[sqlite] Performace regression bug of sqlite 3.6.18

2010-01-12 Thread Hub Dog
*I just checked the **explain result of **sql below under 3.6.17 and 3.6.22, the result is quite different. SQL: -- romermb's incredible combination of events, shared events plus 'marriage' and alternate names as Facts 2010-01-06 -- revised by ve3meo to to bring out RoleName from RoleTableand Shar

Re: [sqlite] Performace regression bug of sqlite 3.6.18

2010-01-11 Thread Max Vlasov
On Mon, Jan 11, 2010 at 4:17 AM, Hub Dog wrote: > I think I found a performance regression bug of sqlite 3.6.1 8. A sql will > cost 1800 seconds to return the query result with sqlite 3.6.18 and with > the > previous version it only cost about 170 seconds. > I have a suggestion for you. If you'r

[sqlite] Performace regression bug of sqlite 3.6.18

2010-01-10 Thread Hub Dog
I think I found a performance regression bug of sqlite 3.6.1 8. A sql will cost 1800 seconds to return the query result with sqlite 3.6.18 and with the previous version it only cost about 170 seconds. I just attach the sql can reproduce the bug under below. But the database is a little large (4M),