x,y are both column of Table "aa", they may have random value . And you think that will affect the performance.?
What I think is Sqlite will first use "where cluase" to get limited value ,then use "group by" in the same . so , the perfomance depend "where" clause , correct? Igor Tandetnik <[EMAIL PROTECTED]> 写道: Bo Lin wrote: > I just found a strange case , can you give me some explaination ? > > The Sql looks like > select x, y, sum(z)/1000 as bw from aa where > a=1 and b=1 and > c =1 and d= 6 group by x, y having count(*) > 1 order by > bw desc limit 10 > > Column "d" is all set to 6 > 1/3 of total record is (a=1,b=1,c=1) > 1/3 of total record is (a=2,b=2,c=2) > 1/3 of total record is (a=3,b=3,c=3) Where do x and y come from, what are their values? Igor Tandetnik __________________________________________________ 赶快注册雅虎超大容量免费邮箱? http://cn.mail.yahoo.com