Skip to content

Masud Rahman

Monday, 14th January 2019 9:40 am
Where: E2-528 EITC

Improving Bug Localization With Context-Aware, Analytics-Driven, Effective Query Reformulation

Software bugs cost about $312 billion per year globally. In 2017 alone, 606 software bugs cost around $1.7 trillion with 3.7 billion people affected and 314 companies impacted. Detection and fixing such bugs upfront consume 50% of the developers' time and efforts. Traditional practices of software debugging are thus neither cost-effective nor sustainable. The very first step of a debugging process is the localization of a bug within the software code, i.e., bug localization. Bug localization is crucial to the whole debugging process since the later steps entirely depend on successfully locating the bugs. Given a bug report, developers often choose a few important keywords from it and then attempt to find out the location of the bug in source code using keyword/code search. Unfortunately, as existing studies show, they fail 88% of the times to choose the right keywords from the bug reports regardless of their development experience. Furthermore, when they choose the queries on an ad hoc basis, majority of their queries are either poor or noisy which require further reformulations. Thus, developers are badly in need of automatic supports in constructing appropriate queries for the bug localization and even for the general-purpose code search.
Source code and regular texts (e.g., news articles) are significantly different from each other in their structures and underlying semantics. Therefore, searching within a codebase is much more challenging than a regular document search (e.g., web search). This talk will (1) demonstrate why appropriate query construction for code search is a major challenge for the developers, and (2) present a suite of novel, state-of-the-art techniques that circumvent such challenges, and support not only bug localization but also code search in general using automated query reformulations.
This speaker is a candidate for a tenure-track position in the Department of Computer Science.
Complete Seminar List
© 2011 University of Manitoba Department of Computer Science
Back to top