An Empirical Study of Software Changes in Statoil ASA - Origin, Priority Level and Relation to Component Size

Abstract

This paper describes the results of analyzing change requests from 4 releases of a set of reusable components developed by a large Oil and Gas company in Norway, Statoil ASA. These components are total 20348 SLOC (Source Lines of Code), and have been programmed in Java. Change requests in our study cover any change in the requirements. We have investigated the distribution of change requests over the categories perfective, adaptive and preventive changes that characterize aspects of software maintenance and evolution. In total there are 208 combined perfective, adaptive and preventive changes. The results reveal that 59% of changes are perfective, 27% of changes are adaptive and 14% of changes are preventive. The corrective changes (223 in total) are excluded in this paper, since they will be analyzed in future work. We have also investigated the relation between customers? and developers? priority on change requests and found no significant difference between customer and developers? priority of change requests. Larger components had more change requests as expected and priority level of change requests increases with component size. The results are important in that they characterize and explain the changes to components. This is an indication as to which components require more effort and resources in managing software changes at Statoil ASA.