>>>>> "Steven" == Steven Roussey <sroussey@stripped>
Steven> I doubt this will show up on a benchmark, since I think real world access
Steven> creates an environment where this bug occurs. The numbers below are an
Steven> average from a 100 iteration test program:
Steven> select count(*) from tbl where a=# and c=0;
Steven> The test seems to go quick but pauses from 1 to 5 times for several seconds
Steven> (that is, a query takes several seconds). This is what kills the average.
Steven> select a from tbl where a=# and c=0;
Steven> count number of rows in result set.
Steven> Again, tbl has a unique index(a,b) and a non-unique index(a,c). The number
Steven> of rows we are getting always falls between 0 and 1000.
Steven> I've run this many times. Sometimes both will take much longer (something
Steven> else on the server is causing the pause), but statistically the numbers
Steven> above consistant and very repeatable.
Steven> A factor of 100x difference! We are immediately converting to the second
Steven> Steven Roussey
Could you mail me a copy of the test program ?
select count(*) should be much faster!