Archives of the TeradataForum
Message Posted: Tue, 09 Jul 2002 @ 16:32:39 GMT
| Subj: | | Re: Uniform Distributions from RANDOM number function |
| |
| From: | | Maxwell, Donald |
Before everyone begins to think I just want to start an argument here, I would like to clarify:
I suggested this as a "longshot" idea, and personally, I'd have a hard time justifying this approach in my own production environment (in
other words - "a snowball's chance in hell" of actually making it to production).
But, I continue to disagree with the objections raised so far. The colliding uniqueness problem is solved by concatenating the unique
batch row identifier (BYTE datatype ). Concatenation of BYTE datatypes is supported. For that matter, one could define a two-column UPI on
ROWID, BatchROWID without concatenation.
More to the point, there is no need to "re-sequence" the uniqueness value or otherwise modify the surrogate to hold sequential or other
"relationship" information. It is a "pure" surrogate, whose only requirement is that it is different than the next surrogate!
Donald
| 
