--- Chris Eich <[EMAIL PROTECTED]> wrote: > Joe, the numbers are 100, 4236301 and 100 respectively (as you suspected, > there are few devices, many intervals, and all devices have intervals). > > Richard, sqlite_stat1 has these rows for the dpi indexes: > INSERT INTO "sqlite_stat1" VALUES('device_perf_interval','dpi1','4236301 > 2118151 21182 21182 1'); > INSERT INTO "sqlite_stat1" VALUES('device_perf_interval','dpi2','4236301 > 2118151 21182 14122 14122'); > > Before ANALYZE, SQLite wanted to run the query as (dpi, d). After, it runs > as (d, dpi) and does use three columns from dpi1 (thanks Igor!):
That's all fine and good, but the question remains - is your query faster pre-ANALYZE (no sqlite_stat1 table) or is it faster post-ANALYZE? Is the plan (dpi, d) faster than (d, dpi) ? ____________________________________________________________________________________ Looking for a deal? Find great prices on flights and hotels with Yahoo! FareChase. http://farechase.yahoo.com/ ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------