Re: relations aren't types?

From: Marshall Spight <>
Date: Mon, 29 Dec 2003 00:43:52 GMT
Message-ID: <cNKHb.64146$VB2.125134_at_attbi_s51>

"Bob Badour" <> wrote in message
> Sorry. Time is just as atomic as any other type including relation types.

I'm not clear on why types have to be atomic. I have not seen a reason for it.

For example, consider the TimeOfDay type. Perhaps it has logical components hour, minute, and second, and possibly something to decide whether it's currently AM or PM. The AM/PM operator need be defined only in terms of the hour attribute.

Now a given implementation of TimeOfDay may well use a single long or some such, but that still doesn't make the *type* atomic, only that specific implementation of the type.

I can imagine that one could have a kind of "type project" operator that gives me just the Hour attribute from the TimeOfDay, and that brought along the AM/PM operator, since it was defined only in terms of attributes that remained after the project.

If one had these operators on (some) types, I don't think one could continue to call those types "atomic." I do believe that the system will necessarily have some atomic types, but I don't see that there has to be very many of them; say, less than ten.

> > Agreed. Would you object to a relation with an attribute of type
> > Alpha?
> I cannot remember: Is alpha the type with an empty set of values and the
> entire universe of operations? Or is alpha the type with the entire universe
> of values and the empty set of operations?

Allow me to suggest a mnemonic: alpha is the first letter in the greek alphabet; it is at the "top" of the sequence. Omega is the last letter; it is at the "bottom" of the sequence. Alpha corresponds to what lattice theory calls "top" and omega to "bottom." Alpha/top is the maximal superclass in the type lattice; it has all values in it.

Alpha/top as an attribute type has some modest usefulness in cases of extreme genericity. Omega/bottom as an attribute type means the relation could never have any rows, which distinctly limits its utility.

Marshall Received on Mon Dec 29 2003 - 01:43:52 CET

Original text of this message