Re: join on three tables is slow

Gerry Reno <greno@verizon.net> writes:
> Pavel Stehule wrote:
>> there is diference in agg position. Send, please, query and explain
>> analyze output.

[ explain analyze output ]

The rowcount estimates seem pretty far off, even for simple cases that
I'd expect it to get right, eg

>          ->  Seq Scan on res_partner_address a  (cost=0.00..88.40 
> rows=16 width=552) (actual time=0.851..16.131 rows=559 loops=1)
>                Filter: ((("type")::text = 'default'::text) OR ("type" IS 
> NULL))

Are the ANALYZE stats up to date for these tables?

What PG version is this, anyway?
        regards, tom lane