Re: Modeling As Data Or Metadata?

What i do is to differentaite between quantitative , qualitative and descriptive observations : for descr attributes you use free text, for qualitative you use either a domain or a lookup table (e.g. high, medium, low) and for quant you use numeric value and a unit (This is a scenario that makes J celko whince...)

Now to collect all your dynamic attributes together as columns you just forfeited the declararive power of sql - you need to dynamically make up the sql to pivot rows out to columns. This you can do with a statement that contains a selfjoin for every dynamic attribute. This statement has a regular form and is easy manipulated in code (factory /builder class on client OR a stored procedure if your sql support execute immediate of dynamically constructed sql)

In my mind the next best thing after stable attributes is a predictable pattern of changes for new attributes.

I would appreciate any feedback as much as you on this ;-) Received on Tue Mar 26 2002 - 15:28:38 CET