VTT Publications 727. 93 p. + app. 118 p.

Abstract — In this paper a new maintenance scenario is outlined based on refactoring techniques. Specifically, refactoring techniques are classified according to two opposed program properties: understanding and efficiency. Understanding oriented refactoring disassembles the program preparing it for maintenance, whereas those oriented to efficiency rearrange it for running. Also, we show the challenges raised from this new perspective on maintenance. At present, this ongoing research is being carried out in the functional setting.

Abstract—Programmers performing a change task must understand the existing software in addition to performing the actual change. This process is likely to be affected by characteristics of the task. We investigated whether the nature of a task has any relationship with when a programmer edits code during a programming session. We characterized differences in editing behaviour with three types of editing styles: EDIT-FIRST, EDIT-LAST, and EDIT-THROUGHOUT. We based our analysis on the interaction history of over 4000 programming sessions collected as part of the development history of open source projects. Our results showed that an enhancement task (as opposed to a bug fix) was less likely to be associated with a high fraction of source code edit events at the beginning of the programming session. To our surprise, we also found that the presence of a stack trace in a bug report did not significantly affect the editing style of the programming session. Keywords-Program comprehension; Development interaction history; Mining software archives.

Programmer interaction histories are event sequences that have been culled from the use of instrumented IDEs, and can be mined to better understand the work habits of developers. A prerequisite for this analysis is an understanding of where the boundaries between distinct tasks lie. Asking developers to indicate manually when they switch tasks is disruptive to their normal work flow, so researchers have sought ways to infer task boundaries automatically based on the content of the interaction histories. Coman previously reported a fully automated algorithm that achieved 80% accuracy in a lab validation study. In this paper, we evaluate the use of Coman’s algorithm within an industrial setting. We found two problems: first, a large number of the tasks identified are in fact only sessions of a larger task; second, the demonstrable eÆects of interruptions are not considered. We argue that the problem of task boundary detection consists of two subproblems: first, detecting task sessions; and second, linking task sessions. Coman’s algorithm eÆectively addresses the first, while the second remains unsolved. 1.