Re: Question about function body checking and 8.1

Alvaro Herrera wrote:
> On Tue, Mar 22, 2005 at 06:09:28PM +0100, Peter Eisentraut wrote:
> > tony_caduto(at)amsoftwaredesign(dot)com wrote:
> > > currently when you execute a CREATE OR REPLACE FUNCTION there is
> > > virtually no checking of the function body for variable type
> > > compatibility or if the SQL statements are valid.
> >
> > Only for very small values of "no". Please provide an example.
>
> Here's a simple one:

Well, mentioning that he was talking about PL/pgSQL would have
helped. :)

Implementing this type of syntax checker isn't hard since the code is
already there, but there might be a small, ugly problem. IIRC, the
parser and/or semantic analyzer of PL/pgSQL relies on knowing whether
the function is called as a trigger. You don't know that at creation
phase. So you'd need to relax the syntax checking in some ill-defined
ways and propagate that relaxation flag all over the place. But
besides that, this is really just a typing exercise for someone.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/