【156期】面试官：谈谈 SimpleDateFormat 的线程安全问题与解决方案
共 19853字，需浏览 40分钟
·
2021-03-17 00:12
阅读本文大概需要 6 分钟。
来自：cnblogs.com/zemliu/p/3290585.html
1. 原因
Date parse() {
calendar.clear(); // 清理calendar
... // 执行一些操作, 设置 calendar 的日期什么的
calendar.getTime(); // 获取calendar的时间
}
2. 问题重现
import java.text.ParseException;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
public class DateFormatTest extends Thread {
private static SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd");
private String name;
private String dateStr;
private boolean sleep;
public DateFormatTest(String name, String dateStr, boolean sleep) {
this.name = name;
this.dateStr = dateStr;
this.sleep = sleep;
}
@Override
public void run() {
Date date = null;
if (sleep) {
try {
TimeUnit.MILLISECONDS.sleep(2000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
try {
date = sdf.parse(dateStr);
} catch (ParseException e) {
e.printStackTrace();
}
System.out.println(name + " : date: " + date);
}
public static void main(String[] args) throws InterruptedException {
ExecutorService executor = Executors.newCachedThreadPool();
// A 会sleep 2s 后开始执行sdf.parse()
executor.execute(new DateFormatTest("A", "1991-09-13", true));
// B 打了断点,会卡在方法中间
executor.execute(new DateFormatTest("B", "2013-09-13", false));
executor.shutdown();
}
}
parse() {
calendar.clear()
// 这里打一个断点
calendar.getTime()
}
首先A线程跑起来以后会进入sleep
B线程跑起来, 卡在断点处
A线程醒过来, 执行 calendar.clear(), 并将设置sdf.calendar的date为1991-09-13, 此时 A B 的 calendar 都为 1991-09-13
让断点继续执行, 输出如下
A : date: Fri Sep 13 00:00:00 CDT 1991 B : date: Fri Sep 13 00:00:00 CDT 1991
3. 解决方案
假如说日期函数处理类使用的是new SimpleDateFormat的方法,那么这里就会有1000次sdf的创建和销毁
Java中提供了一种ThreadLocal的解决方案,它的工作方式是,每个线程只会有一个实例,也就是说我们执行完这1000个任务,总共只会实例化4个sdf.
import java.text.ParseException;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.HashMap;
import java.util.Map;
public class DateUtil {
/** 锁对象 */
private static final Object lockObj = new Object();
/** 存放不同的日期模板格式的sdf的Map */
private static Map<String, ThreadLocal<SimpleDateFormat>> sdfMap = new HashMap<String, ThreadLocal<SimpleDateFormat>>();
/**
* 返回一个ThreadLocal的sdf,每个线程只会new一次sdf
*
* @param pattern
* @return
*/
private static SimpleDateFormat getSdf(final String pattern) {
ThreadLocal<SimpleDateFormat> tl = sdfMap.get(pattern);
// 此处的双重判断和同步是为了防止sdfMap这个单例被多次put重复的sdf
if (tl == null) {
synchronized (lockObj) {
tl = sdfMap.get(pattern);
if (tl == null) {
// 只有Map中还没有这个pattern的sdf才会生成新的sdf并放入map
System.out.println("put new sdf of pattern " + pattern + " to map");
// 这里是关键,使用ThreadLocal<SimpleDateFormat>替代原来直接new SimpleDateFormat
tl = new ThreadLocal<SimpleDateFormat>() {
@Override
protected SimpleDateFormat initialValue() {
System.out.println("thread: " + Thread.currentThread() + " init pattern: " + pattern);
return new SimpleDateFormat(pattern);
}
};
sdfMap.put(pattern, tl);
}
}
}
return tl.get();
}
/**
* 是用ThreadLocal<SimpleDateFormat>来获取SimpleDateFormat,这样每个线程只会有一个SimpleDateFormat
*
* @param date
* @param pattern
* @return
*/
public static String format(Date date, String pattern) {
return getSdf(pattern).format(date);
}
public static Date parse(String dateStr, String pattern) throws ParseException {
return getSdf(pattern).parse(dateStr);
}
}
import java.text.ParseException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
public class Test {
public static void main(String[] args) {
final String patten1 = "yyyy-MM-dd";
final String patten2 = "yyyy-MM";
Thread t1 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("1992-09-13", patten1);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
Thread t2 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("2000-09", patten2);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
Thread t3 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("1992-09-13", patten1);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
Thread t4 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("2000-09", patten2);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
Thread t5 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("2000-09-13", patten1);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
Thread t6 = new Thread() {
@Override
public void run() {
try {
DateUtil.parse("2000-09", patten2);
} catch (ParseException e) {
e.printStackTrace();
}
}
};
System.out.println("单线程执行: ");
ExecutorService exec = Executors.newFixedThreadPool(1);
exec.execute(t1);
exec.execute(t2);
exec.execute(t3);
exec.execute(t4);
exec.execute(t5);
exec.execute(t6);
exec.shutdown();
sleep(1000);
System.out.println("双线程执行: ");
ExecutorService exec2 = Executors.newFixedThreadPool(2);
exec2.execute(t1);
exec2.execute(t2);
exec2.execute(t3);
exec2.execute(t4);
exec2.execute(t5);
exec2.execute(t6);
exec2.shutdown();
}
private static void sleep(long millSec) {
try {
TimeUnit.MILLISECONDS.sleep(millSec);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
单线程执行:
put new sdf of pattern yyyy-MM-dd to map
thread: Thread[pool-1-thread-1,5,main] init pattern: yyyy-MM-dd
put new sdf of pattern yyyy-MM to map
thread: Thread[pool-1-thread-1,5,main] init pattern: yyyy-MM
双线程执行:
thread: Thread[pool-2-thread-1,5,main] init pattern: yyyy-MM-dd
thread: Thread[pool-2-thread-2,5,main] init pattern: yyyy-MM
thread: Thread[pool-2-thread-1,5,main] init pattern: yyyy-MM
thread: Thread[pool-2-thread-2,5,main] init pattern: yyyy-MM-dd
1个线程执行这6个任务的时候,这个线程首次使用过的时候会new一个新的sdf,并且以后都一直用这个sdf,而不是每次处理任务都新建一个新的sdf
2个线程执行6个任务的时候也是同理,但是2个线程的sdf是分开的,每个线程都有自己的"yyyy-MM-dd", "yyyy-MM"的sdf,所以他们不会有线程安全安全问题
推荐阅读：
【155期】面试官：你遇到过log4j2线程阻塞的场景吗，如何解决呢？
【154期】面试官：你能说说 Elasticsearch 查询数据的工作原理是什么吗？
【153期】面试官：谈谈常用的Arraylist和Linkedlist的区别
微信扫描二维码，关注我的公众号
朕已阅 