Re: Question about Date & Darwen <OR> operator

"Marshall Spight" <marshall.spight_at_gmail.com> wrote in message news:1125706291.163098.312610_at_g14g2000cwa.googlegroups.com...
> Mikito Harakiri wrote:
>>
>> >From http://c2.com/cgi/wiki?RelationalAlgebra
>>
>> a OR b : An extended form of union; if the headings of the operands
>> differ, then "missing" attributes take on all possible values. Thus the
>> result may be very large or even infinite. When the operands have the
>> same heading, then this is the same as a traditional SQL UNION, except
>> that all duplicates are always removed.
>>
>> This informal description matches the other alternative. What is the
>> formal definition?
>
> I haven't seen one. D&D don't really do anything with formal methods
> that I've seen.
>
> I would propose something like
>
> Given
> A:(a:Ta,ab:Tab)
> B:(b:Tb,ab:Tab)
>

I am not sure I understand the notation. Could you clarify what '(a:Ta, ab:Tab)' is ?
> A <OR> B = { (a,ab,b) |
> ((a,ab) in A cross product Tb)
> union
> ((b,ab) in B cross product Ta)
> }

Regardless of what Ta, Tb stand for, I object to the (a, aab) notation which denotes an ordered pair. Since a tuple is a set of (A,T,v) triples, should not it be {a, ab} ?

>
> Is that sufficiently formal? What would constitute a
> sufficiently formal form?
>
>
> Marshall
>
Received on Sun Sep 04 2005 - 14:20:34 CEST