RE: [U2] Order of criteria

2005-02-24 Thread Stevenson, Charles
EXPLAIN is Universe-only, not Unidata. Sorry about that. It goes anywhere on the RetrieVe command line. If run from tcl, output is to the screen. -Original Message- From: Don Verhagen Charles, What is the syntax for the "EXPLAIN" keyword I don't see it documented in Unidata (6.X) I t

RE: [U2] Order of criteria

2005-02-24 Thread Don Verhagen
Charles, What is the syntax for the "EXPLAIN" keyword I don't see it documented in Unidata (6.X) I tried: SELECT MY.FILE WITH MY.FILE="SOMETHING" EXPLAIN All it did was select the records, does it EXPLAIN in a log file or somewhere else? Thanks, -- Donald Verhagen Application Deve

Re: [U2] Order of criteria

2005-02-24 Thread Roger Glenfield
Under Universe 9.6, I found that it was much faster to do two selects, the first one with just the index and then the second for the rest of the conditions. The new versions are supposed to optimize for indexing and also for tfile conversions. Roger Kevin King wrote: As I understand it, the q

RE: [U2] Order of criteria

2005-02-24 Thread Kryka, Richard
I agree with Kevin's answer. Also, it may depend on which U2 product/pick-like product you are using. I always try to order the WITH statements so the one selecting the fewest records is first. However, I always place the fields with translates last since I know that translates are much more exp

RE: [U2] Order of criteria

2005-02-24 Thread Stevenson, Charles
Generally speaking, you've got it right, John. A few caveats, notes, gotchas: - Use EXPLAIN keyword to get RetrieVe to show you what it is going to do. - If a field is indexed, and no select list is active when the retrieve command starts, the index will be applied before any actual select & da

RE: [U2] Order of criteria

2005-02-24 Thread Kevin King
As I understand it, the query optimizer can shuffle things around as necessary, but the real issue is whether any of the fields are indexed. If the fields are indexed you might get better response by selecting those using an index first, and then subselecting the remaining ones after the initial s