Thread: Problem with return type of function ??? (corrected)
Hello, (CORRECTED VERSION OF MY PREVIOUS EMAIL) I have a strange problem, because it worked in a fonction for a table, and now I created the same (?) function for another table and it doesn't work... The function is accepted but at runtime I get : ERREUR: wrong record type supplied in RETURN NEXT CONTEXTE : PL/pgSQL function "find_sn_live" line 26 at return next Does someone maybe knows what it could be ? This is (a part of) my function : > CREATE OR REPLACE FUNCTION rma.find_sn_live (varchar(30)) RETURNS SETOF rma.serial_number > AS $_$ > DECLARE > serialnumber ALIAS FOR $1; > row RECORD; > BEGIN > > FOR row IN > SELECT * FROM rma.serial_number WHERE sn=serialnumber > LOOP > RETURN NEXT row; > END LOOP; > > END; > $_$ > LANGUAGE plpgsql STRICT; Thanks a lot for any help ! Denis -- Sent via pgsql-sql mailing list (pgsql-sql@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-sql
Denis BUCHER <dbucherml@hsolutions.ch> writes:
> I have a strange problem, because it worked in a fonction for a table,
> and now I created the same (?) function for another table and it doesn't
> work...
> The function is accepted but at runtime I get :
> ERREUR: wrong record type supplied in RETURN NEXT
> CONTEXTE : PL/pgSQL function "find_sn_live" line 26 at return next
Does that table have any dropped columns? If you don't remember
whether you ever dropped any columns, a quick look into pg_attribute
will tell you:
select attname from pg_attribute where attrelid = 'rma.serial_number'::regclass;
plpgsql isn't tremendously good with rowtypes that contain dropped
columns. I believe this particular case is fixed in CVS HEAD, but the
patch was a bit invasive and won't get back-ported to existing releases.
The workaround is to drop and recreate the table without any dropped
columns.
regards, tom lane
Hello Tom,
Tom Lane a écrit :
> Denis BUCHER <dbucherml@hsolutions.ch> writes:
>> I have a strange problem, because it worked in a fonction for a table,
>> and now I created the same (?) function for another table and it doesn't
>> work...
>
>> The function is accepted but at runtime I get :
>
>> ERREUR: wrong record type supplied in RETURN NEXT
>> CONTEXTE : PL/pgSQL function "find_sn_live" line 26 at return next
>
> Does that table have any dropped columns? If you don't remember
> whether you ever dropped any columns, a quick look into pg_attribute
> will tell you:
> select attname from pg_attribute where attrelid = 'rma.serial_number'::regclass;
Yes !
attname
------------------------------tableoidcmaxxmaxcminxminctidsn_idsnno_clientno_art_bwsn_fc_date........pg.dropped.6................pg.dropped.7........desc_frsn_cm_dateno_factureno_commande
(17 lignes)
> plpgsql isn't tremendously good with rowtypes that contain dropped
> columns. I believe this particular case is fixed in CVS HEAD, but the
> patch was a bit invasive and won't get back-ported to existing releases.
> The workaround is to drop and recreate the table without any dropped
> columns.
OH, I see...
Thanks a lot, I will try this later today when nobody uses the
application...
Denis
Tom Lane wrote: > Denis BUCHER <dbucherml@hsolutions.ch> writes: > > Does that table have any dropped columns? If you don't remember > whether you ever dropped any columns, a quick look into pg_attribute > will tell you: > select attname from pg_attribute where attrelid = 'rma.serial_number'::regclass; > > plpgsql isn't tremendously good with rowtypes that contain dropped > columns. I thought that only applied to columns dropped after the function was defined. Live and learn. -- Richard Huxton Archonet Ltd
Hi Tom, Another question : Tom Lane a écrit : > Denis BUCHER <dbucherml@hsolutions.ch> writes: >> I have a strange problem, because it worked in a fonction for a table, >> and now I created the same (?) function for another table and it doesn't >> work... > >> The function is accepted but at runtime I get : > >> ERREUR: wrong record type supplied in RETURN NEXT >> CONTEXTE : PL/pgSQL function "find_sn_live" line 26 at return next > > Does that table have any dropped columns? If you don't remember > whether you ever dropped any columns, a quick look into pg_attribute > will tell you: > select attname from pg_attribute where attrelid = 'rma.serial_number'::regclass; > > plpgsql isn't tremendously good with rowtypes that contain dropped > columns. I believe this particular case is fixed in CVS HEAD, but the > patch was a bit invasive and won't get back-ported to existing releases. > The workaround is to drop and recreate the table without any dropped > columns. To do this it will be a little complicated because of table dependencies... And it could bug again at the next DROP COLUMN... Is there a way to change my function (RETURN SETOF part) to specify the column names/types ? Thanks a lot again Denis
Richard Huxton <dev@archonet.com> writes:
> Tom Lane wrote:
>> plpgsql isn't tremendously good with rowtypes that contain dropped
>> columns.
> I thought that only applied to columns dropped after the function was
> defined. Live and learn.
There are/were some variants that go away if you recreate the function
or start a fresh session (to clear the compiled-function cache). Not
all, unfortunately.
regards, tom lane
Denis BUCHER <dbucherml@hsolutions.ch> writes: > To do this it will be a little complicated because of table > dependencies... And it could bug again at the next DROP COLUMN... Is > there a way to change my function (RETURN SETOF part) to specify the > column names/types ? No, not really. You could maybe un-drop the columns with some manual surgery on pg_attribute, but it doesn't seem like that's going to lead to a nice solution. If you were really desperate you could try back-porting the patch: http://archives.postgresql.org/pgsql-committers/2009-08/msg00068.php but I wouldn't want to vouch for its safety, considering it hasn't been through a beta test cycle yet. regards, tom lane
Tom Lane a écrit : > Denis BUCHER <dbucherml@hsolutions.ch> writes: >> To do this it will be a little complicated because of table >> dependencies... And it could bug again at the next DROP COLUMN... Is >> there a way to change my function (RETURN SETOF part) to specify the >> column names/types ? > > No, not really. You could maybe un-drop the columns with some manual > surgery on pg_attribute, but it doesn't seem like that's going to lead > to a nice solution. > > If you were really desperate you could try back-porting the patch: > http://archives.postgresql.org/pgsql-committers/2009-08/msg00068.php > but I wouldn't want to vouch for its safety, considering it hasn't > been through a beta test cycle yet. Finally there was no real dependencies of that table, and I've found out that the "dependency" was my function ! Using SETOF (table) makes impossible to DROP the table. Therefore I DROP my function and was able to follow you advice, and it worked perfectly ! Thanks a lot for your help (as well as Richard's) Denis