Re: Stored fields ordered left to right

From: Bob Badour <bbadour_at_golden.net>
Date: Mon, 29 Dec 2003 13:51:16 -0500
Message-ID: <1pydnZ_Ro6ck6m2iRVn-gQ_at_golden.net>


"Marshall Spight" <mspight_at_dnai.com> wrote in message news:jnZHb.693968$Fm2.598344_at_attbi_s04...
> "Bob Badour" <bbadour_at_golden.net> wrote in message
news:Ws6dnWXlao4FKnKiRVn-tA_at_golden.net...
> > "Marshall Spight" <mspight_at_dnai.com> wrote in message
> > news:dKIHb.681736$Tr4.1704642_at_attbi_s03...
> > >
> > > I was speaking of logically distinguishing attributes. I don't
> > > see how the physical level is even relevant here.
> >
> > You don't see how logically distinguishing attributes by physical
position
> > violates physical independence and confuses logical and physical
issues?!?
>
> Again, I don't see the physical level being discussed here.

If implicit order is not physical, where does it come from?

> I don't
> see that position or order are necessarily physical; they can be
> logical. In this case, they are.

Nothing is more physical than position ie. location.

> > > > You are confusing an external physical representation with a logical
> > > > representation.
> > >
> > > Interesting distinction, but not one that I can follow without further
> > > information. Do you have a reference for further reading?
> >
> > Um, everything that has ever been written on logical data models and the
> > relational model in particular.
>
> Your citation lacks a certain hoped-for specificity.

As does your claimed lack of comprehension.

> > What exactly do you not understand? Do you
> > understand external vs. internal? Do you understand physical vs.
logical?
> > Actually, you don't have to answer the last question because it is clear
you
> > do not.
>
> I don't understand why you believe that order is necessarily physical.

The only logical orders are conventional collating sequences of domains. All other order is physical. It implies physical location whether absolute or relative. If not physical, where does the order come from?

> > > > > It doesn't affect the semantics of relations or relational
operators;
> > > > > it just affects how attributes are identified.
> > > >
> > > > Huh? Of course it affects the semantics if positional ordering has
> > meaning!
> > >
> > > Mumble. Operations like union, intersection, difference, are identical
> > > either way. Join needs some work, but it's not what I'd call a huge
issue.
> >
> > No, they are not identical. Consider the following:
> >
> > R1 = { { A=1, B=2 } }
> > and
> > R2 = { { B=2, A=1 } }
> >
> > What is R3 = R1 union R2?
> > What is R4 = R1 intersect R2?
> > What is R5 = R1 minus R2?
> >
> > If position matters, the answers are:
> > R3 = { { A=1, B=2 }, { A=2, B=1 } }
> > R4 = { }
> > R5 = { { A=1, B=2 } }
>
> That's not correct. If we are to do an apples-to-apples comparison
> of relations with named attributes vs. ordered attributes, the
> corresponding ordered relation would be:

If order has meaning, stick with the order I gave you for the operations.

> (I chose attribute A to map to first posititon, and attribute B to map to
> second position.)

It is not your choice. I already gave you the order.

> > If position does not matter, the answers are:
> > R3 = { { A=1, B=2 } }
> > R4 = { { A=1, B=2 } }
> > R5 = { }
>
> As I said, it is simply a question of how one identifies the attributes.

You argue for implicit meaning encoded in order. I correctly identified the attributes in my example, and I showed that if order has meaning the results differ. Order requires an additional step and greater user knowledge to achieve correct results, which means the operations differ.

Omitting the semantic identifiers for the attributes only confuses matters. Received on Mon Dec 29 2003 - 19:51:16 CET

Original text of this message