java

关注公众号 jb51net

关闭
首页 > 软件编程 > java > 阿里巴巴TransmittableThreadLocal使用

关于阿里巴巴TransmittableThreadLocal使用解读

作者：林犀居士

文章主要介绍了三种ThreadLocal的使用：ThreadLocal、InheritableThreadLocal和TransmittableThreadLocal,ThreadLocal和InheritableThreadLocal在单线程和部分情况下可以正常工作,但TransmittableThreadLocal在处理线程池时表现更佳

前言

ThreadLocal在上下文的数据传输上非常的方便和简洁。

工业实践中，比较常用的有三个，ThreadLocal、InheritableThreadLocal、TransmittableThreadLocal，那么他们三个之间有什么区别呢？

常见的三种ThreadLocal比较

ThreadLocalInheritableThreadLocalTransmittableThreadLocal
来源jdkjdk阿里开源
单线程数据传输支持支持支持
new线程数据传输不支持支持支持
线程池数据传输不支持部分支持【简单场景】支持

针对线程池的数据传输，InheritableThreadLocal仅仅能在一些简单场景下做到，下面就用一个案例来说明

package com.tml.mouseDemo.core.threadLocalDemo;

import java.util.concurrent.ThreadFactory;
import java.util.concurrent.atomic.AtomicInteger;

public class SimpleThreadFactory implements ThreadFactory {

    private static AtomicInteger atomicInteger = new AtomicInteger(1);
    @Override
    public Thread newThread(Runnable r) {

        Thread t = new Thread(r);
        t.setName("tml-"+atomicInteger.getAndIncrement());
        return t;
    }
}
package com.tml.mouseDemo.core.threadLocalDemo;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class ThreadLocalDemo1 {

    private static ThreadLocal<String> threadLocal = new InheritableThreadLocal<>();

    private static ExecutorService service = Executors.newFixedThreadPool(2, new SimpleThreadFactory());


    public static void main(String[] args) throws InterruptedException {


        threadLocal.set("hello main");
        for (int i = 0; i < 2; i++) {
            service.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }

        //修改threadLocal中的值
        threadLocal.set("hello world");
        Thread.sleep(2000);
        for (int i = 0; i < 2; i++) {

            service.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }


        ThreadUtils.printLog("get data " + threadLocal.get());

        service.shutdown();

    }
}

2025-01-10 19:41:50 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 19:41:50 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 19:41:52 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 19:41:52 | INFO | main | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 19:41:52 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main

从运行结果来看，前面的两次循环提交的任务，在子线程中确实是能正常的获取主线程设置的变量，即hello main，但是紧接着，我修改了主线程上绑定的变量为hello world，然后继续循环两次提交两个任务，这个时候子线程中获取的线程变量依然是hello main，这明显是与是期望不一致的。

从这个层面来讲，InheritableThreadLocal确实在线程池的层面支持不够友好，可以说仅支持部分简单场景。

根本原因就死线程池的池化机制，从上面的运行日志上也可以看出，提交了4个任务执行的线程依然是两个。线程池中的线程是复用的，InheritableThreadLocal是在创建子线程的时候，会将主线程上绑定的数据拷贝过来，如果我不创建新的线程，但是主线程上绑定的数据改变了呢？那我依然还是读取到之前拷贝的数据。

这个就是InheritableThreadLocal的短板。针对这个问题，阿里开源的TransmittableThreadLocal就能顺利丝滑的解决这个问题。

TransmittableThreadLocal实践

maven依赖

<dependency>
     <groupId>com.alibaba</groupId>
    <artifactId>transmittable-thread-local</artifactId>
     <version>2.12.4</version>
</dependency>

装饰Runnable任务

package com.tml.mouseDemo.core.threadLocalDemo;

import com.alibaba.ttl.TransmittableThreadLocal;
import com.alibaba.ttl.TtlRunnable;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class ThreadLocalDemo1 {

    private static ThreadLocal<String> threadLocal = new TransmittableThreadLocal<>();

    private static ExecutorService service = Executors.newFixedThreadPool(2, new SimpleThreadFactory());


    public static void main(String[] args) throws InterruptedException {


        threadLocal.set("hello main");
        for (int i = 0; i < 2; i++) {
            service.execute(TtlRunnable.get(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            }));
        }

        //修改threadLocal中的值
        threadLocal.set("hello world");
        Thread.sleep(2000);
        for (int i = 0; i < 2; i++) {

            service.execute(TtlRunnable.get(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            }));
        }


        ThreadUtils.printLog("get data " + threadLocal.get());

        service.shutdown();

    }
}

2025-01-10 19:57:03 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 19:57:03 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 19:57:05 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 19:57:05 | INFO | main | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 19:57:05 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world

与第一个案例的差异点在于，使用了

public static TtlRunnable get(@Nullable Runnable runnable) {
    return get(runnable, false, false);
}

来增强了了Runnable任务，执行的结果也是符合预期。

装饰线程池

package com.tml.mouseDemo.core.threadLocalDemo;

import com.alibaba.ttl.TransmittableThreadLocal;
import com.alibaba.ttl.TtlRunnable;
import com.alibaba.ttl.threadpool.TtlExecutors;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class ThreadLocalDemo1 {

    private static ThreadLocal<String> threadLocal = new TransmittableThreadLocal<>();

    private static ExecutorService service = Executors.newFixedThreadPool(2, new SimpleThreadFactory());

    private static ExecutorService ttlExecutorService = TtlExecutors.getTtlExecutorService(service);



    public static void main(String[] args) throws InterruptedException {



        threadLocal.set("hello main");
        for (int i = 0; i < 2; i++) {
            ttlExecutorService.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }

        //修改threadLocal中的值
        threadLocal.set("hello world");
        Thread.sleep(2000);
        for (int i = 0; i < 2; i++) {

            ttlExecutorService.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }


        ThreadUtils.printLog("get data " + threadLocal.get());

        service.shutdown();

    }
}

2025-01-10 20:05:05 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 20:05:05 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 20:05:07 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 20:05:07 | INFO | main | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 20:05:07 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world

与上一个案例的差异点在于，这里没有包装Runnable任务，而是包装了线程池，使用了

public static ExecutorService getTtlExecutorService(@Nullable ExecutorService executorService) {
    if (TtlAgent.isTtlAgentLoaded() || executorService == null || executorService instanceof TtlEnhanced) {
        return executorService;
    }
    return new ExecutorServiceTtlWrapper(executorService, true);
}

包装了ExecutorService，执行结果也是符合预期

使用java Agent无侵入增强线程池

package com.tml.mouseDemo.core.threadLocalDemo;

import com.alibaba.ttl.TransmittableThreadLocal;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class ThreadLocalDemo1 {

    private static ThreadLocal<String> threadLocal = new TransmittableThreadLocal<>();

    private static ExecutorService service = Executors.newFixedThreadPool(2, new SimpleThreadFactory());

    public static void main(String[] args) throws InterruptedException {



        threadLocal.set("hello main");
        for (int i = 0; i < 2; i++) {
            service.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }

        //修改threadLocal中的值
        threadLocal.set("hello world");
        Thread.sleep(2000);
        for (int i = 0; i < 2; i++) {

            service.execute(() -> {
                String s = threadLocal.get();
                ThreadUtils.printLog("get data " + s);

            });
        }


        ThreadUtils.printLog("get data " + threadLocal.get());

        service.shutdown();

    }
}

项目运行的时候，需要添加额外的jvm启动参数，如下

2025-01-10 20:11:59 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 20:11:59 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello main
2025-01-10 20:12:01 | INFO | tml-2 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 20:12:01 | INFO | main | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world
2025-01-10 20:12:01 | INFO | tml-1 | com.tml.mouseDemo.core.threadLocalDemo.ThreadUtils | get data hello world

总结

阿里巴巴的TransmittableThreadLocal是继承自InheritableThreadLocal，对他的功能进行了增强，增强的点也主要是在线程池的支持上。

通过上面的三个案例，可以看到TransmittableThreadLocal是非常灵活的，大家可以根据自己的需要，选择对应的方式来实现。

TransmittableThreadLocal的官网

以上为个人经验，希望能给大家一个参考，也希望大家多多支持脚本之家。

您可能感兴趣的文章:
阅读全文