Re: PostgreSQL 7.4beta5 vs MySQL 4.0.16 with

On Thu, 30 Oct 2003 mallah@trade-india.com wrote:

> >> So its not just PostgreSQL that is suffering from the bad SQL but MySQL also.  But the
> >> question is my does PostgreSQL suffer so badly ??  I think not all developers write very nice
> >> SQLs.
> >>
> >> Its really sad to see that a fine peice of work (RT) is performing sub-optimal becoz of
> >> malformed SQLs.  [ specially on database of my choice ;-) ]
> >
> > Post EXPLAIN ANALYZES of the queries you're running, then maybe you'll be able to get some
> > useful help from this list.  Until then, it's very hard to speculate as to why PostgreSQL is
> > slower.  -sc
>
> Here It is:
>
> in case they are illegeble please lemme know i will attach it as .txt
> files.
>
> Slower One:
>
> explain analyze SELECT DISTINCT main.* FROM Groups main , Principals Principals_1, ACL ACL_2
> WHERE ((ACL_2.RightName = 'OwnTicket')OR(ACL_2.RightName = 'SuperUser')) AND  ( (
> ACL_2.PrincipalId = Principals_1.id AND ACL_2.PrincipalType = 'Group' AND (   main.Domain =
> 'SystemInternal' OR main.Domain = 'UserDefined' OR main.Domain = 'ACLEquivalence') AND main.id =
> Principals_1.id) OR ( ( (main.Domain = 'RT::Queue-Role' AND main.Instance = 25)  OR ( main.Domain
> = 'RT::Ticket-Role' AND main.Instance = 6973)  )  AND main.Type = ACL_2.PrincipalType AND main.id
> = Principals_1.id) ) AND (ACL_2.ObjectType = 'RT::System'  OR (ACL_2.ObjectType = 'RT::Queue' AND
> ACL_2.ObjectId = 25) )  ORDER BY main.Name ASC ;

Note here:

Merge Join
    (cost=1788.68..4735.71 rows=1 width=85)
    (actual time=597.540..1340.526 rows=20153 loops=1)
    Merge Cond: ("outer".id = "inner".id)

This estimate is WAY off.  Are both of those fields indexed and analyzed?
Have you tried upping the statistics target on those two fields?
I assume they are compatible types.

You might try 'set enable_mergejoin = false' and see if it does something
faster here.  Just a guess.