本文借由并发环境下使用线程不安全的SimpleDateFormat优化案例,帮助大家理解ThreadLocal.
最近整理公司项目,发现不少写的比较糟糕的地方,比如下面这个:
public class DateUtil {
private final static SimpleDateFormat sdfyhm = new SimpleDateFormat(
"yyyyMMdd");
public synchronized static Date parseymdhms(String source) {
try {
return sdfyhm.parse(source);
} catch (ParseException e) {
e.printStackTrace();
return new Date();
}
}
}
首先分析下:
该处的函数parseymdhms()使用了synchronized修饰,意味着该操作是线程不安全的,所以需要同步,线程不安全也只能是SimpleDateFormat的parse()方法,查看下源码,在SimpleDateFormat里面有一个全局变量
protected Calendar calendar;
Date parse() {
calendar.clear();
... // 执行一些操作, 设置 calendar 的日期什么的
calendar.getTime(); // 获取calendar的时间
}
该clear()操作会造成线程不安全.
此外使用synchronized 关键字对性能有很大影响,尤其是多线程的时候,每一次调用parseymdhms方法都会进行同步判断,并且同步本身开销就很大,因此这是不合理的解决方案.
改进方法
线程不安全是源于多线程使用了共享变量造成,所以这里使用ThreadLocal
public class DateUtil {
private static Map> sdfMap = new HashMap>();
private static Logger logger = LoggerFactory.getLogger(DateUtil.class);
public final static String MDHMSS = "MMddHHmmssSSS";
public final static String YMDHMS = "yyyyMMddHHmmss";
public final static String YMDHMS_ = "yyyy-MM-dd HH:mm:ss";
public final static String YMD = "yyyyMMdd";
public final static String YMD_ = "yyyy-MM-dd";
public final static String HMS = "HHmmss";
private static SimpleDateFormat getSdf(final String pattern){
ThreadLocal sdfThread = sdfMap.get(pattern);
if (sdfThread == null){
//双重检验,防止sdfMap被多次put进去值,和双重锁单例原因是一样的
synchronized (DateUtil.class){
sdfThread = sdfMap.get(pattern);
if (sdfThread == null){
logger.debug("put new sdf of pattern " + pattern + " to map");
sdfThread = new ThreadLocal(){
@Override
protected SimpleDateFormat initialValue() {
logger.debug("thread: " + Thread.currentThread() + " init pattern: " + pattern);
return new SimpleDateFormat(pattern);
}
};
sdfMap.put(pattern,sdfThread);
}
}
}
return sdfThread.get();
}
public static Date parseDate(String date,String pattern){
if(date == null) {
throw new IllegalArgumentException("The date must not be null");
}
try {
return getSdf(pattern).parse(date);
} catch (ParseException e) {
e.printStackTrace();
logger.error("解析的格式不支持:"+pattern);
}
return null;
}
public static String formatDate(Date date,String pattern){
if (date == null){
throw new IllegalArgumentException("The date must not be null");
}else {
return getSdf(pattern).format(date);
}
}
}
测试
在主线程中执行一个,另外两个在子线程执行,使用的都是同一个pattern
public static void main(String[] args) {
DateUtil.formatDate(new Date(),MDHMSS);
new Thread(()->{
DateUtil.formatDate(new Date(),MDHMSS);
}).start();
new Thread(()->{
DateUtil.formatDate(new Date(),MDHMSS);
}).start();
}
日志分析
put new sdf of pattern MMddHHmmssSSS to map thread: Thread[main,5,main] init pattern: MMddHHmmssSSS thread: Thread[Thread-0,5,main] init pattern: MMddHHmmssSSS thread: Thread[Thread-1,5,main] init pattern: MMddHHmmssSSS
分析
可以看出来sdfMap put进去了一次,而SimpleDateFormat被new了三次,因为代码中有三个线程.那么这是为什么呢?
对于每一个线程Thread,其内部有一个ThreadLocal.ThreadLocalMap threadLocals的全局变量引用,ThreadLocal.ThreadLocalMap里面有一个保存该ThreadLocal和对应value,一图胜千言,结构图如下:
那么对于sdfMap的话,结构图就变更了下
1.首先第一次执行DateUtil.formatDate(new Date(),MDHMSS);
//第一次执行DateUtil.formatDate(new Date(),MDHMSS)分析
private static SimpleDateFormat getSdf(final String pattern){
ThreadLocal sdfThread = sdfMap.get(pattern);
//得到的sdfThread为null,进入if语句
if (sdfThread == null){
synchronized (DateUtil.class){
sdfThread = sdfMap.get(pattern);
//sdfThread仍然为null,进入if语句
if (sdfThread == null){
//打印日志
logger.debug("put new sdf of pattern " + pattern + " to map");
//创建ThreadLocal实例,并覆盖initialValue方法
sdfThread = new ThreadLocal(){
@Override
protected SimpleDateFormat initialValue() {
logger.debug("thread: " + Thread.currentThread() + " init pattern: " + pattern);
return new SimpleDateFormat(pattern);
}
};
//设置进如sdfMap
sdfMap.put(pattern,sdfThread);
}
}
}
return sdfThread.get();
}
这个时候可能有人会问,这里并没有调用ThreadLocal的set方法,那么值是怎么设置进入的呢?
这就需要看sdfThread.get()的实现:
public T get() {
Thread t = Thread.currentThread();
ThreadLocalMap map = getMap(t);
if (map != null) {
ThreadLocalMap.Entry e = map.getEntry(this);
if (e != null) {
@SuppressWarnings("unchecked")
T result = (T)e.value;
return result;
}
}
return setInitialValue();
}
也就是说当值不存在的时候会调用setInitialValue()方法,该方法会调用initialValue()方法,也就是我们覆盖的方法.
对应日志打印.
put new sdf of pattern MMddHHmmssSSS to map thread: Thread[main,5,main] init pattern: MMddHHmmssSSS
2.第二次在子线程执行DateUtil.formatDate(new Date(),MDHMSS);
//第二次在子线程执行`DateUtil.formatDate(new Date(),MDHMSS);`
private static SimpleDateFormat getSdf(final String pattern){
ThreadLocal sdfThread = sdfMap.get(pattern);
//这里得到的sdfThread不为null,跳过if块
if (sdfThread == null){
synchronized (DateUtil.class){
sdfThread = sdfMap.get(pattern);
if (sdfThread == null){
logger.debug("put new sdf of pattern " + pattern + " to map");
sdfThread = new ThreadLocal(){
@Override
protected SimpleDateFormat initialValue() {
logger.debug("thread: " + Thread.currentThread() + " init pattern: " + pattern);
return new SimpleDateFormat(pattern);
}
};
sdfMap.put(pattern,sdfThread);
}
}
}
//直接调用sdfThread.get()返回
return sdfThread.get();
}
分析sdfThread.get()
//第二次在子线程执行`DateUtil.formatDate(new Date(),MDHMSS);`
public T get() {
Thread t = Thread.currentThread();//得到当前子线程
ThreadLocalMap map = getMap(t);
//子线程中得到的map为null,跳过if块
if (map != null) {
ThreadLocalMap.Entry e = map.getEntry(this);
if (e != null) {
@SuppressWarnings("unchecked")
T result = (T)e.value;
return result;
}
}
//直接执行初始化,也就是调用我们覆盖的initialValue()方法
return setInitialValue();
}
对应日志:
Thread[Thread-1,5,main] init pattern: MMddHHmmssSSS
总结
在什么场景下比较适合使用ThreadLocal？stackoverflow上有人给出了还不错的回答。
When and how should I use a ThreadLocal variable?
One possible (and common) use is when you have some object that is not thread-safe, but you want to avoid synchronizing access to that object (I'm looking at you, SimpleDateFormat). Instead, give each thread its own instance of the object.
参考代码:
https://github.com/nl101531/JavaWEB 下Util-Demo
参考资料:
深入浅出的学习Java ThreadLocal
SimpleDateFormat的线程安全问题与解决方案
以上就是本文的全部内容，希望对大家的学习有所帮助，也希望大家多多支持考高分网。
