pg.ddx.io pgsql-sql@postgresql.org mailing list archive
help / color / mirror / Atom feedUsing SETOF in plpgsql function
8+ messages / 6 participants
[nested] [flat]
* Using SETOF in plpgsql function
@ 2000-08-23 10:07 hlefebvre <hlefebvre@lexbase.net>
0 siblings, 2 replies; 8+ messages in thread
From: hlefebvre @ 2000-08-23 10:07 UTC (permalink / raw)
To: pgsql-sql
Hello,
I'd like to return a set of integer in an pl/pgsql function. How can I
do that ?
I've tried things like that, put I've an error when executing :
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
DECLARE ID INTEGER;
BEGIN
select a into id from foo;
return ID ;
END;
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
DECLARE ID setof INTEGER;
BEGIN
select a into id from foo;
return ID ;
END;
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
BEGIN
select a into id from foo;
return select a from foo;
END;
Any help is welcomed.
Thanks.
^ permalink raw reply [nested|flat] 8+ messages in thread
* RE: Using SETOF in plpgsql function
@ 2000-08-23 12:15 Graham Vickrage <graham@digitalplanit.com>
parent: hlefebvre <hlefebvre@lexbase.net>
1 sibling, 1 reply; 8+ messages in thread
From: Graham Vickrage @ 2000-08-23 12:15 UTC (permalink / raw)
To: hlefebvre <hlefebvre@lexbase.net>; pgsql-sql
As far as i know, you can only return single values from functions at the
moment.
Regards
Graham
-----Original Message-----
From: pgsql-sql-owner@hub.org [mailto:pgsql-sql-owner@hub.org]On Behalf
Of hlefebvre
Sent: 23 August 2000 11:08
To: pgsql-sql@postgresql.org
Subject: [SQL] Using SETOF in plpgsql function
Hello,
I'd like to return a set of integer in an pl/pgsql function. How can I
do that ?
I've tried things like that, put I've an error when executing :
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
DECLARE ID INTEGER;
BEGIN
select a into id from foo;
return ID ;
END;
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
DECLARE ID setof INTEGER;
BEGIN
select a into id from foo;
return ID ;
END;
CREATE FUNCTION SP_UPLES() RETURNS setof INTEGER AS '
BEGIN
select a into id from foo;
return select a from foo;
END;
Any help is welcomed.
Thanks.
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 12:35 hlefebvre <hlefebvre@lexbase.net>
parent: Graham Vickrage <graham@digitalplanit.com>
0 siblings, 0 replies; 8+ messages in thread
From: hlefebvre @ 2000-08-23 12:35 UTC (permalink / raw)
To: Graham Vickrage <graham@digitalplanit.com>; +Cc: pgsql-sql
Graham Vickrage wrote:
>
> As far as i know, you can only return single values from functions at the
> moment.
>
> Regards
>
> Graham
Hum, this is possible a least in SQL functions.
But maybe impossible in PL/PGSQL
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 13:46 Andreas Tille <tillea@rki.de>
parent: Jan Wieck <janwieck@Yahoo.com>
2 siblings, 0 replies; 8+ messages in thread
From: Andreas Tille @ 2000-08-23 13:46 UTC (permalink / raw)
To: ; +Cc: pgsql-sql
On Wed, 23 Aug 2000, Jan Wieck wrote:
> You can't. Not with PL/pgSQL nor with any other PL or C. The
> problem is nested deeper and requires the planned querytree
> redesign to get solved.
>
> Before you ask: The mentioned redesign will NOT be done for
> 7.1, and I'm not sure if we will be able to do it for 7.2
> yet.
Just to make sure you've got to know that this is a feature needed
by many users: Add me to the list of users who have a big need for
this!
Kind regards
Andreas.
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 13:48 Karel Zak <zakkr@zf.jcu.cz>
parent: Jan Wieck <janwieck@Yahoo.com>
2 siblings, 0 replies; 8+ messages in thread
From: Karel Zak @ 2000-08-23 13:48 UTC (permalink / raw)
To: Jan Wieck <janwieck@Yahoo.com>; +Cc: hlefebvre <hlefebvre@lexbase.net>; pgsql-sql
On Wed, 23 Aug 2000, Jan Wieck wrote:
> Before you ask: The mentioned redesign will NOT be done for
> 7.1, and I'm not sure if we will be able to do it for 7.2
> yet.
I hope that 7.2 :-), my query/plan cache is still outside current
interest and if core developers not will work on something like query
path redesign, the query/plan cache will still out.. :-(
Or already test query/plan cache anyone? IMHO it is good merge-able
to current source too.
Karel
PS. sorry of my small sigh for this.... :-)
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 14:24 Tom Lane <tgl@sss.pgh.pa.us>
parent: Jan Wieck <janwieck@Yahoo.com>
2 siblings, 1 reply; 8+ messages in thread
From: Tom Lane @ 2000-08-23 14:24 UTC (permalink / raw)
To: Jan Wieck <janwieck@Yahoo.com>; +Cc: hlefebvre <hlefebvre@lexbase.net>; pgsql-sql
Jan Wieck <janwieck@Yahoo.com> writes:
> hlefebvre wrote:
>> I'd like to return a set of integer in an pl/pgsql function. How can I
>> do that ?
> You can't. Not with PL/pgSQL nor with any other PL or C. The
> problem is nested deeper and requires the planned querytree
> redesign to get solved.
Not really. Coincidentally enough, I am just in the middle of removing
execQual.c's hard-wired assumption that only SQL-language functions
can return sets. (This is a side effect of fixing the function manager
so that SQL functions can be called in all contexts, eg used as index
functions.) If you want to fix plpgsql so that it retains state and
can produce multiple elements of a set over repeated calls, the same
way that SQL functions do, then it could be done today.
We may well want to rip out that whole approach to set functions later
when we redo querytrees, but the real limitation so far has been bogus
assumptions in the function-call API, not querytrees.
regards, tom lane
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 14:27 Jan Wieck <janwieck@Yahoo.com>
parent: hlefebvre <hlefebvre@lexbase.net>
1 sibling, 3 replies; 8+ messages in thread
From: Jan Wieck @ 2000-08-23 14:27 UTC (permalink / raw)
To: hlefebvre <hlefebvre@lexbase.net>; +Cc: pgsql-sql
hlefebvre wrote:
> Hello,
>
> I'd like to return a set of integer in an pl/pgsql function. How can I
> do that ?
You can't. Not with PL/pgSQL nor with any other PL or C. The
problem is nested deeper and requires the planned querytree
redesign to get solved.
Before you ask: The mentioned redesign will NOT be done for
7.1, and I'm not sure if we will be able to do it for 7.2
yet.
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #
^ permalink raw reply [nested|flat] 8+ messages in thread
* Re: Using SETOF in plpgsql function
@ 2000-08-23 16:53 Jan Wieck <janwieck@Yahoo.com>
parent: Tom Lane <tgl@sss.pgh.pa.us>
0 siblings, 0 replies; 8+ messages in thread
From: Jan Wieck @ 2000-08-23 16:53 UTC (permalink / raw)
To: Tom Lane <tgl@sss.pgh.pa.us>; +Cc: Jan Wieck <janwieck@Yahoo.com>; hlefebvre <hlefebvre@lexbase.net>; pgsql-sql
Tom Lane wrote:
> Jan Wieck <janwieck@Yahoo.com> writes:
> > hlefebvre wrote:
> >> I'd like to return a set of integer in an pl/pgsql function. How can I
> >> do that ?
>
> > You can't. Not with PL/pgSQL nor with any other PL or C. The
> > problem is nested deeper and requires the planned querytree
> > redesign to get solved.
>
> Not really. Coincidentally enough, I am just in the middle of removing
> execQual.c's hard-wired assumption that only SQL-language functions
> can return sets. (This is a side effect of fixing the function manager
> so that SQL functions can be called in all contexts, eg used as index
> functions.) If you want to fix plpgsql so that it retains state and
> can produce multiple elements of a set over repeated calls, the same
> way that SQL functions do, then it could be done today.
Not that easy. PL/pgSQL isn't a state machine. The
precompiled code is kind of a nested tree of statements. A
RETURN causes a controlled return() through all nested levels
of the PL executors C calls. This might close SPI calls in
execution as well. Imagine a code construct like
FOR rec IN SELECT * FROM customer LOOP
RETURN rec.cust_id AND RESUME;
END LOOP;
which would be the correct syntax for returning sets. What
happens in PL/pgSQL while execution is, that at the beginning
of the loop the SPI query for SELECT is performed, and then
the loop executed for all rows in the SPI result set. And of
course, you can have nested loops, why not.
Now you want to return the first value. If you really return
to the fmgr at this time, the connection to the SPI manager
must be closed, loosing the result set. So how to continue
later?
If we want to make it now for sets of scalar values (not
tuple sets), we could add another feature to the fmgr and the
PL handlers, which we need later anyway.
In the case of a call to a PL or C function returning a set,
the fmgr creates a temp table and calls the function which
fills the temp table with all the return values. Now fmgr
changes the execution trees func node in a way that it is
operating like an SQL function - holding a seqscan over the
temp table. After the last result is returned, the temp table
is removed. This'd work for tuple sets as well (so the temp
table then is our tuple-source).
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #
^ permalink raw reply [nested|flat] 8+ messages in thread
end of thread, other threads:[~2000-08-23 16:53 UTC | newest]
Thread overview: 8+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2000-08-23 10:07 Using SETOF in plpgsql function hlefebvre <hlefebvre@lexbase.net>
2000-08-23 12:15 ` Graham Vickrage <graham@digitalplanit.com>
2000-08-23 12:35 ` hlefebvre <hlefebvre@lexbase.net>
2000-08-23 14:27 ` Jan Wieck <janwieck@Yahoo.com>
2000-08-23 13:46 ` Andreas Tille <tillea@rki.de>
2000-08-23 13:48 ` Karel Zak <zakkr@zf.jcu.cz>
2000-08-23 14:24 ` Tom Lane <tgl@sss.pgh.pa.us>
2000-08-23 16:53 ` Jan Wieck <janwieck@Yahoo.com>
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox