集合为什么不从排序列表中删除重复项？

Question

Why doesn't the set remove duplicates from the sorted list ? I have sorted the dates in ascending order which the first printed line shows but the set isn't removing the duplicates. Why is that ? The program prints:
[Mon Apr 20 12:27:47 CDT 2009, Mon Apr 20 12:27:47 CDT 2009, Sun Dec 20 12:27:47 CST 2009] [Sun Dec 20 12:27:47 CST 2009, Mon Apr 20 12:27:47 CDT 2009, Mon Apr 20 12:27:47 CDT 2009]

Shouldn't creating a set remove the duplicate date from the collection ?

    def void testLoadDoc()
     {
     Date date1 = getCurrentDate(3,20,2009)
     Date date2 = getCurrentDate(11,20,2009)
     Date date3 = getCurrentDate(3,20,2009)


     List<Date> dates = new ArrayList<Date>();
     dates.add(date2);
     dates.add(date1);
     dates.add(date3);

     Collections.sort(dates, new CurrencyDateComparator());
     Set uniqueDates = new HashSet(dates)

     println dates
     println uniqueDates

 }


 private Date getCurrentDate(int month, int day, int year)
 {

     Calendar cal = Calendar.getInstance();
     cal.set(YEAR, year);
     cal.set(MONTH, month);
     cal.set(DAY_OF_MONTH, day);
     return cal.getTime();

 }
}

class CurrencyDateComparator implements Comparator
{
    /* (non-Javadoc)
     * @see java.util.Comparator#compare(java.lang.Object, java.lang.Object)
     */
    public int compare(Object arg0, Object arg1)
    {
        Date p = (Date) arg0;
        Date q = (Date) arg1;

        if (p.before(q))
        {
            return -1;
        }
        else if (p.after(q))
        {
            return 1;
        }
        else
        {
            return 0;
        }
    }

    public boolean equals(Object o)
    {
        if (o instanceof CurrencyDateComparator)
        {
            CurrencyDateComparator c = (CurrencyDateComparator) o;

            return this.equals(o);
        }
        else
        {
            return false;
        }
    }

Answer 1

All your dates have different millisecond values, that are not printed in your traces. Set the milliseconds to 0.

Answer 2

First

java.util.Date carries time information with it, you need to zero out the time information then set what you want to be set or exclude the time portion from the comparison inside your custom Comparator .

Second

Your custom Comparator is buggy and redundant. java.util.Date already implements Comparable so you don't need to compare dates again, just called .compare() on the first Date instance with the second one as an argument.

Third

Also remember that MONTHS are ZERO based in Caledar. January == 0 not 1 .

Fourth

Never rely on the default toString() on java.util.Date , always use a SimpleDateFormatter instance that displays the entire time stamp down to the milliseconds including a TimeZone ( yyyy-MM-dd'T'HH:mm:ssZ ), preferably ISO-8601 and UTC TimeZone for deterministic behavior.

Answer 3

When you use Calendar.getInstance() it returns "now" to millisecond precision. Changing the YMD fields does not clear any of that out, so your two dates are likely off by a millisecond or twenty, which is not included in the default toString() implementation that you're seeing output from.

If you want to set a calendar to a specific time, it is simplest to clear() it first before starting to set fields.

Answer 4

When you call Calendar.getInstance() , you get a Calendar initialized with the current time. This includes hours, minutes, seconds, and milliseconds. Call clearTime() on it to zero them out.

Answer 5

You can use this collection instead

Set<Date> dates = new TreeSet<Date>(CurrencyDateComparator.INSTANCE);
dates.add(date2);
dates.add(date1);
dates.add(date3);

enum CurrencyDateComparator implements Comparator<Date> {
    INSTANCE;

    static final long MILLI_PER_DAY = 86400000L;
    public int compare(Date d1, Date d2) {
        // older versions of Java can use Double.compare()
        return Long.compare(d1.getTime() / MILLIS_PER_DAY, d2.getTime() / MILLIS_PER_DAY);
    }
}