pg.ddx.io  pgsql-sql@postgresql.org mailing list archive  
help / color / mirror / Atom feed
Surprising sequence scan when function call used
3+ messages / 3 participants
[nested] [flat]

* Surprising sequence scan when function call used
@ 2000-10-25 17:26  Will Fitzgerald <fitzgerald@inetmi.com>
  0 siblings, 2 replies; 3+ messages in thread

From: Will Fitzgerald @ 2000-10-25 17:26 UTC (permalink / raw)
  To: pgsql-sql

I have a table, login, which has a field by the same name; there's an index
on that field. I was surprised to discover that a SELECT which compares the
login field to a constant uses an Index scan, but if it is compared to a
function call--for example, lower()--a sequence scan is forced.

Any idea why?

Here are more details:


vdsq=> select version();
                               version
---------------------------------------------------------------------
 PostgreSQL 7.0.2 on i686-pc-linux-gnu, compiled by gcc egcs-2.91.66
(1 row)

vdsq=> \d login
                                Table "login"
  Attribute   |    Type     |                    Modifier
--------------+-------------+-----------------------------------------------
-
 id           | integer     | not null default nextval('login_id_seq'::text)
 login        | char(8)     | not null
 password     | char(8)     | not null
 jobid        | integer     | not null
 type         | smallint    | not null
 entryid      | integer     |
 lastactivity | timestamp   |
 trackid      | varchar(50) |
 roundid      | integer     |
Indices: idx_login_entryid,
         idx_login_jobid,
         idx_login_login,
         idx_login_password,
         idx_login_type,
         login_pkey

vdsq=> explain select * from login where login.login = 'foo';
NOTICE:  QUERY PLAN:

Index Scan using idx_login_login on login  (cost=0.00..582.61 rows=609
width=62)

EXPLAIN
vdsq=> explain select * from login where login.login = lower('foo');
NOTICE:  QUERY PLAN:

Seq Scan on login  (cost=0.00..1361.86 rows=609 width=62)

EXPLAIN




^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: Surprising sequence scan when function call used
@ 2000-10-26 03:22  Tom Lane <tgl@sss.pgh.pa.us>
  parent: Will Fitzgerald <fitzgerald@inetmi.com>
  1 sibling, 0 replies; 3+ messages in thread

From: Tom Lane @ 2000-10-26 03:22 UTC (permalink / raw)
  To: Will Fitzgerald <fitzgerald@inetmi.com>; +Cc: pgsql-sql

"Will Fitzgerald" <fitzgerald@inetmi.com> writes:
> vdsq=> explain select * from login where login.login = lower('foo');

> Seq Scan on login  (cost=0.00..1361.86 rows=609 width=62)

7.0 is a little bit stupid about cross-data-type comparisons (lower()
yields text, not char(n)).  This example works OK in current sources,
but until 7.1 comes out you'll need to write something like
	where login.login = lower('foo')::char;
Or change the login field to type text...

			regards, tom lane



^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: Surprising sequence scan when function call used
@ 2000-10-26 04:08  Hosokawa Tetsuichi <hosokawa@good-day.co.jp>
  parent: Will Fitzgerald <fitzgerald@inetmi.com>
  1 sibling, 0 replies; 3+ messages in thread

From: Hosokawa Tetsuichi @ 2000-10-26 04:08 UTC (permalink / raw)
  To: pgsql-sql

see
functional index

On Wed, 25 Oct 2000 13:26:51 -0400
"Will Fitzgerald" <fitzgerald@inetmi.com> wrote:

> I have a table, login, which has a field by the same name; there's an index
> on that field. I was surprised to discover that a SELECT which compares the
> login field to a constant uses an Index scan, but if it is compared to a
> function call--for example, lower()--a sequence scan is forced.
> 
> Any idea why?
--
* Hosokawa Tetsuichi
* hosokawa@good-day.co.jp



^ permalink  raw  reply  [nested|flat] 3+ messages in thread


end of thread, other threads:[~2000-10-26 04:08 UTC | newest]

Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2000-10-25 17:26 Surprising sequence scan when function call used Will Fitzgerald <fitzgerald@inetmi.com>
2000-10-26 03:22 ` Tom Lane <tgl@sss.pgh.pa.us>
2000-10-26 04:08 ` Hosokawa Tetsuichi <hosokawa@good-day.co.jp>

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