I have run tests on my UniData 6.0.9 system on Solaris 2.8.

I was testing what affected the sort order of
SELECT JOB WITH CONO = "001"
from an ascending ID sort before re-indexing (with an additional field) to a descending
ID sort after.

I tested every combination of the selected field having a format of "3L" or  "3R", of 
being defined as single (S) or multi (M) (not MV!) valued, the presence of a long 
field called DESC or JOB_DESC or not being included, and setting the alternate key 
length to 8 or 20.

The only investigated factor that changed the sort order of the select list was 
whether the field CONO, which is a virtual attribute that is part of the key, was 
defined as single valued "S" or multi-valued "M" in field 6 of the dictionary.  When 
CONO was single valued the SELECT provided an ascending key sort and when CONO was 
multi-valued the SELECT provided a descending key sort.

I thought that the field had always been defined with "M" in field 6 as a result of an 
automated conversion from Pick days to UniData and it certainly provided an ascending 
sort until I deleted and recreated the indices.  Who knows?

Karjala


>>> [EMAIL PROTECTED] 10/06/2004 5:18:31 PM >>>
Oops!

...

I will make a more rigorous round of tests tomorrow to try to pin down the change 
because I cannot yet replicate the changes I observed.  The tests will include the 
presence and absence of the DESC field, and if that makes a difference in a non-SQL 
environment, adding it with a different name.

Karjala


...

-----Original Message-----
From: Karjala Koponen [mailto:[EMAIL PROTECTED] 
...

I recently added an additional alternate index to one of my files.
...
The sort order of
SELECT JOB WITH CONO = "001"
changed from an ascending ID sort before the re-indexing to a descending
ID sort after.

I did the same things on UniData 5.2 on AIX and UniData 6.0.9 on Solaris
2.8 with the same results.
...
Karjala
-------
u2-users mailing list
[EMAIL PROTECTED]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to