Re: Table(s) definition problem
Date: Sat, 28 Feb 2004 01:41:37 -0500
Message-ID: <40403821.60601_at_charter.net>
--CELKO-- wrote:
>>>I have a very wide table (over 1000 attributes). I can group the >>
> attributes into several, ~20, disjoint sets where the elements of each
> set occur together. <<
>
> That makes no sense. A table models a set of things of one kind; each
> rows is one instance of that kind of thing; each column models an
> attribute by storing scalar values. This nightmare is a blob of ~20
> different things!! What would the name of this thing be in a logical
> data model? -- something along the lines of
> "Bagpipes_Automobiles_SuperModels_ .._Alligators" for ~20 nouns where
> the undescore can be read as "or maybe" ??
Not really. They are related (correlated?) groups of attributes that belong to one entity. Consider owned cars. You can buy or sell them at any mileage, and you only have records from when you buy them to when you sell them. One correlated group of data is the work done (and maybe condition) at each major mileage point in the life of the cars. But you will have 30000 mile data on only those cars you bought before that time and sold after it; similarly the 60000 mile maintenance, etc. Do I have one table with all the maintenance (and condition, but this is amenable to a 1:N relation from car to condition, since the condition would be evaluated on the same attributes each time) or the base car table and some fixed number of maintenance tables, one for each of the major maintenance intervals? They have to be separate tables (so a 1:N relation from car to maintenance table don't seem to work, or rather pushes the problem down 1 level) because of the different type of maintenence work required and done at each milestone. Received on Sat Feb 28 2004 - 07:41:37 CET