nested query too expensive - Mailing list pgsql-admin

From Sub Director - Sistemas Informáticos
Subject nested query too expensive
Date
Msg-id 003f01c31ae4$b15a59e0$db00a8c0@219
Whole thread Raw
List pgsql-admin
This a relatively simple nested query that we try to use, but it finish in a "seq scan" with a
too high cost, so we had to use a little orthodox solution creating a temporal table into the
terminal and scanning this table row's one by one making individual querys for each one.
 
Any body knows how to make the query work in "index scan" mode ?
 
________________________________________________________
 
explain select w.*,b.nombre from (select nro_insc,cod_estab,cuitempre,impuesto,sum(monto_impo)
 as totret,sum(monto_rete) as suma_rete,tipodoc,documento from detadj where nro_insc=390009
 and cod_estab=0 and ano=2003 and mes=4 and per=2 and sec=0 group by nro_insc,cod_estab,
 cuitempre,impuesto,tipodoc,documento) w LEFT OUTER JOIN retper b on (w.tipodoc=b.tipodoc
 and btrim(w.documento) like btrim(b.documento) and btrim(w.cuitempre) like btrim(b.cuitempre)
 and w.nro_insc=b.nro_insc and w.cod_estab=b.cod_estab)
 
_______________________________________________
 
TABLES STRUCTURE:
 
 Table "retper"  ( 180.000 rows )
 
  Column   |     Type      | Modifiers
-----------+---------------+-----------
 tipodoc   | integer       |
 documento | character(20) |
 nombre    | character(40) |
 domicilio | character(40) |
 puerta    | integer       |
 localidad | character(15) |
 provincia | character(15) |
 ningbru   | character(20) |
 c_postal  | character(8)  |
 cuitempre | character(20) |
 nro_insc  | integer       |
 cod_estab | integer       |
 graba     | date          |
 hora      | character(4)  |
 opera     | integer       |
 puesto    | integer       |
 crc       | character(4)  |
Indexes:
         cuitemp_btrim,
         docu_btrim,
         retper_cod_estab,
         retper_cuitempre,
         retper_documento,
         retper_nombre,
         retper_nro_insc,
         retper_tipodoc
 
________________________________________________
 
Table "detadj"  ( 18.500.000 rows )
 
   Column   |         Type          | Modifiers
------------+-----------------------+-----------
 cuitempre  | character varying(20) |
 sec        | numeric(10,0)         |
 per        | numeric(10,0)         |
 mes        | numeric(10,0)         |
 ano        | numeric(10,0)         |
 nro_insc   | numeric(10,0)         |
 cod_estab  | numeric(10,0)         |
 nobli      | character varying(20) |
 cod_act    | character varying(20) |
 tipo_agen  | character varying(1)  |
 monto_impo | double precision      |
 alicuota   | double precision      |
 monto_rete | double precision      |
 tipodoc    | numeric(10,0)         |
 documento  | character varying(20) |
 impuesto   | numeric(10,0)         |
 tipo_dato  | numeric(10,0)         |
 id         | character varying(11) |
 tipo_comp  | numeric(10,0)         |
 letra      | character varying(1)  |
 terminal   | numeric(10,0)         |
 numero     | character varying(20) |
 fecha      | date                  |
 ningbru    | character varying(20) |
 graba      | date                  |
 hora       | character varying(4)  |
 opera      | numeric(10,0)         |
 puesto     | numeric(10,0)         |
Indexes:
         ano_detadj,
         ano_mes_per,
         cod_estab,
         cuitempre,
         cuitempre_btrim,
         documento_btrim,
         impue,
         mes_detadj,
         nro_insc_detadj,
         per_detadj,
         sec
 
________________________________________
 
QUERY:
 
# explain select w.*,b.nombre from (select nro_insc,cod_estab,cuitempre,impuesto,sum(monto_impo) as totret,sum(monto_rete) as suma_rete,tipodoc,documento from detadj where nro_insc=390009 and cod_estab=0 and ano=2003 and mes=4 and per=2 and sec=0 group by nro_insc,cod_estab,cuitempre,impuesto,tipodoc,documento) w LEFT OUTER JOIN retper b on (w.tipodoc=b.tipodoc and btrim(w.documento) like btrim(b.documento) and btrim(w.cuitempre) like btrim(b.cuitempre) );
 

RESULTS:
 
NOTICE:  QUERY PLAN:
 
Nested Loop  (cost=4999.30..21256.26 rows=1 width=220)
  ->  Subquery Scan w  (cost=4999.30..4999.34 rows=1 width=106)
        ->  Aggregate  (cost=4999.30..4999.34 rows=1 width=106)
              ->  Group  (cost=4999.30..4999.33 rows=2 width=106)
                    ->  Sort  (cost=4999.30..4999.30 rows=2 width=106)
                          ->  Index Scan using ano_mes_per on detadj  (cost=0.00..4999.29 rows=2 width=106)
->  Seq Scan on retper b  (cost=0.00..9821.23 rows=214523 width=96)
 
________________________________________
 
E. Caillava

pgsql-admin by date:

Previous
From: Robert Treat
Date:
Subject: Re: problems with moving a database
Next
From: Juan Miguel
Date:
Subject: Postgresql suddenly shutdown. Please help