Re: Much Ado about Nothing

>This would mean that SUM(A) should really be Null if any A is Null?

Assuming one even wants the concept of Null in a database, yes. If one defines addition such that (A+Null) yields Null, then SUM(A) should yield Null if any A is Null.

When the user gets a suprising Null as the result, the user will have to decide how to rephrase the query to yield understandable and consistent results. Giving the unsuspecting user and incorrect result is unforgiveable.

>Didn't either Codd, Date, or both, demand multiple kinds of Null, such
>as one for "value exists but is unknown at this time", another for "not
applicable", etc.?

Codd introduced the concept of Null rather late in the game and then suggested 4-valued logic as a way around the problems of Null. Date is dead-set against Null because it violates 1nf.

>As for duplicate tuples in a table being evil, I'm nodding my head to
>the point of spinal injury, but would this imply that duplicate tuples
>in all relations are evil, and aren't the result sets from SELECT also
>considered to be relations?

Select should also yield 1nf relations. Duplicate rows have absolutely no place in an RDBMS. Users have no need for duplicate rows and no way to understand duplicate rows. Received on Sun Mar 11 2001 - 05:30:46 CET