【156期】面試官：談?wù)?SimpleDateFormat 的線程安全問題與解決方案
閱讀本文大概需要 6 分鐘。
來自：cnblogs.com/zemliu/p/3290585.html
1. 原因
Date parse() {
calendar.clear(); // 清理calendar
... // 執(zhí)行一些操作, 設(shè)置 calendar 的日期什么的
calendar.getTime(); // 獲取calendar的時間
}
2. 問題重現(xiàn)
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 會sleep 2s 后開始執(zhí)行sdf.parse()
executor.execute(new DateFormatTest("A", "1991-09-13", true));
// B 打了斷點,會卡在方法中間
executor.execute(new DateFormatTest("B", "2013-09-13", false));
executor.shutdown();
}
}
parse() {
calendar.clear()
// 這里打一個斷點
calendar.getTime()
}
首先A線程跑起來以后會進入sleep
B線程跑起來, 卡在斷點處
A線程醒過來, 執(zhí)行 calendar.clear(), 并將設(shè)置sdf.calendar的date為1991-09-13, 此時 A B 的 calendar 都為 1991-09-13
讓斷點繼續(xù)執(zhí)行, 輸出如下
A : date: Fri Sep 13 00:00:00 CDT 1991 B : date: Fri Sep 13 00:00:00 CDT 1991
3. 解決方案
假如說日期函數(shù)處理類使用的是new SimpleDateFormat的方法,那么這里就會有1000次sdf的創(chuàng)建和銷毀
Java中提供了一種ThreadLocal的解決方案,它的工作方式是,每個線程只會有一個實例,也就是說我們執(zhí)行完這1000個任務(wù),總共只會實例化4個sdf.
import java.text.ParseException;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.HashMap;
import java.util.Map;
public class DateUtil {
/** 鎖對象 */
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>>();
/**
* 返回一個ThreadLocal的sdf,每個線程只會new一次sdf
*
* @param pattern
* @return
*/
private static SimpleDateFormat getSdf(final String pattern) {
ThreadLocal<SimpleDateFormat> tl = sdfMap.get(pattern);
// 此處的雙重判斷和同步是為了防止sdfMap這個單例被多次put重復(fù)的sdf
if (tl == null) {
synchronized (lockObj) {
tl = sdfMap.get(pattern);
if (tl == null) {
// 只有Map中還沒有這個pattern的sdf才會生成新的sdf并放入map
System.out.println("put new sdf of pattern " + pattern + " to map");
// 這里是關(guān)鍵,使用ThreadLocal<SimpleDateFormat>替代原來直接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>來獲取SimpleDateFormat,這樣每個線程只會有一個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("單線程執(zhí)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("雙線程執(zhí)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();
}
}
}
單線程執(zhí)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
雙線程執(zhí)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個線程執(zhí)行這6個任務(wù)的時候,這個線程首次使用過的時候會new一個新的sdf,并且以后都一直用這個sdf,而不是每次處理任務(wù)都新建一個新的sdf
2個線程執(zhí)行6個任務(wù)的時候也是同理,但是2個線程的sdf是分開的,每個線程都有自己的"yyyy-MM-dd", "yyyy-MM"的sdf,所以他們不會有線程安全安全問題
推薦閱讀：
【155期】面試官：你遇到過log4j2線程阻塞的場景嗎，如何解決呢？
【154期】面試官：你能說說 Elasticsearch 查詢數(shù)據(jù)的工作原理是什么嗎？
【153期】面試官：談?wù)劤Ｓ玫腁rraylist和Linkedlist的區(qū)別
微信掃描二維碼，關(guān)注我的公眾號
朕已閱 