Re: Eliminating Nulls

From: Mark D Powell <>
Date: 3 Jul 2005 14:34:50 -0700
Message-ID: <>

I was thinking about two different situations. One of which was based on an example of a special situation that I cannot remember the details on but I had ran accross it in two books on SQL Server T-SQL. I had hoped to look it up and add to the thread but time did not permit. Now I am out of the office till next Wednesday and it is likely to be busy.

The other situation which I should have mentioned is where the key columns to a table ended up being null. When this condition arises the problem often traces to the parent tables not being properly normalized to begin with. Sometimes what you will find is that the customer talks about A, A, and more A. So A ends up in the design with a relationship to B. But if you study the raw data elements what you really have is that A is a derived result of C and D. B is in fact realted to either C, D , or the combination of C and D. These conditions are often not detected until the developers tries to defind a FK and the FK ends up being only on a partial PK. Another give away to a failure to normalize is where the update ends up being dependent not on the PK columns but on another column. Where every DBA is going to recognize this as a problem and will raise a red flag during design the fact is all too often the DBA does not see the design prior to implimentation.

HTH -- Mark D Powell -- Received on Sun Jul 03 2005 - 23:34:50 CEST

Original text of this message