SybTimestamp DateType miss Milliseconds in jre1.3.1!

etienne · **Posted:** Wed Aug 17, 2005 4:50 pm

Hi,

I know there is a lot of mail on the Date nanosecond issue. But I found a strange issue, just wanna know if anyone have found this problem too.

I use a Sybase database. My Pojo have a java.util.Date GeneratedDateTime property mapped on a datetime row of a mapped table.

In dbVisualiser (under jre1.4 probably) I see that this table contain this data >2004-04-06 07:30:21.303< in this column for a given row. This means 6 April 2004, 7:30:21 + 303 milliseconds.

When I use the JRE1.4 to make a Unit testing, it works as it should.

Code:

// convert util use the SimpleDateFormater to create a Date from a String
Date generatedDateTime = DateConvertUtil.convertStringToDate("2004-04-06 07:30:21.303", "yyyy-MM-dd hh:mm:ss.SSS");
// vo is the Pojo object get from the DAO...
assertEquals(generatedDateTime, vo.getGeneratedDateTime());

But if I set my JRE to 1.3.1 (as my production environment is currenly set) the Unit test fail!

I have this strange error:

Code:

junit.framework.AssertionFailedError: expected:<Tue Apr 06 07:30:21 EDT 2004> but was:<2004-04-06 07:30:21.303>
   at junit.framework.Assert.fail(Assert.java:47)

It seems that when using JRE1.3 the milliseconds part is not load from the database!

But in the debug mode, I see that "303" was actually load! But it said that 303 is now the amount of nano seconds, not milliseconds! Is that a bug of com.sybase.jdbc2.tds.SybTimestamp?

From my debugger:
GeneratedDateTime is the value expected from the construction of a Date following this pattern : yyyy-MM-dd hh:mm:ss.SSS. The date is ok.
expected= Date (id=46)
cal= null
fastTime= 1081251021303

vo.getGeneratedDateTime() is loaded from the DB. This date miss the milliseconds part!
actual= SybTimestamp (id=166)
cal= null
fastTime= 1081251021000
nanos= 303333333

Does anyone knows how to get ride of that?

Mapping of this property:

Code:

      <property name="generatedDateTime" type="java.util.Date">
         <column name="GeneratedDateTime" scale="23" precision="3"
            not-null="false" sql-type="datetime" />
      </property>

It seems quite straight foward.

Thanks

Etienne.

etienne · **Posted:** Wed Aug 17, 2005 4:58 pm

In fact the problem is on the other side, Hibernate did load the 303 milliseconds because 303333333 nanoseconds is in fact 303 milliseconds.

But now I dont see why the Date Object I create with the SimpleDateFormat is not equal to 2004-04-06 07:30:21.303...

expected= Date (id=46)
cal= null
fastTime= 1081251021303

Any idea?

Etienne.

Jason Zhicheng Li · **Joined:** Tue Aug 16, 2005 11:06 pm **Posts:** 46

Note that Date equals() is based on how many milliseconds since January 1, 1970, 00:00:00 GMT.

etienne · **Posted:** Thu Aug 18, 2005 9:46 am

Yes, I know that Date is compared in milliseconds. The problem is not directly related to Hibernate (sorry Hibernate Group), but how the comparison is made between 2 dates (Date and SybTimestamp) in 2 JRE:

Between the expected Date "1081251021303 milliseconds" and the actual SybTimestamp date get from Hibenate/Sybase "1081251021000 milliseconds + 303333333 nanoseconds." Theses Date are not compared the same way in 2 JRE.

In IBMJRE 1.3.1 it is not equals, and in SunJRE1.4 it is consider equal!

I am quite confuse here.

If anyone did have this problem it could help me. I will maybe be less strict on the Unit Testing then, because Hibernate is the one who give me the right date.

this is my converter, is the error there?

Code:

   public static Date convertStringToDate(String stringValue, String format) {
      SimpleDateFormat formatter = new SimpleDateFormat(format);
      try {
         return formatter.parse(stringValue);
      } catch (ParseException e) {
         throw new UserTypeParsingRtException(
               "Cannot convert this String to date " + stringValue, e);
      }
   }

Thanks all.

Etienne.