zz_11@mail.bg writes:
> I am running a relativ complex query on pg 8.3.5 and have (possible)
> wrong query plan.
> ...
> If I run the query without thle last part : and n.num like '191%'
> it work ok as speed ~ 30 sec on not very big db.
> If I run the full query it take very long time to go ( i never waited
> to the end but it take > 60 min.)
I'm betting that it's badly underestimating the number of rows
satisfying the LIKE condition:
> -> Index Scan using
> i_nomen_num on a_nomen n (cost=0.00..56.39 rows=24 width=128)
> Index Cond:
> (((num)::text >= '191'::text) AND ((num)::text < '192'::text))
> Filter:
> ((num)::text ~~ '191%'::text)
Is 24 the right number of rows for that, or anywhere close? If not, try
raising the statistics target for this table.
regards, tom lane