Thread: PERFORM statement inside procedure

Rajat Katyal wrote:
> Hi:
>
> In postgres documentation its written that if we execute query as
> PERFORM /query /inside our stored procedure/;/ *then the special
> variable FOUND is set to true if the query produced at least one row, or
> false if it produced no rows.*
> **
> But FOUND variable is always returning true even my query is
> returning *0 records.*

FOUND appears to work correctly in the hundreds of stored procedures I wrote
last month.  (At least, I haven't found any problems _yet_)

You might do well to post more details about your usage.  Best would be to
post the actual stored procedure you're having problems with, along with the
version of Postgres in use.

--
Bill Moran
Potential Technologies
http://www.potentialtech.com

Bill Moran <wmoran@potentialtech.com> writes:
> Rajat Katyal wrote:
>> But FOUND variable is always returning true even my query is
>> returning *0 records.*

> FOUND appears to work correctly in the hundreds of stored procedures I wrote
> last month.  (At least, I haven't found any problems _yet_)

Works for me too, in recent releases.  I believe PERFORM did not originally
set FOUND ... are you reading the documentation that goes with your server
version?

[ digs in CVS logs... ]  Here we go:

2002-06-24 19:12  tgl

    * src/pl/plpgsql/src/pl_exec.c: plpgsql's PERFORM statement now
    sets FOUND depending on whether any rows were returned by the
    performed query.  Per recent pgsql-general discussion.

So it should work in 7.3 or later.

            regards, tom lane

"Rajat Katyal" <rajatk@intelesoftech.com> writes:
> checkPKSql := ''select * from "transform_customer_billing" '';
> checkPKSql := checkPKSql || '' where "inv_no" = '' || quote_literal(new=
> ."inv_no");
> PERFORM checkPKSql;

You seem to be confusing PERFORM with EXECUTE.  They are very different.
The above PERFORM is really equivalent to
    SELECT 'select ...';
which naturally yields a row containing a not-very-useful string value.

            regards, tom lane

Actually my problem is PERFORM is not updating the FOUND variable to false
even when my query return no rows. Can you please tell me the better way to
use PERFORM so that by running my select query I just come to know whether
it returns 0 rows or not.

Regards,
Rajat.
----- Original Message -----
From: "Tom Lane" <tgl@sss.pgh.pa.us>
To: "Rajat Katyal" <rajatk@intelesoftech.com>
Cc: <pgsql-general@postgresql.org>
Sent: Wednesday, April 07, 2004 11:15 AM
Subject: Re: [GENERAL] PERFORM statement inside procedure


> "Rajat Katyal" <rajatk@intelesoftech.com> writes:
> > checkPKSql := ''select * from "transform_customer_billing" '';
> > checkPKSql := checkPKSql || '' where "inv_no" = '' || quote_literal(new=
> > ."inv_no");
> > PERFORM checkPKSql;
>
> You seem to be confusing PERFORM with EXECUTE.  They are very different.
> The above PERFORM is really equivalent to
> SELECT 'select ...';
> which naturally yields a row containing a not-very-useful string value.
>
> regards, tom lane
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster
>

On Wednesday 07 April 2004 08:30, Rajat Katyal wrote:
> Actually my problem is PERFORM is not updating the FOUND variable to false
> even when my query return no rows. Can you please tell me the better way to
> use PERFORM so that by running my select query I just come to know whether
> it returns 0 rows or not.

No, what Tom's saying is that PERFORM doesn't take a string - you are passing
it a string. Since perform is equivalent to SELECT that means you are doing
  SELECT 'SELECT true'
which returns one row, containing one column: 'SELECT true'

If you want to execute a string as a query, you use EXECUTE. Execute slows the
process down because it actually runs the query.

If you want to use PERFORM you'll need to write something like:

PERFORM SELECT true FROM transform_customer_billing WHERE inv_no = NEW.inv_no;
IF FOUND THEN
...

(The only reason I select true rather than * is so I don't think I'm going to
use the results from this query when I look at the code 12 months from now).

PS - that will slow the process down again. I'm not sure you want to be doing
what you're trying to do. Your original problem (a couple of weeks ago?)
seemed to be inserts/updates were half the speed if you checked for the
existence of the row yourself. Not surprising, you're doing two things. But,
you wanted to do this because you didn't know if you were updating or
inserting.

In cases like this, I prefer to remove the unknown. Where I need a customer to
have a balance total, I add a trigger to the customer table so that every
time a new customer is inserted, so is a zeroed row to the balance table.
Deny deletion of balance rows where its customer still exists and you can
safely issue updates all the time.

--
  Richard Huxton
  Archonet Ltd