Re: constraints in algebra instead of calculus

From: paul c <toledobythesea_at_oohay.ac>
Date: Wed, 23 May 2007 14:27:38 GMT
Message-ID: <ubY4i.212893$aG1.147055@pd7urf3no>

Brian Selzer wrote:
> "paul c" <toledobythesea_at_oohay.ac> wrote in message
> news:nnM4i.206459$DE1.56863_at_pd7urf2no...
>

>>Brian Selzer wrote:
>>
>>>...
>>>V is not in BCNF: since StaffId --> CourseId in Teaches, StaffId --> 
>>>CourseId in V.
>>>...
>>
>>I'll think about your other points, it could be that I misunderstood his 
>>point - meanwhile, I must say I'm not too concerned about V's 'form' as I 
>>think he used it just to make his exposition clearer - in a suitable 
>>language, I would guess, one would not be forced to define a view in order 
>>to express the constraint.
>>

>
>
> Sometimes it makes them more succinct. There is danger, however. I don't
> think it's decidable how a constraint defined on an arbitrary view affects
> the closure of the set of constraints on the database schema. Can a
> constraint defined on a view always be a logical consequence of a set of
> constraints defined on the underlying base relations? I don't think so.
> ...

If view V = A & B, surely constraint C on the view, giving V & C, would be manipulated by an engine with the same effect as the expression (A & C) & (B & C).

I've never heard of anybody requiring expressions to result in BCNF values, surely that would cripple an algebra.

p Received on Wed May 23 2007 - 09:27:38 CDT