Leveraging Software Analytics to Support Software Maintenance and Evolution

今天，软件开发项目产生了大量的数据，包括源代码，开发人员的讨论，功能规范，错误报告，执行跟踪，以及最终用户的反馈等。数据在现代软件开发中起着至关重要的作用，因为它包含了有关软件和服务质量以及软件项目动态的重要知识。在源代码中，开发人员使用应用程序编程接口（API）作为代码重用的手段。API允许开发人员通过向库和框架提供高级功能并封装实现细节来与它们进行交互。其目的是降低开发成本，提高系统的质量。然而，最近的研究表明，API通常涉及主要与其设计，质量以及学习资源有关的挑战，例如高质量，完整和正确的文档或代码示例的可用性。此外，API发展迅速。如此快速的发展使得开发人员很难与API的变化保持一致。显然，需要适当和完整的API文档，以帮助开发人员完成软件维护和开发任务，从而提高他们的生产力。该研究计划的长期目标是通过利用软件分析（即使用数据及其分析进行决策）来帮助增强和重新记录传统的API文档。这一目标将通过为软件组织提供可操作的，准确的，高效的上下文感知的API摘要方法和推荐系统来实现，与过去的研究不同，它不仅描述了API元素的目的及其使用，而且还描述了其重大变化（以及可能的理由），与其他元素的依赖关系及其质量。我们的研究方法将得到广泛认可的采矿开发历史、痕迹、非正式文件（例如，电子邮件跟踪、错误报告和代码审查），以及有效的数据挖掘算法，以发现API元素之间的关系，同时保留它们的更改/使用顺序。从这项研究中产生的方法，系统和工具将通过涉及开源系统和（可能的）工业系统的大规模实证研究进行评估，而它们的有用性将通过与专业开发人员和项目经理的用户研究进行评估。

Software development projects generate impressive amounts of data today; including source code, developers' discussions, feature specifications, bug reports, execution traces, as well as end-users feedback, etc. Data plays an essential role in modern software development because it contains a significant knowledge about the quality of software and services as well as the dynamics of software projects. In source code, developers make use of Application Programming Interfaces (API) as a mean of code reuse. APIs allow developers to interact with libraries and frameworks by providing them with high-level features and encapsulating implementation details. The aim is to reduce the development cost and increase the system's quality. Recent research has shown that APIs, however, often involve challenges which are mainly related to their design, quality, as well as learning resources such as the availability of high-quality, complete, and correct documentation or code examples. In addition, APIs evolve fast. Such rapid evolution makes it difficult for developers to stay tuned with changes to the APIs. Clearly, there is a need for proper and complete API documentation that can help developers during their software maintenance and evolution tasks, and consequently increase their productivity. The long-term goal of this research program is to help enhance and redocument traditional API documentation, by leveraging software analytics, that is the use of data, and their analysis for making decisions. This goal will be achieved by providing software organizations with actionable, accurate, and efficient context-aware API summarization approaches and recommendation systems, that unlike past research, will describe not only the purpose of an API element and its usage, but also its significant changes (and whenever possible their rationale), dependencies with other elements, and its quality. Our research methodology will be supported by widely-acknowledged techniques for mining development history, traces, informal documentation (e.g., email treads, bug reports, and code reviews), as well as efficient data mining algorithms to discover relations between API elements, while preserving the order of their changes/uses. The approaches, systems, and tools resulting from this research will be evaluated through large-scale empirical studies involving open-source systems and (possibly) industrial ones, while their usefulness will be assessed by means of user studies with professional developers and project managers.