Bimodal program analysis in the wild

Abstract

The combination of standard program analyses with machine learning-based techniques has recently been a hot research topic, but is not yet widely adopted by software engineering practitioners. We present such an analysis, aimed at identifying and correcting misleading and uninformative variable names, and discuss the challenges arising from attempting to deploy it in practice.

Bio

Marc Brockschmidt is a researcher at Microsoft Research, Cambridge. His research is focussed on using automatic analyses to make software engineers more productive by identifying bugs as early as possible.