Re: Define "flatten database" ?

From: Jan Hidders <>
Date: Sat, 25 Feb 2006 10:41:40 GMT
Message-ID: <EvWLf.278612$>

Mark Johnson wrote:
> Jan Hidders <> wrote:

>>1NF: in some cases they tell you to immedeately split off the nested 
>>columns into separate relations (thus avoiding redundancy) and in others 
>>the relation is merely flattened and the splitting off is postponed to 
>>the later normalization steps. So the question is this. Can we always 
>>use both approaches and if we can does it matter for the end result 
>>which one we take? 

> Which some, here, find quite objectional is that reference to a
> table/relation as being in any way, 'flat'.

I think that's mainly their problem. :-) Most real experts in database theory are perfectly comfortable with it.

> As for the question, I suspect you had an answer in mind. I don't see
> that it would matter as the 'splitting' is to be done on the same
> columns on the same table, at whatever point one decides to do so.

Really? Note that this is a mathematical theorem about mathematically defined objects. Can you supply a mathematical proof?

> Perhaps I might ask a related question. Do you believe that a table
> corresponds, but perhaps only weakly, or even inconsistently, to a
> "relation" as understood in the set theory of the RM, or do you
> believe it is fair to call the database table a "relation", without
> much qualification?

The term "table" is vague and can have different meanings that might depend upon the context in which it is used. Without that one cannot really answer your question.

  • Jan Hidders
Received on Sat Feb 25 2006 - 11:41:40 CET

Original text of this message