Archives of the TeradataForum

Message Posted: Mon, 03 May 2004 @ 17:15:26 GMT


	<Prev	Next>		<<First	<Prev	Next>	Last>>

Subj:		Re: Converting time in miliseconds to TIEMSTAMP in teradata

From:		Dieter Noeth

Victor Sokovin wrote:

It has just occurred to me to apply the following trick and it seems to be working for the interval test values I have had in my table. I don't have time to do more testing right now so I have no clue whether the trick will always work but I thought I should mention it for those who might be looking now for a quick workaround.

Dieter, if you are reading this: would this modification work in your test case?

If n < 2095 ;-)

I already found a workaround similar to Donald's CAST, but it's also wrong (similar, but not exactly):

       10 * (i * n) / 10  as x,
       cast(i * n as interval hour(4) to second)  as x2,


               i     n                x             x2       cum_sum
     -----------  ----  ---------------  -------------  ------------
     ...
     0:00:01.999  2093   1:09:43.907000    1:09:43.907   1:09:43.907
     0:00:01.999  2094   1:09:45.906000    1:09:45.906   1:09:45.906
     0:00:01.999  2095  -0:01:47.062296   -0:01:47.062   1:09:47.905
     0:00:01.999  2096  -0:01:45.063296   -0:01:45.063   1:09:49.904
     0:00:01.999  2097  -0:01:43.064296   -0:01:43.064   1:09:51.903
     0:00:01.999  2098  -0:01:41.065296   -0:01:41.065   1:09:53.902
     0:00:01.999  2099  -0:01:39.066296   -0:01:39.066   1:09:55.901
     0:00:01.999  2100   1:09:57.900000    1:09:57.900   1:09:57.900
     0:00:01.999  2101   1:09:59.899000    1:09:59.899   1:09:59.899

Now MDiff is sometimes not returning the same difference. And if you calculate the difference between cum_sum and x it's not always 0.

Dieter


	<Prev	Next>		<<First	<Prev	Next>	Last>>

Attachments

Library

Quick Reference