Grant: Collaborative Research and Development Grants

Software developers often have problems in making effective use of new and changing technologies. When they reach a point when the use of a new technology, such as a library or framework, is required in their development task, they often must pause what they are doing, and learn about both the abstract concepts and the detailed API usage from external sources. This context switch can be a significant productivity drain to a developer doing creative design work. With strongly-typed languages such as Java, tools such as autocompletion can be built into the integrated development environment (IDE) to help developers recall API details with little disruption. For dynamically-typed languages such as Javascript, it is much harder to anticipate the likely usage scenario, since the type information of the variables can be hard to infer. In this work, we look to exploit history, by mining both previous usage of the APIs in question from existing code bases, and by processing curated API documentation to search for hints for usage patterns. Using machine learning, natural language processing, and data mining techniques, we will build recommendation engines that can be integrated into the IDE to provide help to developers working with dynamic programming languages, libraries, and frameworks. In this way, developers will encounter fewer interruptions in their work flow, and spend less time reading through API documentation, allowing them to better focus on their development task at hand.*