In software engineering, the developers in order to recognize which ingredients or fragments of any software code that put into practice a definite functionality, they utilize Information Retrieval (IR) methods to mechanically spot the code that implement them. The main contribution of this paper is to study and examine the effects of skipping some textual information, namely, the internal documentations from being integrated when performing source code indexing for locating changes process purposes. In this paper, we performed two experiments over three open systems namely Qt, HippoDraw, and KOFFICE. The first experiment is done with counting the internal documentations when preprocessing the software code for locating changes process, while the other one is when skipping it. We used the standard IR measurements, Recall and Precision, and we computed the Wilcoxon signed-rank test to compare and evaluate the results. The experiments results demonstrate that not all internal documentations should be always considered in the process of locating changes. Cases in point, for the Qt system, the results show that the internal documentations improved the results of locating changes while for HippoDraw and KOFFIC systems, the internal documentations negatively impacted it.
展开▼
