Re: Relation subset operators

On 4 juin, 02:14, paul c <toledobythe..._at_oohay.ac> wrote:
> Cimode wrote:
>
> > While working on aggregation within groupping operations on the db
> > core I design for relation  manipulation, I questionned myself about
> > the opportunity of using new operators to simplify relational division
> > formulation and make it more systematic.   For instance, conside the
> > following questions:
>
> > suppose CAR_SALE relation represented as
>
> > CAR_SALE
> > id         car                     salesman                price           color           date
> > 1          Buick           Henderson               10000       Red         01/01/1990
> > 2          Buick           Wilkinson               10000   Red             02/01/1990
> > 3          Chevrolet               Hutchinson      10000   Red             12/01/1990
> > 4          Buick           Wilkinson               10000   Blue            13/01/1990
> > 5          Chevrolet               Henderson       10000   Red             14/01/1990
> > 6          Buick           Henderson               10000   Blue            16/01/1990
> > 7          Buick           Henderson               10000   Blue            18/01/1990
> > 8          Chevrolet               Parson          10000   Yellow  18/01/1990
>
> >...
>
> Cimode, I'm struggling with this, trying to see the logical starting
> point.  I would try to write this en francais but that would be
> incomprehensible, a teenage friend in France even runs rings around in
> English.   I can't remember when I might have read about logical
> foundations of aggregate operators, but for me they've seemed to
> involve, necessarily, the equivalent of TTM group (I'm always a little
> leary of SQL GROUPBY because I gather it doesn't have a logical
> definition).  Assuming a 1970-Codd-style relation there are many "SUM's"
> inherent in the above table/r-table/relation.  To me, that 'inherency'
> seems similar to the transitive closures that are visible in some other
> relations.  The examples I've seen elsewhere of TTM-style GROUP-ed
> attributes haven't offered any kind of operation that queries a subset
> of the tuples in a GROUPed value.  So, if this makes any sense, from an
> algebraic viewpoint, I see aggregates as being very similar to TCLOSE,
> they must be applied first, before restriction.  Are you following  me?.
my purpose in creating an operator is not to explore aggregation but to allow a practical formulation of some operation such as relational division or specialization by constraint. As for TCLOSE, the operation is close but different in the sense that subtyping allows to get around a lot of restrictions that apply to TCLOSE. Received on Thu Jun 04 2009 - 11:01:13 CEST