Oxygen has been released a few weeks ago. After taking a short summer break, we are starting to think about future directions for Eclipse Code Recommenders.
There are a couple of ideas we’ve mulled over for months already. For this post, I selected two of them because (i) I am very excited about both and (ii) I can’t decide which one to go for first. Thus, I think it’s time to learn what you think about them and help us prioritize… 🙂
Idea #1: Building an AI-powered bug detection
The first idea is pretty easy to explain: Bug detection tools like FindBugs offer rule sets for common Java coding issues only. Hence, they fail to detect misuses for the vast multitude of open-source APIs out there. But API constraints, i.e., contracts on how to use an API, can be inferred from code in a way similar to how Code Recommenders learns which methods you are likely to call next.
So, why not develop a bug detection tool that learns API usage constraints (fully automatically) from code and uses this knowledge to spot bugs other tools can’t find?
Idea #2: Taking Eclipse Code Search to the Web
The second idea is almost as intuitive as the first one—although quite different. Code Recommenders intelligent code completion helps developers to quickly learn how to use an API. But sometimes knowing how likely a call to a method is in a given context is not enough; we need to study the source code ourselves to see the bigger picture and to draw our own conclusions.
Hence, we consider extending Code Recommenders with a (remote) search interface that (i) gives you access to all the code Code Recommenders has analyzed to build its recommendation models, and that (ii) offers a feature set comparable to Eclipse’s built-in (local) code search engine—all tightly integrated into the Eclipse IDE.
Now it’s on you! Shape our roadmap!
What do you think about these two ideas? It would be awesome if you could tell. Below you find two links pointing you to two forms. Please use them and share your thoughts:
Idea #1: Build an AI-powered Bug Detection!
Idea #2: Take Eclipse Code Search to the web!
And if you have a clear preference for one idea, just let us know in the comments!
Thank you very much in advance!
About the author: Dr. Marcel Bruch is the founder of Codetrails and holds a PhD from Darmstadt University of Technology. He leads the Eclipse Code Recommenders project and is responsible for the Automated Error Reporting Initiative at Eclipse.