Message-ID: <412CA7ED.4020704@Yahoo.com>
Date: Wed, 25 Aug 2004 10:53:33 -0400
From: Jan Wieck <JanWieck@Yahoo.com>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US;
	rv:1.7.1) Gecko/20040707
MIME-Version: 1.0
To: "Pedro B." <pedro.borracha@netcabo.pt>
Cc: pgsql-sql@postgresql.org
Subject: Re: PQexec and SPI_exec
References: <412CA074.2000606@netcabo.pt>
In-Reply-To: <412CA074.2000606@netcabo.pt>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

On 8/25/2004 10:21 AM, Pedro B. wrote:

> Hello everyone.
> 
> I'm experiencing some doubts regarding a procedure i have (.c compiled 
> as .so) running as an 'after insert for each row' trigger.
> 
> This trigger is supposed to do a simple query, something like
> SELECT * FROM table order by id where processed=0 limit 1
> 
> It's not the perfect way to get the vars of the insert itself, but the 
> result is a set of 45 columns, and the operations of the trigger are 
> somehow complex, so .c is really a necessity on this one, and as long as 
> this select actually returns the proper values, i can deal with it later.
> 
> But my problem is not one of a structure nature: my problem is the fact 
> that using SPI_exec, and the SPI_getvalue(SPI_tuptable->vals[0], 
> SPI_tuptable->tupdesc, columnX), and 
> DatumGetInt32(DirectFunctionCall1(int4in, 
> CStringGetDatum(SPI_getvalue(SPI_tuptable->vals[0], 
> SPI_tuptable->tupdesc, columnX)))), etc, it all works fine.
> the SPI-running-from-the-triggered-.so can detect the correct values - 
> from the insert that triggered it

If the trigger is supposed to use values from the tuple that triggered 
its call, then it should not select them from the table but use the 
provided tg_trigtuple in the TriggerData structure.

See http://www.postgresql.org/docs/current/static/trigger-interface.html 
for details.

> 
> I would prefer to use the more friendly PQexec and the simpler 
> PQgetvalue(res,0,X), but this last approach does not return the values 
> of the insert that triggered it. It returns the values from the "the 
> last insert before this one". What is the proper way to make this method 
> work?

There is no proper way to make this work at all. What you are doing in 
this case is to have the database server process that is handling your 
query open another client connection to the server, starting another 
database server process. This cannot work.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck@Yahoo.com #