Re: cardinality of 1

pc wrote:
> In another thread, Bob Badour said: "Relation-valued attributes already
> allow us to represent optional data as a constrained set where we can
> enforce a maximum cardinality of 1 by specifying a nullary key."
>
> I'm intrigued by this quote, but I just don't get it. Have looked
> elsewhere (thethirdmanifesto.com among other places). My motive doesn't
> have to do with "missing information", but rather finding out whether
> cardinality in any amount can be determined for an arbitrary table (as
> opposed to DEE or DUM) in a way that is intrinsic.

The 'nullary key' is a key on an empty set of columns. At most one row can exist - any second row would shared the same value and hence raise a constraint violation error.

I'm not really sure what the second half of your last sentence means.   If you're looking for a 'cardinality in any amount' where the amount is not (zero or) one, then the 'nullary key' is not the answer.

Going back to BB's comment (as quote above), I don't think that an RVA is needed to permit a nullary key. Any table (including DEE and DUM) can have a nullary key.

You're original source for information on empty keys (nullary keys) would be in Date's "Writings" series. Offhand, I'm not sure exactly which volume - probably 1989-92 (and I'm too lazy to go into the next room and check for you; get back to me if that causes a major problem.)

-- 
Jonathan Leffler                   #include <disclaimer.h>
Email: jleffler_at_earthlink.net, jleffler_at_us.ibm.com
Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/