The ORM Foundation

I have a question with regard to one of the solutions to an exercise from the previous edition of the text book "Information Modelling and Relational Databases" (2001). I have re-drawn the solution diagram using the current ORM2 notation (from Exercise 5.4 (4)):

The reference scheme is disjunctive in nature. This suggests that the resulting primary key will have null values in one or more of its component attributes (which, as we know, is a big no-no). Is this a situation where there is a mismatch between the conceptual level and the relational (implementation) level?

The error message that I got seemed to agree with me. It stated:

"Error    1    Entity type 'Member' in model 'ORMModel1' requires a local preferred reference scheme or supertype(s) that provide preferred identification.    ORMModel1.orm    0    0"

I understand from the text that disjunctive reference schemes are a murky area. Is there any room for them at all?

Cheers

Hi Andy,
Thanks for the question.
There are several points to make about your example.

Entity Types (page 63 of  Terry's  2001 book and page 66 of the 2008 version)
An ORM Object Type can be either an "Entity" or a "Value"
In your diagram the ORM Object "Member" is an Entity type and not a Value type.
This means that the Object Type "Member" is a name for the set of all possible instances of Member.
Sets don't have duplicates so "Member" needs a reference.

In my example I have chosen ".id" but you can chose others.
I used the NORMA Fact Editor to type in Member(.id) has GivenName() 
The empty parentheses after GivenName tell NORMA that this is a Value Type.

Your "Member" object is coloured red because the fact does not conform to the ORM2 rules.
This is what the error message is saying.

Allowing for multiple given names
Your example allows for up to three given names, two of which are optional.
Since, as you say, it is wise to avoid nulls my example is coded in a way that (a) allows for as many given names as you want and (b) avoids creating nulls in the database.

I suggest that you make liberal use of the Verbalizer window. This will help you to identify the exact meaning of any facts that you enter and to correct them as you add them to the model.

Hope this helps
Ken

 Thanks Ken. You have confirmed what I suspected. The solution given was incorrect. I never understand why textbooks contain incorrect solutions.

The verbalisation of the facts makes sense:

For each Surname, GivenName, GivenName, and GivenName,
at most one Member has that Surname and
has first that GivenName and
has second that GivenName and
has third that GivenName.

I believe that it is not a reference scheme at all. It is just a uniqueness relationship which is apty notated using the external uniqueness constraint.

But the disjuntive problem is not good for a primary reference scheme. And of course, the easy way out is to make up a sequential identifier - that solves the primary reference scheme problem every time.

Cheers

Hi Andy,

Andy Reiser:

The solution given was incorrect.

Well, before I agree with you on this, I would need to review the example you mention in order to see what it was trying to illustrate.
I could not find the example from your reference.

Please confirn that the book in which you found the example is "Information Modeling and Relational Databases" 2001 by Terry Halpin?
If so, please will you give me the page number of the book.

Thanks
Ken

 Hi Ken,

 It was not an example in the book. It was one of the solutions to an exercise from the book. The text book is "Information Modelling and Relational Databases" (2001). And the solution was to Exercise 5.4 (4).

Another problem with the solution was that instead of using a P in a circle (as per ORM1 requirements to depict a primary reference scheme), it featured a U in a circle (external uniqueness). There was no reference mode. So, the only possible reference scheme was the names.

Hi Dave,
Thanks for the clarification.
I found the exercise on page 206 but I could not see any "solution".
Where did you see the "solution"?

Anyway, my "solution" to this exercise is shown below.
The frequency constraints are explained in NORMA Tutorial 4.

Ken

Thanks Dave and Ken.

I got the solution in a tutorial at university a couple of years ago. Our tutor had the solutions to the even numbered questions, which he would put on the screen for us to copy down.

I have a big question mark next to it in my notes and vaguely recall that the tutor was not able to provide a good explanation which explained away the messy situation that would arise with null values in the primary key.

Maybe it was just an example to raise a discussion about disjunctive reference schemes.

Thanks for your replies. I really do appreciate it.

Hi Andy,

Thanks for the clarification.
I recently joined the computer science faculty of the University of Lincoln (http://www.lincoln.ac.uk/home/) with a view to introducing ORM to the syllabus and to the university management system. So with my university hat on, I'd like to contact your former tutor with a view to collaborating on ORM education and training. So please will you let me know which university and the name of the tutor and I will follow it up from there.

Thanks

Ken  

Hi Ken,

Last year, I lectured ORM content as a part-time (sessional) lecturer at Queensland University of Technology (QUT). It was based on the 2001 text (so, not ORM2), as the current textbook only just came out as the semester was underway.

I'm not sure who is taking that unit this semester, but the member of academic staff who is responsible for it is Prakash Bhandari. I'm happy to PM you his details, if you like.

Regards

Hi Dave, 

Dave Rogers:

I'm happy to PM you his details, if you like.

Yes please, that would be very helpful.
One of my academic objectives is to use this website to collaborate with and to give support to all academics who teach ORM - wherever they are.
So if you know of any other Universities or Colleges who are teaching ORM, please let me know.

Ken 

Hi Ken,

I learnt ORM at QUT as well. My tutor no longer works there. I believe he is now working in the industry. Sorry I cannot help you out with a contact.

Andy

 An observation....

While perusing this thread, I noticed that the solution proposed by Ken does not contain the same information as that in the opening post by Andy.  In Andy's, I know the full name of the member, but in Ken's I do not know in what order the given name's occur.  (Or does the frequency constraint add information (viz, sequence) to the solution?  [I'm still learning NORMA].)  I don't know what the original exercise was (not having access to the book), so I don't know whether name order was important. 

As a side note to anybody modeling people's names, please be aware that some people (certainly in Australia) only have one name, such as  "Rainbow", "Serenity", etc.  So none of the posted models in this thread would support this situation!

regards, Peter

Hi Peter,
You are right that my model does not show the order in which names occur.
My model allows a person to have one, two or three given names.
This underlines that point that an object-role model says exactly what you want it to say - no more and no less.
The best way to check the meaning of an object-role model is to use the verbalizer.
The verbalizer makes it easy for people to say "No -that is not what I meant" - as you have done in this instance.

Other modeling approaches such as UML and ER require people to learn the meaning of the UML or ER symbols before they can understand the exact meaning of the model diagram. This is one of the reasons why ORM is more cost-effective than other information modeling approaches.  

The following example shows the elementary facts in my model together with their automatically generated verbalizations.
Ken

PS - if the diagram is a bit fuzzy on your screen, just click on the diagram and you should see a clearer image.  

Hi Reiner,

That subset constraint is definitely an improvement. And you would probably need another one running up to the 'has first' fact type from the 'has second' fact type.

Personally, I think disjunctive reference is a really bad idea. Primary Keys (even composite keys) should never be able to contain an attribute with a null value in it. This is an example of where the conceptual schema does not map well to the relational schema, in my opinion. But I'm sure the question was just there to get students thinking about these issues, rather than putting the solution forward as an ideal solution.