Re: dimensional model question

On 2006-11-13, brw wrote:

> I kinda think I need to separate (a sales table and an expenses
> table), since sales will use different dimensions than expenses
> (right?).

The usual Kimball-style rationale is that facts measure business processes living at the intersection of the relevant dimensions. If the dimensions change, normally so does the process, and so you also need a new fact table. I think this is sound advice in the requirements engineering sense: you need to focus on what you're measuring and why.

The slightly more theory-backed answer is that as far as these concepts apply to only parts of schemas, even in dimensional models the fact part ought to be fully normalized and free of nulls. Hence, everything suggested by conventional relational theory applies. In this case you seem to suggest that the minimal candidate keys of the two relations you're contemplating will be only partially overlapping, or even disjoint. Given the functional dependencies implied by that, normalization too speaks against unifying the relations.

-- 
Sampo Syreeni, aka decoy - mailto:decoy_at_iki.fi, tel:+358-50-5756111
student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2