Re: Functional dependency is multivalued dependency
From: paul c <toledobythesea_at_oohay.ac>
Date: Sun, 30 Apr 2006 14:45:21 GMT
Message-ID: <5445g.93879$7a.33811_at_pd7tw1no>
>
>
> Question - in the RL approach, if pi_XY(I) is (XY \/ I), what does XY
> stand for? (eg., if XY isn't a projection, is it a 'bag'?)
>
> BTW, 8.2.2 seems to be Heath's theorem, but Alice book doesn't seem to
> mention that.
>
> p
>
Date: Sun, 30 Apr 2006 14:45:21 GMT
Message-ID: <5445g.93879$7a.33811_at_pd7tw1no>
paul c wrote:
> Mikito Harakiri wrote:
>
>> The proof of proposition 8.2.2 (page 164, the Alice book) begins with >> observation that given three sets of attributes X, Y and Z such that >> set Z complements sets X an Y in the relation header of the relalation >> I, the pi_XY(I) /\ pi_XZ(I) is always a superset of I. Is it obvious? >> >> In RL we do it in couple of steps: >> >> (XY \/ I) /\ (XZ \/ I) <= I ? >> ...
>
>
> Question - in the RL approach, if pi_XY(I) is (XY \/ I), what does XY
> stand for? (eg., if XY isn't a projection, is it a 'bag'?)
>
> BTW, 8.2.2 seems to be Heath's theorem, but Alice book doesn't seem to
> mention that.
>
> p
>
(I wasn't expecting that it would be a 'bag', it's just that I thought the symbol '\/' stood for semi-union which I think RL intends to operate on relations.) So if we start with the value I that has X and Y attributes, how do we get a value for XY without projection?
p Received on Sun Apr 30 2006 - 16:45:21 CEST