Re: By The Dawn's Normal Light

"Dawn M. Wolthuis" <dwolt_at_tincat-group.comREMOVE> wrote in message news:clecle$p04$1_at_news.netins.net...
> "Marshall Spight" <mspight_at_dnai.com> wrote in message
> news:ZMxed.303809$D%.296516_at_attbi_s51...
>
> > Sometimes you have ordered data, and sometimes you have unordered
> > data. Which primitive operations you want depends on which one
> > you have. Each model handles one well and ignores the other.
> >
> > I propose that the ideal model would handle both, and have
> > relatively simple ways of transforming one into the other.
>
> I'm in.

Deal!

> > Lists are simpler and sometimes therefore the better choice.
> > Relations are more powerful, but that can mean more
> > complexity than you need.
>
> Functions are even more powerful -- you know you have a unique primary key
> that, effectively, maps to the remainder of the data for that, uh, record.
> I'm not sure what the current view of that is with relations, but I believe
> that each relation must have one or more (sets of) candidate keys. In that
> case, we can view them as functions as well, choosing one of the candidate
> key sets as the "from" and the remainder of the attributes as the "to".

I guess I'm not sure what "the current view" is either, but I would certainly insist that relations have at least one candidate key, and thus function is a subtype of relation.

I am reminded of a term Joshua Bloch uses: the "power to weight ratio." (Here "weight" is conceptual weight.) Lists, functions, and relations all do very well by this measure.

Marshall Received on Sun Oct 24 2004 - 06:54:35 CEST