Re: How to model searchable properties of an entity

> Let us say I have Part# and it has very large number of flexible
> attributes defined by the user, such as color, width, height, etc...
> Depending on Part number, these attributes may change.

There is an inherent mismatch between tables containing rows with uniform number of columns and "flexible attributes defined by the user". Its like asking how to fit a circle in a square. The solution will always be less than ideal.

One, less than ideal method, but without NULLs and minimal duplicate data uses generic modelling (ie data in a few tables, where columns in effect become rows). A recent thread titled "Nearest Common Ancestor Report" touches upon generic modelling. The cost is usually decreased utility of SQL, increased reliance on coding, decreased performance, etc.

I am currently developing an experimental db which allows each thing to have variable number of relationships to other things. If you would like to see its advantages/disadvantages in this particular case, please post some sample data. Received on Fri Aug 20 2004 - 17:45:01 CEST