On Thu, Oct 7, 2010 at 2:53 PM, Pavel Stehule <pavel.stehule@gmail.com> wrote:
> Hello,
>
> today I found a few bugs:
>
> a) parser allow a labels on invalid positions with strange runtime bug:
>
> postgres=# CREATE OR REPLACE FUNCTION foo()
> RETURNS void AS $$
> BEGIN
> FOR i IN 1..2
> <<<invalidLabel>>
> LOOP
> RAISE NOTICE '%',i;
> END LOOP;
> END;
> $$ LANGUAGE plpgsql;
> CREATE FUNCTION
>
> ERROR: column "invalidlabel" does not exist
> LINE 2: <<<invalidLabel>>
> ^
> QUERY: SELECT 2
> <<<invalidLabel>>
> CONTEXT: PL/pgSQL function "foo" line 3 at FOR with integer loop variable
> postgres=#
I'm not sure if I'd call that a bug, but it does look like a somewhat
odd error message, at least at first glance.
> b) SRF functions must not be finished by RETURN statement - I know, so
> there is outer default block, but it looks like inconsistency for SRF
> functions, because you can use a RETURN NEXT without RETURN. It maybe
> isn't bug - but I am filling it as inconsistency.
>
> postgres=# CREATE OR REPLACE FUNCTION fg(OUT i int)
> RETURNS SETOF int AS $$
> BEGIN
> FOR i IN 1..3
> LOOP fg.i := i;
> RETURN NEXT;
> END LOOP;
> END;
> $$ LANGUAGE plpgsql;
> CREATE FUNCTION
>
> postgres=# select fg();
> fg
> ----
> 1
> 2
> 3
> (3 rows)
I don't see what's wrong with this.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company