Re: Examples of SQL anomalies?
Date: Wed, 9 Jul 2008 23:41:28 -0700 (PDT)
On Jul 9, 1:31 pm, "David Cressey" <cresse..._at_verizon.net> wrote:
> "Marshall" <marshall.spi..._at_gmail.com> wrote in message
> > On Jul 7, 1:21 pm, JOG <j..._at_cs.nott.ac.uk> wrote:
> > > On Jul 5, 1:09 pm, "David Cressey" <cresse..._at_verizon.net> wrote:
> > > > "Marshall" <marshall.spi..._at_gmail.com> wrote in message
> > > >news:6ad51b62-e66a-4daa-b21c-c361fd6b22f8_at_8g2000hse.googlegroups.com...
> > > > > What can be meaningfully asked is determined by the schema.
> > > > > If the schema specifies that the weight attribute is nullable,
> > > > > then the question of how much a shipment weighs in total
> > > > > is a question that cannot be asked.
> > > > > Marshall
> > > > Bingo!
> > > However, one /can/ validly ask "please give me the minimum the
> > > shipment weighs" and this may still be very useful.
> > Well, that's assuming that all the shipments have a positive
> > weight. What if we might ship helium baloons? Then you
> > can't set the minimum.
> > OK I was just having some fun there.
> > > This is of course
> > > not a defence of null markers (obviously not. its me), but rather just
> > > a precaution against ruling out all questions of irregular data in
> > > blanket fashion. Our aim should be to provide frameworks that allows
> > > us to ask these questions with syntactic correctness /and/ as
> > > parsimoniously as possible.
> > Sure.
> > My idea of the phrasing of the sum() over a nullable weight
> > is "what is the sum of the weights of the items in the shipment
> > for which the weight has been entered into the db?"
> That is, IIUYC , the sum of all the wieghts, as far as the db knows them,
> This amounts to the "open world assumption" with regard to weights of items.
> If one is going to adopt a "closed world assumption" with regard to weights
> of entered items, then it would seem to me that the weights column would
> have to be declared not nullable. Meaning that, if an item is entered nto
> a row, the weight column may not be left null.
> In the case of not nullable columns, this entire subthread is moot.
> Brian seems to be willing to bounce back and forth between OWA and CWA,
> without making any entries in the schema to reflect which assumption is
> operative. Or maybe it's just a disconnect between Brian and SQL.
I think I agree with all of that. (But I am not so confident that I
what does or does not constitute CWA/OWA under a system with NULLs.)
Marshall Received on Thu Jul 10 2008 - 01:41:28 CDT