            long start = System.currentTimeMillis();
            voList.stream().forEach(item -> {
                item.getLevel();
                System.out.print("");
            });
            long end = System.currentTimeMillis();
            System.out.println("        stream: " + (end - start));
        }
    }

    /**
     * @description: 内部类：parallelStream 线程类
     */
    class ParallelStreamThread extends Thread {
        @Override
        public void run() {
            long start = System.currentTimeMillis();
            voList.parallelStream().forEach(item -> {
                item.getLevel();
                System.out.print("");
            });
            long end = System.currentTimeMillis();
            System.out.println("parallelStream: " + (end - start));
        }
    }

    /**
     * @description: 初始化集合
     */
    private static void initList(int quantity) {
        List<CloudLogVo> cloudLogVoList = new ArrayList<>(quantity);
        for (int i = 0; i < quantity; i++) {
            String value = String.valueOf(i);
            cloudLogVoList.add(new CloudLogVo(value, value, value, value));
        }
        voList = Collections.synchronizedList(cloudLogVoList);
    }

    public static void main(String[] args) throws InterruptedException {
        // 遍历量级
        int quantity = 1 * 1000 * 1000;
        initList(quantity);
        // 统计遍历耗时
        TraversalTimeConsume.ForIThread forIThread = new TraversalTimeConsume().new ForIThread();
        TraversalTimeConsume.ForeachThread foreachThread = new TraversalTimeConsume().new ForeachThread();
        TraversalTimeConsume.StreamThread streamThread = new TraversalTimeConsume().new StreamThread();
        TraversalTimeConsume.ParallelStreamThread parallelStreamThread = new TraversalTimeConsume().new ParallelStreamThread();

        new Thread().sleep(2000);
        forIThread.start();
        foreachThread.start();
        streamThread.start();
        parallelStreamThread.start();
    }
}

测试量级

测试数据量级如下

data: ['10万','30万','40万','50万','60万','70万','100万','150万','200万','300万']

测试结果

纵坐标为耗时/毫秒，横坐标为集合量级。数据量级越大，统计耗时数据误差越小。

集合在耗时程序中流遍历、迭代遍历的效率测量对比

遍历结构体存在耗时程序时，综合来看:

增强for循环表现较好，foreach效率整体较高
在 65 万级数据量左右，parallelStream 效率会超过 foreach，随着数据量级增加，差距增大
stream 与 parallelStream 在 50 万数量级别之后，parallelStream 遍历效率更高，并且随着数据量增大，差距越大