Hi Pablo,

>Don't mind this is a Carlos Yohn question and he solved it, the answer is in the newsgroup.

For any lurkers, my reply was:

Hi Carlos,

>>I've found a situation in which 'general' is very faster than 'machine'. ¿Can you test it and send me some comments?<<

That's interesting. I ran your sample with extended Rushmore showplan info, that is SYS(3054,11) and the difference in speed is that GENERAL and SPANISH both use the index on CAMPO1 in table MUCHOS, while the MACHINE sequence doesn't use the existing index, but rather creates a temporary one. That happens as soon as a join spans more than one field, eg. you select on CAMPO1 and CAMPO2. If you only use one field, all three COLLATE sequences are equally fast.

Nonetheless, especially with numeric fields the GENERAL collate sequence can cause wrong results, and I'd prefer correct results over fast results. *s*

Anyway, I pass this bug on to the Microsoft team. I hope it gets fixed soon.


Christof