Re: Functional Dependencies > Uniqueness Constraints

Marshall wrote:
> Jon Heggland wrote:

>> By that reasoning, FDs shouldn't be included if you have general
>> database constraints.

>> And I want those.

Yes.

> Given that an FD is just a particular kind of general database
> constraint, does that introduce any particular difficulty? Well,
> I guess it does actually, in that I expect it is undecidable whether
> a given constraint is an FD or not. But that doesn't mean we
> can't have a canonical expression of a constraint that is an
> FD, and be limited by the fact that unless we tell the system
> it's an FD (by using the canonical form) it won't know that
> it is.

And the exact same reasoning goes for keys / uniqueness constraints.

> It seems to me this problem crops up in a variety
> of places, such as determining whether an operator is
> commutative based on its constraints.

I agree that high-level abstraction of classes of integrity constraints (keys, foreign keys, FDs) is a good thing. I just don't see why you want to discard keys from that list.

-- 
Jon