Improving Software Quality through Effective Decision Support

点击翻译按钮获取中文摘要

Software is considered to be of high-quality if it functions correctly and exhibits a variety of non-functional requirements, such as being performant, secure and easy to change. High-quality software benefits users who can focus on their work rather than battling with buggy or slow software. High-quality software benefits the development organizations who create the software by shortening development schedules and by lowering software maintenance costs. A critical factor that has been shown to have an impact on software quality is the design of the software. Although evidence indicates that inappropriate design choices and design flaws lead to lower software quality, there is little support to help software developers make appropriate design choices that lead to high-quality software. The long-term objective of this research program is to assist software developers to make informed and appropriate design decisions as they work that lead to the production of higher-quality software. This research will enable design information about a software project, that has typically been invisible and inaccessible, to be externalized and to be available for supporting informed decision-making within a project, for improving development approaches through analysis of design trends across projects, and for supporting the education of good design practices. Three objectives are planned for the research program. First, we will seek to externalize design decisions for a system through the development of algorithms and tools to mine and represent design decisions from a project's artifact history, such as from pull requests and code reviews. Where appropriate, we will associate these decisions with code impacted by the decision. The goal for this objective is to enable a repository of externalized design information to be built from a variety of open source projects. Second, we will seek to automatically identify the context of a developer's work as they work to enable the provision of intelligent design advice. We will develop algorithms and techniques to automatically form a developer's work context to enable useful design recommendations. Third, we will show that it is possible to create useful design recommendation tools that lead to the development of higher-quality software. We will leverage the work from the first two objectives to provide this intelligent design advice at particular intervention points during development, such as code reviews. We will experimentally show that the recommendations have a desired effect on software quality.