Re: Requirements for update languages?

From: Lauri Pietarinen <lauri.pietarinen_at_atbusiness.com>
Date: Fri, 15 Nov 2002 13:49:31 +0200
Message-ID: <3DD4DF4B.1C26A696_at_atbusiness.com>


Jan Hidders wrote:

> Lauri Pietarinen wrote:
> >> >
> >> > B1 x B2 = { ( x, n+m ) | (x, n) in B1, (x, m) in B2 }
> >>
> >> Ahem, that should of course be n.m, i.e., n times m, and not n+m. :-(
> >
> >Yes, it did look a bit strange...
> >
> >But just to clarify things, n and m are cardinalities, right?
> >
> >B1 = {(0,2 )} and B2 = {(1,1),(2,1)}
> >
> >if applied to the original example?
> >
> >But these are sets? So what do they have to do
> >with bags?
>
> They are bags, or, to be more precies, that's how bags are usually defined:
> functions that map some set to the set of natural numbers, and functions in
> turn are usually defined as binary relations that are functional and total.
> And, yes, that confirms Date's point.

One further observation came to mind:

With the bag-representation you just mentioned it is possible to remove the duplicates, or reduce the number of them just by "updating" the counter. This is not possible in SQL (without some internal rowid).

So would it be fair to say that SQL does not even really adher to a bag-model?

regards,
Lauri Pietarinen Received on Fri Nov 15 2002 - 12:49:31 CET

Original text of this message