bug in plpgsql??? - Mailing list pgsql-bugs

From Alexander Zagrebin
Subject bug in plpgsql???
Date
Msg-id GPEFIGKNJAIDINJJPGOOAEMCCJAA.alexz@visp.ru
Whole thread Raw
Responses Re: bug in plpgsql???  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-bugs
Hi!

I try to work with trees in Postgres 7.1.1
My test script is
====================================================================
CREATE TABLE hierarchy

    parent      INTEGER,
    child       INTEGER
);

INSERT INTO hierarchy VALUES (1, 2);
INSERT INTO hierarchy VALUES (2, 4);
INSERT INTO hierarchy VALUES (3, 5);
INSERT INTO hierarchy VALUES (4, 5);

CREATE FUNCTION is_child(INTEGER, INTEGER) RETURNS BOOLEAN AS
'
DECLARE
-- Don't works!!!
    rec hierarchy%ROWTYPE;
-- Works!!!
    rec RECORD;
BEGIN
    RAISE DEBUG ''child is %, parent is %'', $1, $2;
    FOR rec IN SELECT * FROM hierarchy WHERE child = $1 LOOP
        RAISE DEBUG ''found parent %'', rec.parent;
        IF rec.parent = $2 OR is_child(rec.parent, $2) THEN
            RETURN TRUE;
        END IF;
    END LOOP;
    RETURN FALSE;
END;
'
LANGUAGE 'plpgsql';

SELECT is_child(5, 1);
====================================================================
If I declare "rec" as "RECORD", then all works fine:

2001-05-22 16:22:01 DEBUG:  child is 5, parent is 1
2001-05-22 16:22:01 DEBUG:  found parent 3
2001-05-22 16:22:01 DEBUG:  child is 3, parent is 1
2001-05-22 16:22:01 DEBUG:  found parent 4
2001-05-22 16:22:01 DEBUG:  child is 4, parent is 1
2001-05-22 16:22:01 DEBUG:  found parent 2
2001-05-22 16:22:01 DEBUG:  child is 2, parent is 1
2001-05-22 16:22:01 DEBUG:  found parent 1

But if "rec" is declared as "hierarchy%ROWTYPE", then
execution fails with such log messages:

2001-05-22 16:11:21 DEBUG:  child is 5, parent is 10
2001-05-22 16:11:21 DEBUG:  found parent 3
2001-05-22 16:11:21 DEBUG:  child is 3, parent is 10
Server process (pid 62170) exited with status 139 at Tue May 22 16:11:21
2001
Terminating any active server processes...
Server processes were terminated at Tue May 22 16:11:21 2001
Reinitializing shared memory and semaphores
....

Is this bug or feature?

Alexander Zagrebin
--

pgsql-bugs by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: psql dumps core when entering invalid backslash command
Next
From: Ondrej Palkovsky
Date:
Subject: Failing backend on plpgsql function