=?iso-8859-1?Q?g=FCnter_strubinsky?= <[EMAIL PROTECTED]> writes:
> Sorry, wrong copy!
Okay, looking more closely, you've got two problems here:
> CREATE OR REPLACE FUNCTION public.norm(int8, record)
> RETURNS int8 AS
> ...
> LANGUAGE 'plpgsql' VOLATILE;
plpgsql functions don't support inputs declared as type "record"; you
have to use a more specific type name. norm() would have failed at
runtime had you ever gotten that far.
> declare
> currec record;
> ...
> select norm(catrec.cat1,currec) into dmy;
plpgsql also doesn't support passing whole record variables into SQL
commands. The error message about this is unhelpful :-(. (I just
improved the message in CVS tip, but actually making it work seems like
a much larger project...)
In this particular case I think you can end-run around it by not
importing the row into plpgsql at all. Can't you combine
select * into currec from denorm where theKey=catrec.cat1;
select norm(catrec.cat1,currec) into dmy;
into
select norm(catrec.cat1,denorm.*) into dmy
from denorm where theKey=catrec.cat1;
This should work if norm()'s second argument is declared as type
"denorm" (equivalent to denorm%rowtype).
If you don't want to tie norm() to the denorm row type, you'll need to
pass in the values it needs as separate scalars.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match
