Reusability Metrics for Program Source Code Written in C Language and Their Evaluation

Abstract

There are various approaches to quantitatively and statically measuring the reusability of program source code; however, empirical demonstrations of the effectiveness of such approaches by considering actual reuse in actual development projects or of the magnitude of their effect on actual reusability have not been reported in depth. In this paper, we identified a set of metrics that are thought to be effective for measuring the reusability of C language program source code. Subsequently, for ten projects involved in development with existing software modification and adoption, during which conventional source code in an old project are extensively reused and adopted to a new project, we compared values of the static metrics identified and the reuse results before and after the development. Statistical analysis demonstrated that some of our metrics are effective for actual software development, and we accurately determined the magnitude of their effect on actual reusability. More concretely, it was found that when the percentage of files used outside the belonging directory is small and the number of function calls is small, the complexity of source code as the material of reuse and factors that are affected by the source code are limited, indicating high reusability.