A Simple, Efficient and Robust Code Completion System for ESTI Development Environment

作为加拿大西部最古老的IT公司之一，ESTI不仅为客户构建定制软件，还为客户的现有软件提供再工程和专业服务。因此，ESTI的技术领导和开发人员必须处理不同种类，领域的软件，最重要的是，由第三方开发团队开发的软件。在开发过程中，开发人员严重依赖各种框架和库，并使用相应的API（应用程序接口）。问题是，由于各种因素（例如示例和文档不足），API很难学习。虽然现代IDE具有代码完成功能，但在某种意义上它是有限的，因为它们完全基于相应编程语言的编译器功能。特定领域和种类的软件可能已经用一组专用的框架和库（沿着其他面向领域的技术）开发，并且因此由某个开发团队用某组API开发。这个问题对ESTI来说更为重要，因为他们还需要在日常活动中处理由第三方开发团队开发的不同领域和品种的客户特定软件。ESTI的开发人员经常面临困难，并花费大量时间手动寻找给定上下文的适当代码完成，他们主要是在临时基础上这样做。我们的研究工作将解决该公司面临的日益增长的需求这一重要问题，并将通过考虑主题系统中已经存在的内容（即，现有的源代码示例），并考虑开发人员的上下文。这种自动化代码完成支持可能会提高开发人员的生产力，使他们更容易使用第三方客户端软件，避免使用已经使用的框架和库引入外部依赖关系，并产生更容易理解的代码。因此，这将有助于提高产品质量和服务，以及公司的整体收入。

As one of the oldest IT companies in Western Canada, ESTI not only builds custom software for its clients, but also provides reengineering and professional services to the existing software of its clients. As such technical leads and developers at ESTI have to deal with software of different varieties, domains and most importantly, software developed by third party development teams. During development the developers heavily rely on various frameworks and libraries and use their corresponding APIs (application program interfaces). The problem is that APIs are hard to learn due to various factors, such as inadequate examples and documentation. Although modern IDEs have a code completion feature, it is limited in the sense that they are solely based on the compiler features of the corresponding programming languages. Software of a particular domain and variety might have been developed with a dedicated set of frameworks and libraries (along with other domain oriented technologies), and thus with a certain set of APIs by a certain development team. This issue is even more crucial for ESTI since they also need to deal with client-specific software of different domains and varieties developed by third party development teams in their daily activities. Developers at ESTI often face difficulties and spend a significant amount of time manually looking for appropriate code completions for a given context and they do this mostly on ad-hoc basis. Our research work will address this non-trivial problem the company is facing with growing need, and will deliver an automatic code completion tool by taking into account what is already present in the subject systems (i.e., existing source code examples) and by considering the context of the developer. Such automated code completion support is likely to increase developers' productivity, make it easier for them to even work with third party client software, avoid introducing external dependencies by using the already used frameworks and libraries in them, and results in more understandable code. This would thus contribute to improved product quality and service, and overall revenue of the company.