Re: Detection of which attributes should get set in update trigger - Mailing list pgsql-general

From Thiemo Kellner
Subject Re: Detection of which attributes should get set in update trigger
Date
Msg-id e4bc5896-86f9-44a2-8a97-9aa2324bc84c@gelassene-pferde.biz
Whole thread Raw
In response to Re: Detection of which attributes should get set in update trigger  ("David G. Johnston" <david.g.johnston@gmail.com>)
List pgsql-general

Thanks for the reply. I confirm the behaviour. Your explanation makes sense if I consider having read that with an update the is a complete new record version written.

Am 10.11.2023 um 14:35 schrieb David G. Johnston:
On Friday, November 10, 2023, Thiemo Kellner <thiemo@gelassene-pferde.biz> wrote:
Hi all

I have a view, that is a join over 4 tables (the setup of the rule question). I want to update the central table over the view (by an instead-of trigger). How can I determine, whether an attribute should get set to NULL, "new.XYZ is null" or whether it should be left alone. Is there a leave-me-alone function determining, whether an attribute is in the set clause of an update statement. I am thinking along the line code in the trigger function like

update TABLE
   set XYZ = case
                  when leave-me-alone (new.XYZ) then old.XYZ
                  else new.XYZ
             end;

And what about the where condition... hm, guess I am a bit confused.

IIUC, The NEW record is complete, with every value as it should be in the newly saved tuple.  There is no leave-me-alone concept.  You don’t get to know if the column was specified or not in the update command.

David J.

pgsql-general by date:

Previous
From: "David G. Johnston"
Date:
Subject: Re: Detection of which attributes should get set in update trigger
Next
From: pf@pfortin.com
Date:
Subject: Aren't regex_*() functions built-in?