Dimensions?

Robert Graham Merkel rgmerk@mira.net
Thu, 28 Sep 2000 01:28:20 +1100


Christopher Browne writes:
 > On Wed, 27 Sep 2000 10:46:27 +1000, the world broke into rejoicing as
 > "Phillip Shelton" <shelton@usq.edu.au>  said:
 > > True. I think what I want is to be able to report on the transaction from
 > > both clothes and shoes.
 > > 
 > > Others have shown that if I use a well defined tree I can roll-up my account
 > > info  and have it do what I want it to.
 > 
 > ... And in perhaps-overly-brief response to the multiply-expressed
 > question,
 >   "What if it's not a tree?"
 > 
 > ... I'll just respond that if it's a tree, then rolling things up to the
 > right point is a completely linear matter, and allows you to maintain
 > the property that it all adds up nicely.
 > 
 > If, instead, it's not a tree, but a general graph, where "Shoes"
 > could occur anywhere, rather than adding things up, reports would
 > need to use more general set-theoretic operations like intersections,
 > unions, and such, where there is considerably more risk of accidentally
 > double-counting transactions and otherwise "messing up," and, arguably
 > more importantly, making it harder to understand/predict exactly what
 > is happening because there's _NOT_ the "linear" relationship that "all
 > the transactions are in this subtree."
 > 
 > The engine code required to support "connected graphs" isn't the nasty
 > part; it's probably tough to get right, but certainly not impossible.
 > The _NASTY_ part would be to put together a user interface that you can
 > implement and that users will understand.

I tend to think that interpreting these kind of complex relationships
is not really the domain of the engine at all - it's the domain of the
individual reports that interpret the data in the engine in different 
ways.

As for the user interface question, ideas are rolling around in the
back of the head, but you're right, it is *nasty*.

------------------------------------------------------------
Robert Merkel	                           rgmerk@mira.net

------------------------------------------------------------