Re: strange explain - Mailing list pgsql-hackers

From Zeugswetter Andreas SB SD
Subject Re: strange explain
Date
Msg-id 46C15C39FEB2C44BA555E356FBCD6FA4961DCA@m0114.s-mxs.net
Whole thread Raw
In response to strange explain  (Oleg Bartunov <oleg@sai.msu.su>)
Responses Re: strange explain  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
> EXPLAIN
> tour=# explain analyze  select * from tours  where
>            ( operator_id in (2,3,4,5,7) and type_id = 4 ) or
>            ( operator_id = 8 and type_id = 3);
> NOTICE:  QUERY PLAN:
>
> Index Scan using type_idx, type_idx, type_idx, type_idx, type_idx, type_idx on tours  (cost=>
> 0.00..12.25 rows=1 width=1091) (actual time=0.27..0.27 rows=0 loops=1)
> Total runtime: 0.44 msec

Actually this plan looks very strange to me. One would expect it to only use
type_idx twice (in lack of a better index (type_id, operator_id)).
Fetch all rows with type_id=4 and then filter the result on operator_id in (...),
then do type_id=3 and filter operator_id=8.
Seems there is room for another performance improvement here :-)

Andreas


pgsql-hackers by date:

Previous
From: Jean-Michel POURE
Date:
Subject: Re: [BUGS] Bug #659: lower()/upper() bug on
Next
From: Hannu Krosing
Date:
Subject: Re: Discontent with development process (was:Re: pgaccess