Self-Improving IDE 2.0?

Code Recommenders Propose IDE With Predictive Text

Jessica Thornsby

Code Recommenders is proposed as an IDE that learns how APIs are used by their clients.

Code Recommenders has been proposed under the Eclipse Technology Container Project. This project will provide tools that automatically analyse large code repositories, extract various data and integrate this information into a context-sensitive IDE, ready to be reused by the developer. The IDE and tools will continuously improve themselves by leveraging implicit and explicit knowledge about how APIs are used by their clients. It is also under consideration whether this ecosystem should leverage the existing Eclipse Marketplace and P2, to scan and update API usages. The project proposal calls this “a collaborative way of sharing knowledge over the IDE,” (also called the “IDE 2.0” in the proposal.) Currently, these systems are fed more or less manually by an administrator, but it is planned that this will graduate to a community-driven process.

Initially, the project will provide additional templates for frequently reoccurring code snippets, a stacktrace search engine, and tools for evaluating API misuse. The initial contribution will come from the Intelligent Code Completion tool, which works by examining how developers use Text widgets in their code, creates a database of the possible usages, builds an intelligent code completion engine and presents its findings to the user. The Extended, usage-driven Javadoc will also be used in the initial contribution.

The project is developed at Darmstadt University of Technology, with Marcel Bruch as lead. Chris Aniszczyk is listed as a mentor.

Inline Feedbacks
View all comments