On 25 Jun 2011, at 7:14pm, Black, Michael (IS) wrote: > I guess my point is why doesn't the optimizer recognize the almost-zero cost > of min/max in the compound query? It assumes instead to do it all in one scan.
Because it's almost impossible to optimize sub-SELECTs. You pretty-much have to perform the sub-SELECT and look at the results before you can figure out how to optimise it. If you want good optimisation, use JOIN, which part of the SQL standard and has a syntax which leads to easier optimization. > select * from (select min(i) from test) as a,(select max(i) from test) as b, > (select count(i) from test) as c; That makes no sense in terms of what you'd really want to do with a database, so the optimizer is going to have trouble with it. Simon. _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
