Java基于线程池的独立任务并发执行器

Posted 2023-02-05 lyloou

目的：

对于多个独立的任务，可以以并发的方式执行任务，以提高 CPU 利用率，提高处理效率。

思路

在一个线程池中，开启指定数量的线程，每个线程从任务队列中获取任务执行。

执行的过程中，判断当前线程是否在执行任务的状态，如果没有执行任务，取一条任务执行，如果正在执行，则跳过，下轮再判断。

在所有任务执行完后，关闭线程池。

需要注意的是数据结构的选择，须选择并发类的数据结构，不然可能出现阻塞，死锁等情况。

（具体逻辑参考源码）

示例

/**
 * 并发执行器示例
 */
public class ConcurrentExecutorTest 

    /**
     * 测试
     */
    public static void main(String[] args) 
        for (int i = 0; i < 100; i++) 
            test();
        
    

    private static void test() 
        Map<String, String> paramMap = new LinkedHashMap<>();
        for (int i = 0; i < 10; i++) 
            paramMap.put("key:" + i, "value:" + i);
        

        final ConcurrentExecutor<String, String, Integer> executor = new ConcurrentExecutor<>(5, paramMap,
                (k, v) -> 
                    ThreadUtil.sleep(10);
                    System.out.println(Thread.currentThread().getName() + "-" + v);
                    final int abs = Math.abs(Objects.hash(v));
                    if (abs % 3 == 0) 
                        int i = 1 / 0;
                    
                    return abs;
                );
        executor.execute();
        System.out.println("success result: " + executor.getSuccessResultMap());
        System.out.println("error result: " + executor.getErrorResultMap());
    

测试结果

pool-1-thread-1-value:0
pool-1-thread-2-value:1
pool-1-thread-4-value:3
pool-1-thread-3-value:2
pool-1-thread-3-value:8
pool-1-thread-1-value:5
pool-1-thread-5-value:4
pool-1-thread-2-value:6
pool-1-thread-4-value:7
pool-1-thread-3-value:9
success result: key:2=231604360, key:0=231604358, key:6=231604364, key:5=231604363, key:3=231604361, key:9=231604367, key:8=231604366
error result: key:1=java.lang.ArithmeticException: / by zero, key:4=java.lang.ArithmeticException: / by zero, key:7=java.lang.ArithmeticException: / by zero

源码

import cn.hutool.core.collection.CollUtil;
import cn.hutool.core.collection.ConcurrentHashSet;
import cn.hutool.core.lang.Assert;
import cn.hutool.core.thread.ThreadUtil;
import cn.hutool.core.util.BooleanUtil;

import java.util.*;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentLinkedQueue;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.function.BiFunction;

/**
 * 并发执行器
 * <p>
 * 适用场景：每个任务是独立的，不耦合的
 *
 * @author lilou
 * @since 2022/6/9 9:05
 */
public class ConcurrentExecutor<K, V, R> 
    private final Map<K, V> paramMap;
    private final Map<K, R> successResultMap;
    private final Map<K, Throwable> errorResultMap;
    private final Set<K> runningKeySet;
    private final Queue<K> candidateKeyQueue;
    private int maxThreadNum;
    private final ExecutorService executorService;
    private final BiFunction<K, V, R> consumer;
    private final Map<Integer, Boolean> currentIndexThreadRunningStatusMap;

    public ConcurrentExecutor(int maxThreadNum, Map<K, V> paramMap, BiFunction<K, V, R> consumer) 
        Assert.notNull(paramMap, "paramMap不可为空");
        Assert.isTrue(maxThreadNum > 0, "maxThreadNum不可小于1");

        final int paramSize = paramMap.size();
        // 如果需要处理的数据小于 maxThreadNum 时，用这个小数
        this.maxThreadNum = Math.min(maxThreadNum, paramSize);
        // 线程池数量不能比1小
        this.maxThreadNum = Math.max(this.maxThreadNum, 1);
        // tips: 如果错误地使用 this.paramMap = paramMap; 多测试几遍会发现，偶尔会陷入了阻塞
        this.paramMap = Collections.synchronizedMap(paramMap);
        this.candidateKeyQueue = new ConcurrentLinkedQueue<>(paramMap.keySet());
        this.runningKeySet = new ConcurrentHashSet<>(paramSize);
        this.consumer = consumer;

        this.executorService = Executors.newWorkStealingPool(this.maxThreadNum);
        this.currentIndexThreadRunningStatusMap = new ConcurrentHashMap<>(this.maxThreadNum);
        this.successResultMap = Collections.synchronizedMap(new LinkedHashMap<>(this.paramMap.size()));
        this.errorResultMap = Collections.synchronizedMap(new LinkedHashMap<>());
    


    public void execute() 
        while (CollUtil.isNotEmpty(paramMap)) 

            // 最多同时有 maxRunningThreadNumber 同时消费 taskMap 中的数据
            for (int i = 0; i < this.maxThreadNum; i++) 
                int currentIndex = i;

                // 没执行完，不要执行
                final Boolean isRunning = currentIndexThreadRunningStatusMap.getOrDefault(currentIndex, false);
                if (BooleanUtil.isTrue(isRunning)) 
                    continue;
                

                // 每个线程只处理和自己相关的
                final K candidateKey = pickCandidateKey();
                // 当前没有对应key的任务
                if (Objects.isNull(candidateKey)) 
                    continue;
                

                executorService.submit(() -> 
                    try 
                        currentIndexThreadRunningStatusMap.put(currentIndex, true);
                        final V data = paramMap.get(candidateKey);
                        final R result = consumer.apply(candidateKey, data);
                        successResultMap.put(candidateKey, result);
                     catch (Exception e) 
                        errorResultMap.put(candidateKey, e);
                     finally 
                        paramMap.remove(candidateKey);
                        candidateKeyQueue.remove(candidateKey);
                        currentIndexThreadRunningStatusMap.remove(currentIndex);
                    
                );
            
        
        executorService.shutdown();
    


    private K pickCandidateKey() 
        for (K candidateKey : candidateKeyQueue) 
            if (!runningKeySet.contains(candidateKey)) 
                runningKeySet.add(candidateKey);
                return candidateKey;
            
        
        return null;
    

    public Map<K, R> getSuccessResultMap() 
        return successResultMap;
    

    public Map<K, Throwable> getErrorResultMap() 
        return errorResultMap;