proposal: table functions and plpgsql

From: "Pavel Stehule" <pavel(dot)stehule(at)gmail(dot)com>
To: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: proposal: table functions and plpgsql
Date: 2008-05-21 16:12:27
Message-ID: 162867790805210912u2b33abbfqae76a7c4b8a4bf17@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hello

I am returning back to my patch and older proposal
http://archives.postgresql.org/pgsql-hackers/2007-02/msg00318.php .

Some work did Neil Conway
http://archives.postgresql.org/pgsql-hackers/2007-07/msg00501.php and
he commited half of this patch - RETURN QUERY part.

Problematic part of my patch is implementation. Tom Lane proposal
implenation RETURNS TABLE only as syntactic sugar for RETURNS SETOF
RECORD. This is not comaptible with potential implementation, because
it adds some default variables. My solution was special argmode, so I
was able don't create default variables for output. My solution wasn't
best too. It was ugly for current plpgsql where is often used RETURN
NEXT statement (PSM doesn't know similar statement). I unlike default
variables - it simply way to variables and column names collision.

I propose following syntax for plpgsql:

CREATE OR REPLACE FUNCTION foo(m integer)
RETURNS TABLE (a integer, b integer) AS $$
DECLARE r foo; -- same name as function, this type has local visibility
BEGIN
FOR i IN 1..m LOOP
r.a := i; r.b := i + 1;
RETURN NEXT r;
END LOOP;
RETURN;
END;
$$ LANGUAGE plpgsql;

In my proposal I don't create any default variables. Result type is
only virtual - I don't need write it to system directory. I thing it's
better than using some specific predeclared type as RESULTTYPE OR
RESULTSET.

What do you thing about?

Regards
Pavel Stehule

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message PFC 2008-05-21 16:18:27 Re: Posible planner improvement?
Previous Message Pavel Stehule 2008-05-21 15:45:32 plpgpsm future