欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

SimpleDateFormat线程不安全问题与ThreadLocal原理

程序员文章站 2024-03-15 16:10:59
...

SimpleDateFormat是JDK中长久以来自带的日期时间格式化类,但是它有线程安全性方面的问题,使用时要避免它带来的影响。

SimpleDateFormat是线程不安全的

写一个SimpleDateFormat在并发环境下简单的例子先。

public class SimpleDateFormatExample {
    private static SimpleDateFormat simpleDateFormat = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");

    public static void main(String[] args) {
        // 创建线程池,最好直接new ThreadPoolExecutor,而不是用Executors工具类
        // ExecutorService threadPool = Executors.newCachedThreadPool();
        ExecutorService threadPool = new ThreadPoolExecutor(
            5, 50, 0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<>(100)
        );

        List<String> dates = Arrays.asList(
            "2019-02-21 15:47:01",
            "2018-03-22 16:46:02",
            "2017-04-23 17:45:03",
            "2016-05-24 18:44:04",
            "2015-06-25 19:43:05",
            "2014-07-26 20:42:06",
            "2013-08-27 21:41:07",
            "2012-09-28 22:40:08",
            "2011-10-29 23:39:09"
        );

        for (String date : dates) {
            threadPool.execute(() -> {
                try {
                    System.out.println(simpleDateFormat.parse(date));
                } catch (ParseException e) {
                    e.printStackTrace();
                }
            });
        }
    }
}

输出非常混乱,抛出大量NumberFormatException,以及得出错误的结果。

Exception in thread "pool-1-thread-2" java.lang.NumberFormatException: For input string: ""
    at java.lang.NumberFormatException.forInputString(NumberFormatException.java:65)
    at java.lang.Long.parseLong(Long.java:601)
    at java.lang.Long.parseLong(Long.java:631)
    at java.text.DigitList.getLong(DigitList.java:195)
    at java.text.DecimalFormat.parse(DecimalFormat.java:2084)
    at java.text.SimpleDateFormat.subParse(SimpleDateFormat.java:1869)
    at java.text.SimpleDateFormat.parse(SimpleDateFormat.java:1514)
    at java.text.DateFormat.parse(DateFormat.java:364)
    at me.lmagics.SimpleDateFormatExample.lambda$main$0(SimpleDateFormatExample.java:42)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)
Exception in thread "pool-1-thread-4" Exception in thread "pool-1-thread-3" java.lang.NumberFormatException: For input string: ".809E.809E22"
    at sun.misc.FloatingDecimal.readJavaFormatString(FloatingDecimal.java:2043)
    at sun.misc.FloatingDecimal.parseDouble(FloatingDecimal.java:110)
    at java.lang.Double.parseDouble(Double.java:538)
    at java.text.DigitList.getDouble(DigitList.java:169)
    at java.text.DecimalFormat.parse(DecimalFormat.java:2089)
    at java.text.SimpleDateFormat.subParse(SimpleDateFormat.java:1869)
    at java.text.SimpleDateFormat.parse(SimpleDateFormat.java:1514)
    at java.text.DateFormat.parse(DateFormat.java:364)
    at me.lmagics.SimpleDateFormatExample.lambda$main$0(SimpleDateFormatExample.java:42)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)
java.lang.NumberFormatException: For input string: ".809E.809E22"
    at sun.misc.FloatingDecimal.readJavaFormatString(FloatingDecimal.java:2043)
    at sun.misc.FloatingDecimal.parseDouble(FloatingDecimal.java:110)
    at java.lang.Double.parseDouble(Double.java:538)
    at java.text.DigitList.getDouble(DigitList.java:169)
    at java.text.DecimalFormat.parse(DecimalFormat.java:2089)
    at java.text.SimpleDateFormat.subParse(SimpleDateFormat.java:1869)
    at java.text.SimpleDateFormat.parse(SimpleDateFormat.java:1514)
Wed Mar 23 16:46:02 CST 1
Tue May 24 18:44:04 CST 2016
    at java.text.DateFormat.parse(DateFormat.java:364)
Wed Mar 23 16:46:02 CST 1
    at me.lmagics.SimpleDateFormatExample.lambda$main$0(SimpleDateFormatExample.java:42)
Thu Feb 21 15:47:01 CST 2019
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
Thu Jun 25 19:43:05 CST 2015
Thu Oct 29 23:39:09 CST 2011
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)

然后通过SimpleDateFormat的源码来分析它线程不安全的根本原因。
SimpleDateFormat是继承自DateFormat类,DateFormat类中维护了一个全局的Calendar变量。

    /**
     * The {@link Calendar} instance used for calculating the date-time fields
     * and the instant of time. This field is used for both formatting and
     * parsing.
     *
     * <p>Subclasses should initialize this field to a {@link Calendar}
     * appropriate for the {@link Locale} associated with this
     * <code>DateFormat</code>.
     * @serial
     */
    protected Calendar calendar;

从注释可以看出,这个Calendar对象既用于格式化也用于解析日期时间。再查看parse()方法接近最后的部分。

        Date parsedDate;
        try {
            parsedDate = calb.establish(calendar).getTime();
            // If the year value is ambiguous,
            // then the two-digit year == the default start year
            if (ambiguousYear[0]) {
                if (parsedDate.before(defaultCenturyStart)) {
                    parsedDate = calb.addYear(100).establish(calendar).getTime();
                }
            }
        }
        // An IllegalArgumentException will be thrown by Calendar.getTime()
        // if any fields are out of range, e.g., MONTH == 17.
        catch (IllegalArgumentException e) {
            ...
        }

        return parsedDate;

可见,最后的返回值是通过调用CalendarBuilder.establish()方法获得的,而它的入参正好就是前面的Calendar对象。

    Calendar establish(Calendar cal) {
        boolean weekDate = isSet(WEEK_YEAR)
                            && field[WEEK_YEAR] > field[YEAR];
        if (weekDate && !cal.isWeekDateSupported()) {
            // Use YEAR instead
            if (!isSet(YEAR)) {
                set(YEAR, field[MAX_FIELD + WEEK_YEAR]);
            }
            weekDate = false;
        }

        cal.clear();
        // Set the fields from the min stamp to the max stamp so that
        // the field resolution works in the Calendar.
        for (int stamp = MINIMUM_USER_STAMP; stamp < nextStamp; stamp++) {
            for (int index = 0; index <= maxFieldIndex; index++) {
                if (field[index] == stamp) {
                    cal.set(index, field[MAX_FIELD + index]);
                    break;
                }
            }
        }

        if (weekDate) {
            int weekOfYear = isSet(WEEK_OF_YEAR) ? field[MAX_FIELD + WEEK_OF_YEAR] : 1;
            int dayOfWeek = isSet(DAY_OF_WEEK) ?
                                field[MAX_FIELD + DAY_OF_WEEK] : cal.getFirstDayOfWeek();
            if (!isValidDayOfWeek(dayOfWeek) && cal.isLenient()) {
                if (dayOfWeek >= 8) {
                    dayOfWeek--;
                    weekOfYear += dayOfWeek / 7;
                    dayOfWeek = (dayOfWeek % 7) + 1;
                } else {
                    while (dayOfWeek <= 0) {
                        dayOfWeek += 7;
                        weekOfYear--;
                    }
                }
                dayOfWeek = toCalendarDayOfWeek(dayOfWeek);
            }
            cal.setWeekDate(field[MAX_FIELD + WEEK_YEAR], weekOfYear, dayOfWeek);
        }
        return cal;
    }

该方法中先后调用了cal.clear()与cal.set(),也就是先清除cal对象中设置的值,再重新设置新的值。由于Calendar内部并没有线程安全机制,并且这两个操作也都不是原子性的,所以当多个线程同时操作一个SimpleDateFormat时就会引起cal的值混乱。类似地,format()方法也存在同样的问题。

为什么用ThreadLocal能解决问题

ThreadLocal即线程本地变量。它用来为每个线程维护一个专属的变量副本,线程对自己的变量副本进行操作时,对其他线程的变量副本没有任何影响。由此可见,它特别适合解决并发情况下变量共享造成的线程安全性问题,前提是各个副本隔离后不影响业务运行。
以上面的SimpleDateFormat问题为例,ThreadLocal可以这样使用。

// 可以直接设置初始值
private static ThreadLocal<SimpleDateFormat> simpleDateFormat = ThreadLocal.withInitial(() ->
    new SimpleDateFormat("yyyy-MM-dd HH:mm:ss")
);
// 也可以调用set()方法
private static ThreadLocal<SimpleDateFormat> simpleDateFormat = new ThreadLocal<>();
simpleDateFormat.set(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss"));
...
// 调用get()方法取得值
System.out.println(simpleDateFormat.get().parse(date));
// 移除
simpleDateFormat.remove();

那么ThreadLocal是采用了什么机制来实现变量副本隔离的呢?在Thread类内部,有如下的定义。

    /* ThreadLocal values pertaining to this thread. This map is maintained
     * by the ThreadLocal class. */
    ThreadLocal.ThreadLocalMap threadLocals = null;

可见每个线程都维护了一个叫ThreadLocalMap的东西,它是ThreadLocal中定义的一个静态内部类。其实现类似于HashMap,但没实现Map接口,数据结构和内部逻辑也有不同。ThreadLocalMap.Entry是这样定义的。

        static class Entry extends WeakReference<ThreadLocal<?>> {
            /** The value associated with this ThreadLocal. */
            Object value;

            Entry(ThreadLocal<?> k, Object v) {
                super(k);
                value = v;
            }
        }

该Entry的键值类型都是确定的。值就是变量的副本,键是对ThreadLocal对象的一个弱引用。由于线程并不能直接访问和存取ThreadLocalMap,只能藉由ThreadLocal进行,因此不同的线程之间的变量副本就实现了隔离。

SimpleDateFormat线程不安全问题与ThreadLocal原理

上面的图来自阿里Java开发手册,清晰地示出了线程、ThreadLocal、ThreadLocalMap三者的引用关系,鼓掌。
另外,ThreadLocal还可能存在内存泄漏的问题,前人已经写过很好的分析文章,如https://www.jianshu.com/p/a1cd61fa22da,下面稍作总结。

  • 回忆Java中的4种引用类型:强、软、弱、虚引用,其引用强度依次递减。其中弱引用只能存活到下一次Young GC发生之前。


    SimpleDateFormat线程不安全问题与ThreadLocal原理
  • ThreadLocalMap.Entry中的键就是弱引用,如果它被回收,会出现key为null但value仍然存在的情况(value是强引用,当然Entry也没有被回收),有内存泄漏风险。ThreadLocal的设计者已经考虑到了这种情况,调用get()/set()/remove()方法时,都会调用expungeStaleEntry()方法来删除这种key已经被回收了的Entry。这段代码很有意思,关键点添加了一点注释。
        private int expungeStaleEntry(int staleSlot) {
            Entry[] tab = table;
            int len = tab.length;

            // expunge entry at staleSlot
            // 将值和Entry都设成null,这样在下一次GC根搜索时均不可达,就被回收了
            tab[staleSlot].value = null;
            tab[staleSlot] = null;
            size--;

            // Rehash until we encounter null
            // 还没完,接下来会继续找key为null的其他Entry,一起删掉
            Entry e;
            int i;
            for (i = nextIndex(staleSlot, len);
                 (e = tab[i]) != null;
                 i = nextIndex(i, len)) {
                ThreadLocal<?> k = e.get();
                if (k == null) {
                    // 找到了,将值和Entry都设成null
                    e.value = null;
                    tab[i] = null;
                    size--;
                } else {
                    int h = k.threadLocalHashCode & (len - 1);
                    if (h != i) {
                        tab[i] = null;

                        // Unlike Knuth 6.4 Algorithm R, we must scan until
                        // null because multiple entries could have been stale.
                        // nextIndex()方法就是(i + 1)%len,ThreadLocalMap是采用线性探测开放定址解决hash冲突的
                        // 这比HashMap的链地址法(数组+链表)简单得多,当然ThreadLocal变量多了之后,解决冲突的时间会边长
                        while (tab[h] != null)
                            h = nextIndex(h, len);
                        tab[h] = e;
                    }
                }
            }
            return i;
        }
  • 但是就算这样设计了,也不能完全防止ThreadLocal内存泄漏,因为它可以是static的,也有可能在分配了变量副本之后没调用任何方法。另外,由于ThreadLocalMap的生命周期和线程一样长,因此不管Entry的键是对ThreadLocal的强引用还是弱引用,都有可能出现ThreadLocal被回收变成null的情况。
  • 所以,完全避免内存泄漏的唯一手段就是在ThreadLocal用完后,调用remove()方法。