Re: [HACKERS] problems with dropped columns

Christopher Kings-Lynne wrote:
> I want to fix this bug, however I can't see how the example below is
> failing... (Obeys dropped columns) I'm not up with my SRFs, so would
> someone be able to post a concise SQL script that demonstrates the failure?
>
> I can see in the code that it should be failing, but I need a demonstrated
> example...

Here is a self contained example using cvs tip:

CREATE TABLE fk_test (f1 int, f2 int);
insert into fk_test(f1, f2) values(1, 21);
insert into fk_test(f1, f2) values(2, 22);

ALTER TABLE fk_test DROP COLUMN f2;
ALTER TABLE fk_test ADD COLUMN f3 int;

insert into fk_test(f1, f3) values(3, 33);
insert into fk_test(f1, f3) values(4, 34);

regression=# select * from fk_test ;
f1 | f3
----+----
1 |
2 |
3 | 33
4 | 34
(4 rows)

CREATE OR REPLACE FUNCTION test()
RETURNS SETOF fk_test AS '
DECLARE
rec fk_test%ROWTYPE;
BEGIN
FOR rec IN SELECT * FROM fk_test LOOP
RETURN NEXT rec;
END LOOP;
RETURN;
END;
' LANGUAGE 'plpgsql';

regression=# select * from test();
f1 | f3
----+----
1 |
2 |
3 |
4 |
(4 rows)

Joe