On Tue, 2007-03-27 at 13:12 -0600, John Stanton wrote: > I suspect that the timing difference is due to page overflows. I did > only a cursory browse of the B-Tree code but it is just a guess. A test > would be to make a simple table with two adjacent integer columns and > time raising an index on one column and on both. If the times are > comparable the speed difference reported in this thread is a page > overflow artifact.
Thanks, John, Good idea -- I'll give that a try. I wouldn't have expected that with such a small record, but it's worth testing. (The schema again, in which "value" is a string of max 15 chars: CREATE TABLE keyword (key, contyp int, imagecount int, searchcat int, value, nextword, sec, ipr, fldseq int); CREATE INDEX valuekey on keyword (value, key); Thanks! Stephen Stephen Toney Systems Planning [EMAIL PROTECTED] http://www.systemsplanning.com ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
