Message-ID: <408D2CE5.1090108@sysd.com>
Date: Mon, 26 Apr 2004 11:38:13 -0400
From: Rich Cullingford <rculling@sysd.com>
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4) Gecko/20030703
MIME-Version: 1.0
To: pgsql-interfaces@postgresql.org
Subject: Tantalizing reference to passing data from NEW variable in a trigger
	function...
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

All,
This question seems to have been asked a number of times on this list, 
recently (including by me) without being answered: how to pass the row 
data stored in the NEW variable from a trigger function into a 
sub-function. None of the obvious syntaxes work, including one by a 
respondent in this list that suggested assigning NEW to a record 
variable, and then passing *that*. After searching up and down the 
archives, I see the following tantalizing snippet (from 6/2003) in the 
context of a rule:

Gunter Diehl <gdid ( at ) gmx ( dot ) de> writes:
 > create funtion f1(v, v) returns void as '...'
 > create rule vupt as on update to v do instead select f1(new, old)
 > While accepting this definitions, pg says it can not handle "whole-tuple
 > references" at runtime.

FWIW, the "old" case works fine.  "new" doesn't work so well because the
new row hasn't been formed into a tuple at the point where the rule
runs; it only exists as a list of variables.  (This is a rather handwavy
explanation, but I think it captures the key point.)  While this could
probably be fixed with some effort, I doubt it's going to happen soon.
Is there any chance of doing what you want with a trigger instead of a
rule?

                                                     ^^^^^

If there's a page somewhere that says how to do this, I'd be extremely 
grateful for a pointer. The obvious maneuver of copying NEW's data 
field-by-field into a composite variable is rather unappetizing when the 
row is wide...
                           Rich Cullingford
                           rculling@sysd.com