Mining Source Code Descriptions from Developer Communications
Sebastiano Panichella,Jairo Aponte, Massimiliano Di Penta,Andrian Marcus,Gerardo Canfora
Dept. of Engineering-RCOST, University of Sannio, Italy
Universidad Nacional de Colombia, Bogota, Colombia
Wayne State University, Detroit USA
Problem:
Absence of comments/descriptions for methods.
Importance/Applications of the technique:
Automatic mining of source code descriptions from bug tracking systems and mailing lists.
Source code re-documentation.
Approach:
1. Downloading emails and tracing them onto classes- using 2 heuristics- if email contains fully qualified class name or file name. For bug ids-matching bug id to closing comments.
2. Extracting paragraphs based on programming language keyword/operator density.
3. Tracing paragraphs onto method based on occurrence of keyword “method” and a method name followed by parenthesis.
4. Filtering paragraphs further by - return types, over-ridding/overloading and method invocations(invoking a method inside another method).
5. Computing textual similarities between paragraphs and methods and ranking them based on similarity measure.
Evaluation:
Critique
