ThreadLocal是线程本地变量,可以应用在分布式系统追踪,事务管理方面,这里先提出几个较难的问题
- ThreadLocal的内部大致实现原理?
- ThreadLocal的Entry的key为何设计成弱引用?
- ThreadLocal的hash碰撞是如何处理的?
- ThreadLocal如何处理主线程传值到子线程?
- 如何让子线程跟随主线程tl值变化而变化?
- 线程池中使用threadLocal如何保持不变?
- 阿里TransmittableThreadLocal?
ThreadLocal的内部大致实现原理?
ThreadLocal内部类有一个内部类ThreadLocalMap, ThreadLocalMap有一个内部类Entry<tl,value>,Entry的key是弱引用。
Thread类有个属性threadLocals,类型是ThreadLocalMap。
当你执行threadLocal.get的时候,他会先获得当前线程,从线程里拿到属性threadLocals,然后从map中获取值
ThreadLocal的Entry的key为何设计成弱引用?
实线为强引用,虚线为弱引用
设想如果是强引用会发生什么?
ThreadLocal tl = new ThreadLocal();
当你tl=null时,1号线断开,但是由于ThreadLocal还被Entry的key强引用着(2号线),不能被回收,就造成了内存泄漏。
key使用弱引用就没有内存泄漏了吗?
当然不是,ThreadLocal确实可以被回收了,但是Entry的value还被threadLocals强引用着(3号线),value还是内存泄露了,
也正是因为这样,ThreadLocal的get,set,remove方法全部存在一个判断如下:
if(null==key){
//手动释放value
value = null;
}
ThreadLocal的hash碰撞是如何处理的?
我们知道hashmap处理hash碰撞的方式是拉链法,或者链地址法,具体做法就是数组+链表/红黑树。
那么ThreadLocal的hash碰撞其实是使用闭散列表,或者开放定址法,具体做法就是如果hash后mod数组长度如果发现下标已经有元素了,那么取数组的下一个下标,直到取到为止。
ThreadLocal如何处理主线程传值到子线程?
使用InheritableThreadLocal
//主线程设置了值
InheritableThreadLocal<Integer> tl = new InheritableThreadLocal<>();
tl.set(1);
//子线程可以拿得到
new Thread(()->{
System.out.println("子线程:"+tl.get());
}).start();
实现原理就是在new Thread里的init会把主线程的inheritableThreadLocals属性深拷贝到子线程的inheritableThreadLocals属性
如何让子线程跟随主线程tl值变化而变化?
阅读如下代码:
@Test
public void testInheritableThreadLocal1() throws InterruptedException {
int i=1;
//主线程设置了值
InheritableThreadLocal<Integer> tl = new InheritableThreadLocal<>();
tl.set(i);
new Thread(()->{
while(true){
//但子线程始终是初始值,因为只有init才会深拷贝父线程的inheritableThreadLocals属性
System.out.println("子线程:"+tl.get());//始终拿到的是1
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}).start();
//主线程不停修改
while(true){
Thread.sleep(1000);
i++;
tl.set(i);
System.out.println("主线程:"+tl.get());
}
}
通过如上代码可知:子线程只会在new Thread的init方法深拷贝inheritableThreadLocals数据,所以之后不管主线程如何修改tl的值,子线程都不会改变。
解决方案:使用线程同步,当父线程值发生改变时通知子线程修改tl的值。
代码如下:
方案一:
int i = 1;
@Test
public void testInheritableThreadLocal2() throws InterruptedException {
Object obj = new Object();
//主线程设置了值
InheritableThreadLocal<Integer> tl = new InheritableThreadLocal<>();
tl.set(i);
//new Thread里的init会把主线程的inheritableThreadLocals属性深拷贝到子线程的inheritableThreadLocals属性
//但子线程可以拿得到
new Thread(()->{
synchronized (obj){
while(true){
tl.set(i);
System.out.println("子线程:"+tl.get());
obj.notify();
try {
obj.wait();
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}).start();
//主线程不停修改
synchronized (obj){
while(true){
i++;
obj.notify();//通知子线程修改线程变量
obj.wait();//释放锁
tl.set(i);//主线程修改线程变量
System.out.println("主线程:"+tl.get());
Thread.sleep(1000);
}
}
}
方案二:(失败,但值得借鉴)使用回调函数callback,当主线程的tl值发生改变,调用子线程的callback接口把tl值传过去
失败原因在注释里有详细说明
@Test
public void testInheritableThreadLocal3() throws InterruptedException {
//主线程设置了值
InheritableThreadLocal<Integer> tl = new InheritableThreadLocal<>();
tl.set(i);
final Callback[] callback = new Callback[1];
new Thread(()->{
//这样还是不行,子线程始终还是最初的值
//因为这行代码是主线程调用的,所以修改的还是主线程的tl值
callback[0] = i -> { tl.set(i); };
while(true){
System.out.println("子线程:"+tl.get());
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}).start();
Thread.sleep(1000);
//主线程不停修改
while(true){
i++;
//调用子线程回调函数callback
callback[0].update(i);
tl.set(i);//主线程修改线程变量
System.out.println("主线程:"+tl.get());
Thread.sleep(1000);
}
}
interface Callback{
void update(int i);
}
线程池中使用threadLocal如何保持不变?
先看错误例子:
public class ThreadLocalVariableHolder2 {
private static ThreadLocal<Integer> variableHolder = new InheritableThreadLocal<Integer>(){
@Override
protected Integer initialValue() {
return 0;
}
};
public static void setValue(int val) {
variableHolder.set(val);
}
public static int getValue() {
return variableHolder.get();
}
public static void remove() {
variableHolder.remove();
}
public static void increment() {
variableHolder.set(variableHolder.get() + 1);
}
public static void main(String[] args) {
ExecutorService executor = Executors.newFixedThreadPool(2);
//线程池会复用线程,使用完线程后一定要清楚threadLocal,不然去执行新任务的时候会使用上一次的threadLocal产生错误结果
for (int i = 0; i < 5; i++) {
executor.execute(() -> {
long threadId = Thread.currentThread().getId();
int before = getValue();
increment();
int after = getValue();
System.out.println("threadId: " + threadId + ", before: " + before + ", after: " + after);
});
}
executor.shutdown();
}
}
输出结果如下:
threadId: 13, before: 0, after: 1
threadId: 12, before: 0, after: 1
threadId: 13, before: 1, after: 2
threadId: 13, before: 2, after: 3
threadId: 12, before: 1, after: 2
让我们关注threadId=12的输出
threadId: 12, before: 0, after: 1
threadId: 12, before: 1, after: 2
这就不符合我们想要的结果了,我们想要的结果是
threadId: 12, before: 0, after: 1
threadId: 12, before: 0, after: 1
即使是同一个线程,但是执行的任务却是不同的,前面的任务肯定不能影响后面的任务啊.
解决方案
在每个任务执行结束后执行remove,把当前线程的tl删掉就不会影响到下一个相同线程了,如下:
try {
long threadId = Thread.currentThread().getId();
int before = getValue();
increment();
int after = getValue();
System.out.println("threadId: " + threadId + ", before: " + before + ", after: " + after);
**加粗样式** }finally {
remove();
}
上面的例子有initialValue,如果我们没有复写该方法呢?
如果还是remove的话,下一次相同线程进来的时候就会空指针,getValue产生的空指针。
解决方案:
之所以这样是因为getValue设计成如果拿到的为null,就会从initValue方法拿。
阿里TransmittableThreadLocal
<dependency>
<groupId>com.alibaba</groupId>
<artifactId>transmittable-thread-local</artifactId>
<version>2.12.1</version>
</dependency>
public class ThreadLocalVariableHolder1 {
private static ThreadLocal<Integer> variableHolder = new TransmittableThreadLocal<Integer>();
public static void setValue(int val) {
variableHolder.set(val);
}
public static int getValue() {
return variableHolder.get();
}
public static void remove() {
variableHolder.remove();
}
public static void increment() {
variableHolder.set(variableHolder.get() + 1);
}
public static void main(String[] args) {
ExecutorService executor = Executors.newFixedThreadPool(2);
variableHolder.set(1);
//线程池会复用线程,使用完线程后一定要清楚threadLocal,不然去执行新任务的时候会使用上一次的threadLocal产生错误结果
for (int i = 0; i < 5; i++) {
executor.execute(TtlRunnable.get(() -> {
long threadId = Thread.currentThread().getId();
int before = getValue();
increment();
int after = getValue();
System.out.println("threadId: " + threadId + ", before: " + before + ", after: " + after);
}));
}
executor.shutdown();
}
}
输出结果:
threadId: 14, before: 1, after: 2
threadId: 13, before: 1, after: 2
threadId: 14, before: 1, after: 2
threadId: 13, before: 1, after: 2
threadId: 14, before: 1, after: 2
结果正确,我们连finally都没写,他是如何做到的?
TtlRunnable.get(task), 这行代码很关键,最终代码如下:
基于这个思想,我改造了一下我的代码:
public class ThreadLocalVariableHolder3 {
private static ThreadLocal<Integer> variableHolder = new InheritableThreadLocal<Integer>();
public static void setValue(int val) {
variableHolder.set(val);
}
public static int getValue() {
return variableHolder.get();
}
public static void remove() {
variableHolder.remove();
}
public static void increment() {
variableHolder.set(variableHolder.get() + 1);
}
public static void main(String[] args) {
ExecutorService executor = Executors.newFixedThreadPool(2);
variableHolder.set(1);
//线程池会复用线程,使用完线程后一定要清楚threadLocal,不然去执行新任务的时候会使用上一次的threadLocal产生错误结果
for (int i = 0; i < 5; i++) {
executor.execute(
TtlRunnable.get(
() -> {
int before = getValue();
long threadId = Thread.currentThread().getId();
increment();
int after = getValue();
System.out.println("threadId: " + threadId + ", before: " + before + ", after: " + after);
},variableHolder)
//我也不需要在这里还原回去
);
}
executor.shutdown();
}
}
class TtlRunnable implements Runnable{
private Runnable task;
private ThreadLocal<Integer> tl;
private TtlRunnable(Runnable task,ThreadLocal<Integer> tl) {
this.task = task;
this.tl = tl;
}
public static Runnable get(Runnable task,ThreadLocal<Integer> tl){
return new TtlRunnable(task,tl);
}
@Override
public void run() {
//备份线程执行之前threadLocal的值
int back = tl.get();
try {
this.task.run();
} finally {
//线程执行完后给恢复回去
tl.set(back);
}
}
}