APSEC '99 Proceedings of the Sixth Asia Pacific Software Engineering Conference
| Hi-index | 0.00 |
In this paper we analyze statistically the efforts for C++ program modification which is needed by a given requirement change during the prototyping development of a certain GUI application. In the analysis we consider both C++ program P to be updated and the resultant C++ program P', and discuss the correlation between the values M(P) of the Object-Oriented metrics obtained from P and the efforts E(P') needed to produce P'. According to the definitions of M(P) and E(P'), we present two approaches in this paper.In the first approach, we take M(P) as the value obtained by applying the metrics to the whole source code P and E(P') as the lines of codes(LOC) that are actually modified or created, respectively. However, the experimental result cannot show strong correlation between M(P) and E(P').Based on the analysis results of the first approach, we propose the second approach to consider the object-oriented properties more directly. The analysis process consists of following three steps: 1) take only the classes in P, classify them according to the MVC paradigm originally for Smalltalk their functions, and then evaluate M(P) for the classified classes, 2) in order to evaluate the efforts for code modification, define E(P') as the heuristic value which is calculated empirically from the numbers of methods and members created or modified, and 3) analyze the relationship between the values of M(P) on the classified classes in P and the values of E(P') on the source codes in P'. From the experimental result, we can prove there exists a high correlation between them.