Re: Record Log Trigger - Mailing list pgsql-sql
From | Ramakrishnan Muralidharan |
---|---|
Subject | Re: Record Log Trigger |
Date | |
Msg-id | 02767D4600E59A4487233B23AEF5C5992A4084@blrmail1.aus.pervasive.com Whole thread Raw |
In response to | Record Log Trigger (lucas@presserv.org) |
List | pgsql-sql |
Hi, I do not think there is a performance issue due trigger on a larger database. For improving performance, it is better towrite triggers for Insert, update, delete separately rather than combining them on a single trigger. Periodic VACCUM and RE-INDEX the table will improve the performance. Regards, R.Muralidharan -----Original Message----- From: pgsql-sql-owner@postgresql.org [mailto:pgsql-sql-owner@postgresql.org]On Behalf Of lucas@presserv.org Sent: Tuesday, May 03, 2005 5:45 PM To: CHRIS HOOVER Cc: pgsql-sql@postgresql.org Subject: Re: [SQL] Record Log Trigger Well... Right, I will use to_timestamp() function instead of now() function. But, what is the performance for those Triggers??? Considering that all tables will have this Trigger and will check for each update or insert. Thanks Quoting CHRIS HOOVER <CHRIS.HOOVER@companiongroup.com>: > One change you might want to look at is not using the now() function. > According to the docs, the now() function always returns the start of the > transaction time. So, if your code is using transaction blocks, the time may > not be what you are expecting. > > This is what I had do to in my trigger to get the current clock time: > to_char(to_timestamp(timeofday(),\'Dy Mon DD HH24:MI:SS.US YYYY\') > > > HTH, > > Chris > ------------------( Forwarded letter 1 follows )--------------------- > Date: Mon, 02 May 2005 16:10:46 -0300 > To: pgsql-sql@postgresql.org.comp > From: lucas@presserv.org.comp > Sender: pgsql-sql-owner+m21335@postgresql.org.comp > Subject: [SQL] Record Log Trigger > > Hi all, > I am building a database in postgresql and I made a function that returns the > system time and the current user... like this: > > CREATE OR REPLACE FUNCTION generate_idx() returns text as > $$ > select to_char(now(),'YYYYMMDDHHMISSUSTZ')||CURRENT_USER; > $$ language 'SQL'; > > CREATE OR REPLACE FUNCTION TG_idxm() RETURNS trigger AS > $$ > BEGIN > NEW.idxm = generate_idx(); > RETURN NEW; > END; > $$ LANGUAGE plpgsql; > > And my all tables have the "idxm" field, its something like a log for the > record, to know Who and When the record have changed. I.e: > > CREATE TABLE products( > id serial primary key, > description varchar(50), > ... > idxm varchar(100) > ); > CREATE TRIGGER TG_products_idxm BEFORE INSERT or UPDATE on products > FOR EACH ROW > EXECUTE PROCEDURE TG_idxm(); > > Okay, it runs fine... but my question is: > Is it right??? In the future (when the database will be bigger with many of > millions records) this functions for each table will depreceate my database > performance??? > Is there any other way to build it??? > > Thank you > > ---------------------------(end of broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > ---------------------------(end of broadcast)--------------------------- TIP 2: you can get off all lists at once with the unregister command (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)