Re: Views for denomalizing
Date: Fri, 4 Feb 2005 09:27:56 -0500
Message-ID: <36hf3cF51qa31U1_at_individual.net>
"Alfredo Novoa" <alfredo_novoa_at_hotmail.com> wrote in message
news:jmu6011qbjoejhumkrlakd5nhscsvfe213_at_4ax.com...
> On Thu, 3 Feb 2005 23:18:29 -0500, "Alan" <not.me_at_rcn.com> wrote:
>
> >Normalization occurs at a logical level (note I did not say "the" logical
> >level
>
> Why not?
>
> >, as I am just distinguishing logical from physical, not getting into
> >logical vs conceptual- IOW, let's say logical and conceptual are the same
> >thing
>
> But they are very different.
I know. It is obvious that I know, as I explained why I was not making that distinction.
>
> >Views are a physical construction
>
> Very very wrong!
Not at all. Proof: Show me a view that has no underlying tables.
>
> Views are as logical as the rest of the relation variables.
>
> It seems that you are calling physical model to the logical model and
> logical model to the conceptual model. This is a very common confusion
> in the literature.
I see. The literature is wrong and you are right.
>
> This also explains your confusion between the conceptual and logical
> levels.
>
> > and have nothing to do
> >with normalization.
>
> I disagree. View values are relations like any other relation. We can
> check whether a view is in a given normal form, but it does not mean
> that we should normalize the views.
Almost. A view is made of normalized or non-normalized tables. It can be determined if the resulting view is normalized or not, but you can't normalize a view. Normalization takes place much earlier, using functional dependencies. A view is not constructed until several steps later. In fact, it would be at theend of the chain.
>
> >attributes
> >[apply FDs to get to the next level, revealing PKs in the process]:
> >relations (entities and relationships - an ERD is often created here)
> >[apply transformation rules* to the ERD to get to the next level]
> >-----boundry between logical and physical----------
>
> An ERD is a conceptual design, and if you apply "transformation rules"
> to the ERD what you get is an incomplete logical design.
Not in the least bit correct. Assuming the ERD itself is correct (a fair assumption), and you follow the rules, you ALWAYS end up with a database that is in at least 3NF.
>
> > Remember, a view is
> >nothing but a stored SQL statement
>
> A view is defined using a query statement, but a view is a relation or
> table variable.
>
>
> Regards
Received on Fri Feb 04 2005 - 15:27:56 CET