Re: A problem in SQL (a real challenge)
Date: 2000/05/09
Message-ID: <UlUR4.237$C9.4096_at_news1.online.no>#1/1
Joe Celko <71062.1056_at_compuserve.com> wrote in message
news:8evc8e$otb$1_at_nnrp1.deja.com...
>
> >> Karl Hewlett suggested a smart solution for this a while ago. This
> solution utilizes the unique rowid of each row in a table: <<
>
> ROWID is both not standard, and not relational.
Not standard, but I suppose every vendor supports it. I'm only familiar with Oracle, however.
> You can use a CURSOR
> and stay within standards, but that also stinks.
I suggested such a solution the last time the question arised.
>You will have do this
> because the table is so screwed up by not having a PRIMARY KEY declared
> that all bets are off. Beat up the programmer to prevent this from
> happening again.
In general I agree, hence the statement "Do note that a table without a
primary key violates 1NF, hence should
generally be avoided." in my last reply. Still, heavy performance demands
could even make me violate some design-rules. All constraints imposed on a
database have an inherent performance penalty. I'm currently working with a
database designed by a major IP-telephony vendor with absolutely no
constraints.
-- ThomasReceived on Tue May 09 2000 - 00:00:00 CEST