Abstract

Maintainability of program parts refers to the ease with which these parts can be modified. Many existing metrics of internal quality attributes such as cohesion, coupling, etc. have been used in this regard. Cohesiveness among program parts is considered as a strong measure for the maintainability of object-oriented components and to predict the probability of being erroneous. Our objective is to propose a novel graph-based cohesion metric to measure the maintainability of different program parts in an object-oriented program and predict their fault proneness. We compute the cohesion of the sliced component as a measure to predict its correctness and preciseness. In addition, we wish to theoretically validate the proposed technique against the existing guidelines of cohesion measurement and compare it with some existing techniques. We propose a new cohesion metric named affected component cohesion (ACCo) to measure the maintainability of different program parts and predict their fault proneness. This metric is based on the hierarchical decomposition slice of an object-oriented program that comprises all the affected program parts. The slices are obtained with respect to some modification made to the program under consideration. It is essential to consider all possible dependence relationships along with the control and data dependences that exist between different program parts of an object-oriented program for a better program comprehension. To represent these dependences, we construct a suitable intermediate graph for an object-oriented program. Then, we compute the slice of the object-oriented program using the intermediate graph to extract the affected program parts. These extracted affected program parts are represented as nodes in the proposed affected slice graph. The critical and sub-critical nodes that require thorough testing are determined by estimating their cohesion measure. The theoretical validation signifies that ACCo satisfies all the existing properties for cohesion measurement. The results obtained are more precise and comparable with other existing approaches. ACCo is a more precise and practical technique to measure the inter-relatedness of affected program parts in an object-oriented program. The discussion on possible threats to its validity demonstrates the scope for improvement of this approach.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call