Re: 4 the FAQ: Are Commercial DBMS Truly Relational?

From: Marshall Spight <mspight_at_dnai.com>
Date: Sun, 10 Oct 2004 15:23:28 GMT
Message-ID: <Qtcad.366672$Fg5.150351_at_attbi_s53>


"Kenneth Downs" <firstinit.lastname_at_lastnameplusfam.net> wrote in message news:pbhbkc.5d7.ln_at_mercury.downsfam.net...
> Marshall Spight wrote:
>
> > "Christopher Browne" <cbbrowne_at_acm.org> wrote in message
> > news:2srmk1F1njk29U1_at_uni-berlin.de...
> >>
> >> If you cut columns off of the result set, it is possible for the
> >> result set to, in fact, not be a "set", but rather a non-unique "bag"
> >> of tuples.
> >> [...]
> >> That is NOT going to be a "set" or a "relation" if some customer made
> >> multiple purchases between those dates.
> >>
> >> The problem that this expresses is that the relational algebra does
> >> not satisfy the property of closure.
> >
> > That's an odd viewpoint.
> >
> > The math books I've read have made the point that when you
> > are talking about sets, then {2, 2} is the same set as {2}.
> > So if you have duplicates after a project, you just throw
> > them away; they "don't count" so to speak.
> >
>
> Therefore, wouldn't the implementation of SQL have some sort of implied
> DISTINCT on every SELECT statement?

We've switched from talking about the relational algebra to talking about SQL.

Yes, the relational algebra has an implied "distinct" on the result of every operation. No, SQL doesn't. This is one of the ways SQL isn't relational.

Marshall Received on Sun Oct 10 2004 - 17:23:28 CEST

Original text of this message