db-torque-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Patryk Laurent <pat...@snt-inc.com>
Subject UPDATEs taking a long time
Date Fri, 24 Jan 2003 13:51:39 GMT
Greetings,

I have discovered that UPDATEs in PostgreSQL of the following structure

UPDATE mytable SET mycolumn=# where id=#

uses Sequential scans (try EXPLAIN, you get Seq Scan) when id is an
indexed primary key (bigint) generated by a sequence.  It is very slow
for large tables.

(To do the UPDATE in Torque, I was just taking the old object, oldRow,
and doing oldRow.setmycolumn(newValue); oldRow.save(); )

To get an Index Scan, it turns out you must enclose the ID number in
single quotes in your query, like this:

UPDATE mytable SET mycolumn=# where id='#'

Are any of the Torque developers aware of this bug/feature?  Torque is
clearly using the sequential scan (it was taking me about 5 seconds to
update a single row in a 3,000,000 row table, just like when I didn't
use quotes.)

Any feedback on this would be much appreciated.  Thanks,

Patryk

Senior Programmer
SUPERnatural tools, Inc.


Mime
View raw message