RE: no index used

I seem to recall having better luck getting the results I wanted with unions instead of or statements, you might try that. Will >===== Original Message From steinhoefel@dcs-systeme.de ===== >Thanks again for your answers, > >I had already tried to change OPTCOMPIND from 2 to 0 - with no effect - and >used "update statistics" before I posted my question. >Today I tried a "update statistics high", and now I get better answers from >the server, but it is by far not what I expect from it. >It now uses simple indexes in the matter I'd expect - most of the time - but >it still refuses using combined indices. There is a combined index on >(kmi,mti) as you can see from my post of the table definition. On that table >I do selects with a lot of conditions, for instance 10 kmi conditions each >one with 40 mti conditons in the following matter: >select pkey from table1 where>kmi = 5 and (mti between 10 and 39 or mti between 60 and 99 or mti or mti >between...) or >kmi = 5262 and (mti between 0 and 4 or mti between 50 and 54 or mti >between...). > >For such selects the system does a table scan. This is not necessary, the >answer is just a small part of the whole table. That these my theory is >rigth I see at the results of selects just with kmi (let out all mti >conditions): it's very fast now, much faster than a table scan. >Even if I try just two kmi each one with one mti condition the Index is not >used. > >Today I encountered another problem: >kmi and mti are spatial index columns for lkr and lkh(x and y). So against >all theory I tried to use lkr and lkh directly, with no spatial index. So I >did a index on lkr, and selects with very small answers where accelerated >fine, but if had big answers (a lot of rows) it was a catastrophe. Selecting >the whole table with the index was as 15 times !!! as slow as a table scan. >Now my question: Is it generally not adviceable to use indices on float >columns in informix? >If I cannot resolve my mti problem I plan to put kmi and mti together into a >double (I need at least 32 bit for kmi and 7 bit for mti). So would this be >a useful try? ------------------------------------------------------------ This e-mail has been sent to you courtesy of OperaMail, as a free service from Opera Software, makers of the award-winning Web Browser, Opera. Visit us at http://www.opera.com/ or our portal at: http://www.myopera.com/ Your free e-mail account is waiting at: http://www.operamail.com/ ------------------------------------------------------------