Re: Extending my question. Was: The relational model and relational algebra - why did SQL become the industry standard?
Date: Tue, 18 Feb 2003 15:01:29 -0800
Message-ID: <D0z4a.21$h_3.106_at_news.oracle.com>
"Alfredo Novoa" <alfredo_at_ncs.es> wrote in message
news:e4330f45.0302181408.121a939f_at_posting.google.com...
> > Why? View updates is about ability to solve equations with relational
> > operators. While one can certainly claim that Set Algebra is superior
to
> > Bag Algebra, you have to demonstrate what exact technical difficulties
are
> > solving equations in the Bag Algebra.
>
> The systematic approaches to the view update problem based on
> predicate logic does not work with bags.
I have a doubt about Date's approach. Is he treating each operation individually; independently of each other? Is he treating each new tuple inserted into a relation individually; unrelated to other tuples inserted (in the same transaction)?
Returning to the example
select id, 'VOICE' type, voice phone
from contact
select id, 'FAX' type, fax phone
from contact
we can easily see that both assumtions aren't correct:
- Sucessively applied operations can't be considered independently of each other. In the above example union can't be considered independently of the antiprojection (BTW, what is the correct term for adding a pseudocloumn?).
- Two tuples inserted together into a derived relation correspond to a single tuple insertion in the base relation. The approach based upon "single tuple" translation is flawed.