Textual information driven software quality improvement

点击翻译按钮获取中文摘要

Society's reliance and dependence on software and information technology is nowhere more obvious than in the ubiquity of hand-held devices. The typical teenager will go no more than a few minutes per day without touching either a cell phone, an MP3 player or a gaming console. Consumers demand more sophisticated programs, more features, and the highest possible quality. Software developers must continuously evolve this huge number of programs. However, evolving programs is difficult for several reasons. High-level documentation is scarce or outdated and, thus, most of the times developers must rely on the only up-to-date sources of information: the source code itself, and, when available and up to date, on textual information contained into bug-tracking systems, mailing lists, forums or blogs. The problem of changing software is further exacerbated information inconsistency. The sources of information are by nature textual, in natural language, and, thus, difficult to use systematically because they are ambiguous, e.g., "path" could refer to a filesystem path for a developer while it may refer to a path on a map for a user of a map navigation system. We conjecture that information inconsistency impairs program understanding, it increases program understanding effort, and, ultimately, fault proneness. The long term goal of this research program is to promote information consistency and thus increase software quality through the identification of textual ambiguities and inconsistency, the recommendation of improvements, and the semi-automatic refactoring of textual information (including code comments and identifiers). More precisely, we will quantify the effect of context and information inconsistency on fault and change proneness and on program understanding effort, when considering as source of textual information the code itself, bug reports, mailing lists, forums and blogs. Once the inconsistency will be identified and quantified appropriate recommendations of improvements will be made, and, based on developers decisions refactoring actions, to mitigate or completely remove inconsistency, will be performed.