CAREER: Program Comprehension in Internet-Scale Code Search for Open Source Reuse in Opportunistic Software Development

点击翻译按钮获取中文摘要

This project will study software developers engaged in the process of searching the Internet for source code that can be adapted to their goals, and it will design tools to support decision-making in this kind of work. It has become commonplace to search the Internet for source code in the course of a software development project. Locating the right component for as-is reuse or a reference example at the right time can have significant impact on the how the project progresses. There are a number of critical decision points in the search process, when the matches returned are eliminated or selected. These judgments require a form of program comprehension, though one that not been studied previously. Current research in program comprehension has been concerned with how software developers form mental models of how a single body of source code works in order to perform a maintenance task. Program comprehension in the context of Internet-scale code search involves discerning the characteristics of a candidate piece of code without becoming entangled in the internals.Internet-scale code search is exploratory, iterative, and open-ended, so both relevance and suitability judgments are made throughout the process. Relevance judgments are made by software developers while identifying promising candidates among the available matches. These decisions are rapid, taking only a few seconds, and use relatively little information. Suitability judgments are made when determining whether a promising candidate is appropriate for the software project. These decisions are slower and typically involve a careful cost-benefit analysis. In this project, both types of judgments will be studied with the aim of creating better tools for Internet-scale code search. By understanding how developers decide between options, the cues that they use to comprehend the match, and the kinds of errors that they make, it is possible to present matches so they can find the desired source code more quickly and accurately. The research will have three components: laboratory experiments to study relevance and suitability judgments in a controlled setting; field studies to examine the search process in context; and refinements to an existing research platform, the Sourcerer code search engine. The evaluation criterion for this project is whether the new knowledge, techniques, and prototypes have reduced errors in relevance and suitability judgments during Internet-scale code search for opportunistic software development.The expected outcomes of the research are: an understanding of the program comprehension that occurs in the context of Internet-scale code search; a model of the search and comprehension process; a characterization of the cues that are used in relevance and suitability judgments; and prototype search tools that support this process. These results could have far-reaching impact, because opportunistic software development and open source reuse have become widely adopted strategies. These practices allow software developers to leverage a global supply chain for software components and knowledge that they can use to implement new product and service ideas to keep up with societal, economic, and corporate demands. Accurate and rapid relevance and suitability judgments are essential for the successful application of these strategies. The educational components of this project will prepare students for the very different world of software development that is currently coming into being.