Re: more closed-world chatter
From: Jon Heggland <jon.heggland_at_idi.ntnu.no>
Date: Sat, 12 May 2007 10:31:03 +0200
Message-ID: <f23u3d$qsu$1_at_orkan.itea.ntnu.no>
>
> Well, the sets identity
>
> A /\ B = A \ (A \ B)
>
> is a special case of
>
> (A /\ B) \/ (A /\ 00) = A \ (A \ B)
>
> hence the term. I have no doubt it has been known since ancient times,
> but I'm not aware if anybody put an emphasis to it like Date did.
Date: Sat, 12 May 2007 10:31:03 +0200
Message-ID: <f23u3d$qsu$1_at_orkan.itea.ntnu.no>
Vadim Tropashko wrote:
> I see. The term semijoin applies to "A \ (A \ B)", while the term for
> "A \ B" is antijoin.
Antijoin can be defined in terms of semijoin and complement, but it seems it's cleaner to do it the other way around, since you then only need the one operator. Nice.
>> What do you mean by the phrase >> "The generalized Date identity"?
>
> Well, the sets identity
>
> A /\ B = A \ (A \ B)
>
> is a special case of
>
> (A /\ B) \/ (A /\ 00) = A \ (A \ B)
>
> hence the term. I have no doubt it has been known since ancient times,
> but I'm not aware if anybody put an emphasis to it like Date did.
I see. So the special case is when A and B have the same attributes, reducing (semi)join to intersection and antijoin to minus. I was thrown off by the general case not being "symmetrical" with regard to A and B, but now I see that if A and B have the same attributes, (A /\ 00) = (B /\ 00), right?
-- JonReceived on Sat May 12 2007 - 10:31:03 CEST