On Jun 14, 2006, at 16:34 UTC, Eduardo wrote: > >I need to select a large set of records out of a table, sort them by > >one column, and then get just a subset of the sorted list. (For > >example, I might want records 40-60 ordered by date, which is a > >completely different set than records 40-60 ordered by user ID.) I > >start with the full list of record IDs I want, and a query something like > >this: > > So you want the records from position 40 to position 60 ?
Yes, if by "position" you are referring to the sorted list. > In DB2 was faster select the 0-59 (1 to 60) records from the date > ordered table and from that selection in inverse order, select the > 0-19 (1-20). It was faster because selecting records from 0 (1) to > N-1 (N) was a lot faster than using the select records M-1(M) to > N-1(N), even when used two times. In my tests, selecting the 20 records I want is no problem (iterating througheven 13000 records once the SELECT returns is lickety-split). It's mainly the sorting that was the problem. (Though the SELECT itself takes several seconds, but I'm assuming that that's already as fast as it can get.) Best, - Joe -- Joe Strout -- [EMAIL PROTECTED] Verified Express, LLC "Making the Internet a Better Place" http://www.verex.com/
