A measurement based comparative evaluation of effectiveness of object-oriented versus conventional procedural programming techniques and languages

Abstract

Many assumptions have been made about positive effects of the object-oriented paradigm on software quality. It has been also widely assumed that the object-oriented programming technique is more powerful than the conventional procedural programming technique and object-based languages provide abstraction far better than procedural languages. However, in reality little or no empirical evidence has been given to support these assumptions. Therefore, there is a desperate need for a quantitative-based comparative evaluation of merits of object-oriented versus conventional procedural program design methods and languages. This paper reports on a pioneer work in this direction. The paper describes the results of the application of well-established software metrics on data-collection through controlled experimentation, where forty subjects developed program samples in Java and C languages for the solution of the same set of problems via object-oriented and procedural structured program design methods respectively. Three important aspects of program quality known as program difficulty, effort to comprehend implementation of the algorithm and level of languages have been measured from Java and C programs using well established and recognized software science metrics.