Radicore Forum
Fast Uncompromising Discussions. FUDforum will get your users talking.

Home » RADICORE » How To » Regarding optional attributes and denormalisation
Regarding optional attributes and denormalisation [message #2809] Thu, 06 October 2011 06:01 Go to previous message
rksk16it is currently offline  rksk16it
Messages: 2
Registered: October 2011
Junior Member
Hi Tony

Though I am not a user of radicore or PHP, I read your articles often, and try to apply them in my own projects. Smile

I was reading your article regarding normalisation and effective database design : http://www.tonymarston.net/php-mysql/database-design.html. And it was like 10th time I was reading it. Cool

In the section of denormalisation, there is a sub-section about optional attributes that exist as a group : http:// www.tonymarston.net/php-mysql/database-design.html#optional. attributes. There you said the following :

R (K, A, B, C, X, Y, Z) where:

   1. Attribute K is the primary key.
   2. Attributes (A B C) exist all the time.
   3. Attributes (X Y Z) exist some of the time (but always as a group under the same circumstances).
   4. Attributes (X Y Z) require special processing.

After denormalising the result is two separate relations, as follows:

    * R1 (K, A, B, C)
    * R2 (K, X, Y, Z) where K is also the foreign key to R1


What I can see is that you said that breaking of the original relation 'R' into 'R1' and 'R2' is denormalisation. But I fail to see any single normalisation rule which is broken.

1st NF : No repeating groups
2nd NF : No partial dependencies
3rd NF : No transitive dependencies
BC NF : Every determinant is key
4th NF : No multivalued dependency (if R doesnt break it, then probably R1 and R2 also dont break it)
5th NF : Cannot be decomposed into furthur smaller relations. (Again if R doesnt break it, R1 and R2 probably also dont)
6th NF (No use discussing)

Also, from the application point of view, R1 can be seen as a class 'customers', and R2 can be as 'customers_with_arrears' which inherits 'customers' as it follows one-to-one relationship with some extra attributes, effectively increasing the specialisation (thus inheritance).

Thus it looks like a better design overall, and I cant see any denormalisation. I would be grateful if you can help clarify it.

Thanks a lot Smile

[Updated on: Thu, 06 October 2011 06:02]

Report message to a moderator

 
Read Message
Read Message
Read Message
Previous Topic: Using UPD4 to process emails
Next Topic: Tutorial Part 4 - Figure 6
Goto Forum:
  


Current Time: Sat Nov 23 20:16:39 EST 2024

Total time taken to generate the page: 0.00901 seconds