Re: Looking for how to model 3D objects in 2D relational databases
Date: 10 Oct 2005 19:06:26 -0700
Message-ID: <1128996386.302993.214310_at_g47g2000cwa.googlegroups.com>
Mikito Harakiri wrote:
> Gene Wirchenko wrote:
> > On 9 Oct 2005 09:01:56 -0700, "Marshall Spight"
> > <marshall.spight_at_gmail.com> wrote:
> >
> > [snip]
> >
> > >The first thing you need to observe is that relational databases
> > >are not 2D. A dimension is a measure independent of any other
> > >dimensions. The columns of a relational table are independente;
> > >they are dimensions.
> >
> > That has bothered me for some time now.
> >
> > Given a PK, the columns are not independent. By that, the
> > dimensionality of a table would be the number of columns in the key.
> > What am I missing?
>
> Exactly. This is why some tables are even called dimensions in DW
> terminology. For example, it would be ridiculos to suggest that
>
> Time(year, month, date)
>
> is a 3 dimensional entity. On the other hand, we have fact tables,
> which could clearly have any number of dimensions.
You have to be really specific when you are talking about dimensions, as their are few terms that possess such a range of subtly different connotations across maths and computer science. (Although to the OP, there are no definitions of a dimension that make RM 2D).
Each Time(year, month, date) may be plotted as a point in a 3 dimensional space, certainly making the entity 3D. Face recognition algorithms use similar techniques, where a 100x100 image might be modelled as a point in a euclidian space with 10,000 dimensions, or its dimensionality reduced to a much more tractable vector of n "feature" coefficients, in an n dimensional space. As such, almost anything can be viewed as a dimensional entity (whether it seems initially intuitive or not - and the model will of course leave that intuition to you). Received on Tue Oct 11 2005 - 04:06:26 CEST