Re: Basic question?What 's the key if there 's no FD(Functional Dependencies)?

From: vldm10 <vldm10_at_yahoo.com>
Date: 2 Nov 2006 13:47:40 -0800
Message-ID: <1162504060.305190.100710@h54g2000cwb.googlegroups.com>

paul c wrote:
> David Cressey wrote:
> > "paul c" <toledobythesea_at_dbms.yuc> wrote in message
> > news:a2c2h.242940$R63.209531_at_pd7urf1no...
> >> vldm10 wrote:
> >>> saturnlee_at_yahoo.com wrote:
> >>>> What 's the key for it? ABC or nothing???
> >>>
> >>> ABC is not the key.
> >>> Example: Let one partricular entity has A,B,C atributes
> >>> and let these atributes take the following values:
> >>>
> >>> A B C
> >>> -----------------------------
> >>> 2 4 6
> >>> 8 4 6
> >>> 2 4 6
> >>>
> >>> ( ABC can be the key only in the trivial cases i.e if an entity has
> >>> the atributes whose values never change)
> ...
> >> 2) I don't know why entities need to be mentioned, either, nor what a
> >> non-"trivial" entity might possibly be.
> >>
> >> p
> >
> > I think he was referring to "trivial functional dependencies". A key
> > determines any subset of itself, trivially. In current parlance, "well,
> > duh!"

>

> I still don't get what entities have to do with FD's. I thought Codd
> came up with functional dependencies for relations, not entities and
> that it is dangerous to mix those terms up, whatever we might think an
> entity is, once we've made a relation to deal with it, we should suspend
> the term as it can lead to all kinds of subjective confusions and just
> talk about tuples or predicates (when people start talking about the
> "real world", for me it's usually a clue that they are about to lapse
> into mysticism!).

>
> p

Let me clarify this more.
We have the real world and the RM.
Simply speaking we have the following schema:

Entities,attributes,.. ---------> Relations, columns,... (*)

If you thing that (*) is a simple thing and somehow automatic, I don't. It is very complex correspondence.

1)
Now if we put on left side of (*) the entity from my example with its three states and try to apply given relation R(A,B,C) it doesn't work.
Here we assume that ABC is the "natural" key.

2)
If you try to apply just the last two states from my example into given relation R(A,B,C), that is case when values are not repeated, you will get something like:

R:          A   B   C
           ------------
            8   4   6
            2   4   6

If you have millions of the rows in this relation, I don't know how you
will determine what keys determine same entity (with different states).

3)
I wrote about this earlier:
Honda store received group of new Honda Civic, all of them with same attributes
and we concluded that in this case "natural" key doesn't work.

4)
As a theoretical question we can consider case when a relation has any finite number of the attributes. This can be very big number - 100 maybe 200.
What will be the "natural" key for the n-m relationship between these entities?
What if these attributes start to change there values? How many dates will be involved in the key?

Vladimir Odrljin Received on Thu Nov 02 2006 - 15:47:40 CST