Dear Researcher

please refer to these attachments if it is helpful for you.

Reagards

http://www.diva-portal.se/smash/get/diva2:597126/FULLTEXT01.pdf

https://engineering.purdue.edu/RVL/Publications/Sarkar05MetricsForAnalyzing.pdf

Basic metric that you should collect is coupling, in-degree and out-degree metrics. As for as the test adequacy is concerned, an impact analysis based criterion will be helpful. Please read our paper:

http://www.interscience.in/IJCSI_Vol3Iss2/24-33.pdf

Very good question. There is actually not a lot of research into associative relationships. Coupling concepts do not address this type of relationship. Since the basic concept laydown by Myers covered only direct module to module relationships. This also suggests that in flow and out flow metrics are not appropriate.

I recommend writing a paper that investiages this relationship. My first though is to define a test criteria for the data structure and then determine if the two modules conform.

Here, the coupling metric needs to be collected not between the two modules but between the file (resource) and the module that is invoking it. Here, as there is no direct dependency between the two modules which are accessing the same resource, definitely the value of coupling between the modules will be 0.

But now, the coupling measure of resource with modules and the in-degree and out-degree of the said resource and the modules are to be estimated. This will definitely help in arriving at dependency measure calculation.

@Carl... Thank for the reply. I think you got the main point. Currently i am working on this issue. Can you give me some idea on test criteria?

Thank you

Generally, establishing specific test criteria is not possible because it is dependent on the properties of the data store, i.e. random v. sequential. It is also dependent on the nature of the fields and their dependencies. My approach is very similar to that used with the triangle problem. Look to the output to construct the cases. In this cause the output is the data store.

When I'm faced with this problem in practice, I construct the equivalence conditions based on the data store, i.e. what is a valid and invalid condition for each data item in the store and the structure itself. If this becomes to difficult, I find that algebraic specification. Next, I construct cases from the data modified by the each module and the algebraic specification for the data structure. If the module effects multiple attributes then I apply combinatorial testing.

Transitive testing is a challenge. Let me know how this works in your research.

Hi,

I think there is some other way too to handle it.

Consider that,  module x is modifying the content of a file F which is actually being used by module y. In this case, the dependency between the modules x and y  is not direct. But module y's functionality is affected due to the change in file F.

In these kinds of situations, it is better to go for Formal Specification based testing. If you are using OO in your designing, then naturally OCL based formal specification will be useful. In such specification, we can specify both pre-conditions and post-conditions of each and every class (in your case, it may be a module containing classes) which will help in deriving test cases to check for the proper working of the modules. For further details about it, you can refer our publication on OCL in generating test cases.

 Thank you Jeya Mala Dharmalingam.

For OO design OCl can be useful. What about procedural based system like C?