pg.ddx.io  pgsql-sql@postgresql.org mailing list archive  
help / color / mirror / Atom feed
Help on indexes
2+ messages / 2 participants
[nested] [flat]

* Help on indexes
@ 2002-05-15 12:08 Stephane DEWITTE <stephane@smeso.fr>
  2002-05-15 15:19 ` Re: Help on indexes Josh Berkus <josh@agliodbs.com>
  0 siblings, 1 reply; 2+ messages in thread

From: Stephane DEWITTE @ 2002-05-15 12:08 UTC (permalink / raw)
  To: pgsql-sql

Hi everybody.

I have some little problem on index utilisation.

Here is the schema of my base :

-----------------------------------
                  Table "mclis"
   Column    |         Type          | Modifiers
-------------+-----------------------+-----------
 cli_cod     | numeric(9,0)          | not null
 cli_civ_cod | character varying(5)  | not null
 cli_nomu    | character varying(32) | not null
 cli_pren    | character varying(32) | not null
 cli_dnai    | date                  | not null
 cli_insee   | character varying(13) | not null
 cli_cle     | character varying(2)  | not null
 cli_mdp_def | character varying(6)  | not null
Indexes: cli_cod_mclis_key,
         cli_insee_mclis_key,
         cli_nomu_mclis_key
Primary key: mclis_pkey


                 Table "mdrs"
   Column   |         Type         | Modifiers
------------+----------------------+-----------
 dr_cod     | numeric(11,0)        | not null
 dr_typ     | character varying(1) | not null
 dr_dcre    | date                 | not null
 dr_dmaj    | date                 | not null
 dr_mnap    | numeric(18,6)        | not null
 dr_typdrg  | character varying(1) | not null
 dr_cli_cod | numeric(9,0)         |
 dr_dpai    | date                 |
 dr_mdr_cod | character varying(3) |
 dr_psa_cod | numeric(6,0)         |
Indexes: dr_cli_cod_mdrs_key,
         dr_cod_mdrs_key,
         dr_psa_cod_mdrs_key,
         idx_dr_dpai,
         idx_dr_mnap
Primary key: mdrs_pkey


                 Table "mdecs"
    Column    |         Type          | Modifiers
--------------+-----------------------+-----------
 dec_cod      | character varying(20) | not null
 dec_dcre     | date                  | not null
 dec_typedec  | character varying(1)  | not null
 dec_deccomp  | character varying(20) |
 dec_dsoins   | date                  | not null
 dec_dr_cod   | numeric(11,0)         | not null
 dec_mntsso   | numeric(12,6)         |
 etar_gmu_cod | character varying(6)  |
 dec_mntmut   | numeric(12,6)         |
 dec_cli_cod  | numeric(6,0)          |
Indexes: dec_cod_mdecs_key,
         dec_dr_cod_mdecs_key
Primary key: mdecs_pkey


                  Table "mldecs"
    Column     |         Type          | Modifiers
---------------+-----------------------+-----------
 ldec_dec_cod  | character varying(20) | not null
 ldec_numlig   | character varying(2)  | not null
 ldec_nexec    | character varying(8)  |
 ldec_act_clas | character varying(3)  | not null
 ldec_mntpaye  | numeric(12,6)         |
 ldec_pu       | numeric(12,6)         | not null
 ldec_txsso    | numeric(5,2)          | not null
 ldec_mntsso   | numeric(12,6)         | not null
 ldec_txmut    | numeric(5,2)          |
 ldec_mntmut   | numeric(12,6)         |
 ldec_mnttot   | numeric(12,6)         | not null
Indexes: idx_ldec_numlig, (on ldec_dec_cod and ldec_numlig)
         ldec_dec_cod_mldecs_key (on ldec_dec_cod)
Primary key: mldecs_pkey
-------------------------
When I make : (case 1)
explain select * from mldecs where ldec_dec_cod = 'x' I obtain :
NOTICE:  QUERY PLAN:
Index Scan using ldec_dec_cod_mldecs_key on mldecs  (cost=0.00..34449.16
rows=8792 width=155)
EXPLAIN

When i make (case 2)
explain select * from mldecs where ldec_dec_cod in  ('x','y') I obtain :
NOTICE:  QUERY PLAN:
Seq Scan on mldecs  (cost=0.00..63928.99 rows=17540 width=155)
EXPLAIN

When i make : (case 3)
explain select * from mclis,mdrs,mdecs,mldecs
where cli_cod = 147854
and dr_cli_cod = cli_cod
and dec_dr_cod = dr_cod
and ldec_dec_cod = dec_cod;

I obtain :
NOTICE:  QUERY PLAN:
Hash Join  (cost=418.82..64348.03 rows=18 width=393)
  ->  Seq Scan on mldecs  (cost=0.00..55136.99 rows=1758399 width=155)
  ->  Hash  (cost=418.79..418.79 rows=12 width=238)
        ->  Nested Loop  (cost=0.00..418.79 rows=12 width=238)
              ->  Nested Loop  (cost=0.00..123.31 rows=8 width=150)
                    ->  Index Scan using cli_cod_mclis_key on mclis
(cost=0.00..6.01 rows=1 width=73)
                    ->  Index Scan using dr_cli_cod_mdrs_key on mdrs
(cost=0.00..116.93 rows=30 width=77)
              ->  Index Scan using dec_dr_cod_mdecs_key on mdecs
(cost=0.00..34.97 rows=8 width=88)

EXPLAIN

I don't understanf why it makes a seq scan on table mldecs on cases 2 and 3,
according that mldecs(ldec_dec_cod) has an index (named
ldec_dec_cod_mldecs_key). The base is vacuumed and analyzed every night. How
can I use the ldec_dec_cod_mldecs_key index on mldecs ?

Regards.
Stephane.






^ permalink  raw  reply  [nested|flat] 2+ messages in thread

* Re: Help on indexes
  2002-05-15 12:08 Help on indexes Stephane DEWITTE <stephane@smeso.fr>
@ 2002-05-15 15:19 ` Josh Berkus <josh@agliodbs.com>
  0 siblings, 0 replies; 2+ messages in thread

From: Josh Berkus @ 2002-05-15 15:19 UTC (permalink / raw)
  To: Stephane DEWITTE <stephane@smeso.fr>; pgsql-sql

Stephane,

> I don't understanf why it makes a seq scan on table mldecs on cases 2
> and 3,
> according that mldecs(ldec_dec_cod) has an index (named
> ldec_dec_cod_mldecs_key). The base is vacuumed and analyzed every
> night. How
> can I use the ldec_dec_cod_mldecs_key index on mldecs ?

First question: Have you actually measured the execution time of the
various queries?  There are many cases where a Seq Scan is faster than
an Index Scan.  For example, if in Case2, 50% of the entries were
values 'X' of 'Y', then a Seq Scan is the better approach.

Please determine that you actually have a problem with slow execution.
 Otherwise, the query parser is just doing its job.

-Josh Berkus



^ permalink  raw  reply  [nested|flat] 2+ messages in thread


end of thread, other threads:[~2002-05-15 15:19 UTC | newest]

Thread overview: 2+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2002-05-15 12:08 Help on indexes Stephane DEWITTE <stephane@smeso.fr>
2002-05-15 15:19 ` Josh Berkus <josh@agliodbs.com>

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox