Re: Logging (was Re: Suggestion GRANT ALTER, TRIGGER ON - Mailing list pgsql-hackers

From scott.marlowe
Subject Re: Logging (was Re: Suggestion GRANT ALTER, TRIGGER ON
Date
Msg-id Pine.LNX.4.33.0305210906010.22799-100000@css120.ihs.com
Whole thread Raw
In response to Re: Logging (was Re: Suggestion GRANT ALTER, TRIGGER ON  (Alvaro Herrera <alvherre@dcc.uchile.cl>)
List pgsql-hackers
On Tue, 20 May 2003, Alvaro Herrera wrote:

> On Tue, May 20, 2003 at 02:42:21PM -0600, scott.marlowe wrote:
> > On Tue, 20 May 2003, Tom Lane wrote:
> > 
> > > Josh Berkus <josh@agliodbs.com> writes:
> > > > I would propose
> > > > SELECT
> > > > INSERT/UPDATE
> > > > DELETE
> > > > DDL
> > > 
> > > Hm, why that particular division --- why separate DELETE but keep
> > > INSERT and UPDATE together?
> > 
> > Why not just use a regex?  Then you could log exactly what you're looking 
> > for.
> 
> I'd think a bitstring or some such is better... with a regex you are
> sure going to drive the performance down.

Sure, that's fine too.  I'd just prefer some way to "wild card" what is 
being logged.  I.e. I can look for specific SQL code and log just that.

Logging all selects doesn't really gain me much over grepping the log 
files, as there will still be tons of selects I'm not interested in on a 
production system.  Logging queries that contain specific keywords (i.e. 
table name, field name things like that) represents a much more useful 
tool to me.



pgsql-hackers by date:

Previous
From: Martijn van Oosterhout
Date:
Subject: Pruning useless tables for queries
Next
From: Tom Lane
Date:
Subject: Corner-case bugs in ThisStartUpID accounting