Abstract:

Assessment is an extremely important part of the process of education. One of the most significant learning goals of the introductory programming courses is to make students able to comprehend an already written program source code. A common way to assess this ability of students is by giving them tasks that consist of a source code snippet and a question regarding its behavior. When creating a large number of tasks containing source code snippets, e.g. for educative assessment needs in a typical introductory programming course attended by a high number of students, teachers need to maintain the task complexity in order to achieve fair assessment. This problem can be addressed by turning to automated production of such questions. However, in order to attain the complexity consistency in the process of automated production of tasks containing source code snippets, there must be a means of automated measurement of source code complexity. Further, in order to enable automated measurement of source code complexity - an appropriate source code complexity metric is needed.

In this paper we present a survey on software complexity metrics. We discuss and compare the properties of the analyzed metrics, and identify the advantages and drawbacks of using each of them. We also analyze their potential for application in educational environment, in the context of the described problem. Finally, we describe the complexity metric that we have defined, and that is completely suitable for the problem under consideration. We analyze its properties and we present the basis of our claim on the metric’s suitability by explaining the reasons in detail.

Keywords:

Software complexity metrics, automated measurement of source code complexity, assessment of students' knowledge in introductory programming courses.