# Re: repeating groups

From: Marshall Spight <marshall.spight_at_gmail.com>
Date: 18 Feb 2006 22:11:44 -0800

Jonathan Leffler wrote:
> mAsterdam wrote:
> > Marshall Spight wrote:
> >>>[...snippage...]
> >> Sure. But there are several ways out of the repeating groups problem
> >>
> >> 1) decomposing relations, aka "classical" 1NF
> >> 2) higher-than-1 cardinality attributes: lists or sets
> >
> > Aren't you jumping over the "order may have meaning" problem
> > here by taking these two together?
>
> It depends whether you take the OR in 'lists or sets' to be inclusive or
> exclusive, doesn't it? If it is inclusive, then lists handle the 'order
> has meaning' case and sets handle the 'order is not significant case'.
> But you've complicated the algebra - instead of just sets, you've now
> got to deal with lists and sets (and the alternatives you list below).

Yes, *exactly*!

> >> 3) Fully nested relations.
> >
> > lists-of-lists? sets-of-sets? lists-of-sets? sets-of-lists?
> >
> >> Actually, 2) was something I hadn't really thought of before:
> >> add *one* level of nesting. This idea is interesting but less
> >> appealing than 3, because it is less regular. I tend to prefer
> >> orthogonal designs; they have fewer arbitrary limitations.
>
> Was it Hoare who said "anything in computer science that isn't recursive
> isn't any good" or words to that effect? (I found two references at
> Google to Jim Gray saying "Anything in computer science that's not
> recursive is no good", but it is not an easy expression to search for.)

> 1-level of nesting would be arbitrary and would very quickly become
> restrictive. Indeed, you'd have to be very careful not to lose the
> closure property of the algebra.

Indeed.

Marshall Received on Sun Feb 19 2006 - 07:11:44 CET

Original text of this message