Re: How will schemas be affected by nested relations?
Date: Tue, 25 Jul 2006 01:56:18 GMT
Message-ID: <6Texg.227758$Mn5.783_at_pd7tw3no>
JOG wrote:
> Marshall wrote:
>> How will schema design be affected by having the ability to use >> nested relations? I have an intuition that it might not be that >> much; that nesting is a *little* useful but not all that *much* useful. >> However, I am concerned that I don't have a normal form to >> inform design choices. >> >> When should relation schemas be nested?
>
> Does Darwen's GUNF from the TTM email list cover this marshall?
>
> To quote Jon Heggland in:
> http://groups.google.com/group/comp.databases.theory/msg/16c6debecdb29896?dmode=source
>
> "Let rv be a relvar with an attribute rva that is of a relation type
> with attributes a1, ..., an. Then rv is in GUNF if and only if there
> exists a possible value r of rv such that there does not exist a
> relation s such that:
> r WHERE COUNT(rva) > 0 = s GROUP ( { a1, .., an } AS rva )"
> ...
I'm particularly interested in single-attribute nested relations where all values are singleton sets, as I think these 'come first', in some way, before we talk about multiple-attribute rva's. It occurred to me that this would allow definition of keys without depending on operators like rename or group/ungroup. This is the only advantage I've been able to think of so far so maybe it's not really important, since people seem happy to use external notions like functional dependencies, but for all I know there could be others, beyond a perhaps more precise way to think of tables. However, I think it doesn't require any real disturbance in the way users think and should probably require only a trivial change to, for example, TTM, even though I haven't got a tidy revision handy.
Hugh Darwen mentioned the world 'artifice' when he was referred (AFAICR) to the example of a catalogue that records relation keys. He thinks a lot deeper than I do, so I would never say he's wrong, but I couldn't help but observe that a dbms is an artifice in the first place, for that matter, so is a relational interpretation. I wondered whether Codd saw the 'artifice' as being a necessary consequence of his Information Principle. Just thinking out loud.
p Received on Tue Jul 25 2006 - 03:56:18 CEST