Re: Postgresql 7.3.2 Crash

On Tue, 25 Mar 2003, Stephan Szabo wrote:

> Okay, I think I've localized the cause (but not a fix).
>
> >     name_id            bigint    not null default 0,
>
> I think the problem occurs with of the hack (mentioned in the last mail)
> because the default expression is of a different type.  I think it occurs
> specifically because the default expression is of a by value type and the
> real type is by reference, but I haven't gone through enough tests to be
> sure (it works if I make the default a bigint, a timestamp column with a
> timestamptz expression works but an abstime doesn't)
>
> Short term workaround is to make the default expression of the same type
> as the column rather than merely something that can be converted to
> that type.


Well, you're right, here is my workaround:
-- purpose: workaround the dumb value to bigint conversion of postgresql
:)
-- usage getmebigint(int);
drop FUNCTION getmebigint(int);
CREATE FUNCTION getmebigint(int) RETURNS bigint AS
   'DECLARE
        id bigint;
    BEGIN
        select $1 into id;
        RETURN id;
    END;'
   language 'plpgsql';


And in table definitions you use getmebigint(0) that makes the
transformation between value type and bigint type

        name_id      bigint not null default getmebigint(0),

Awful but is working until you'll find the problem.

Question: there isn't any cast operator like this?:
        name_id      bigint not null default bigint(0)


Anyway, thankz for the fast reply.

Regards,
Adrian Pop