BuGL -- A Cross-Language Dataset for Bug Localization
This dataset addresses a gap for researchers needing to investigate bug localization across multiple programming languages, though it is incremental as it builds on existing single-language datasets.
The authors tackled the lack of a cross-language dataset for bug localization by creating BuGL, a large-scale dataset with over 10,000 bug reports from open-source projects in C, C++, Java, and Python.
Bug Localization is the process of locating potential error-prone files or methods from a given bug report and source code. There is extensive research on bug localization in the literature that focuses on applying information retrieval techniques or machine learning/deep learning approaches or both, to detect location of bugs. The common premise for all approaches is the availability of a good dataset, which in this case, is the standard benchmark dataset that comprises of 6 Java projects and in some cases, more than 6 Java projects. The existing dataset do not comprise projects of other programming languages, despite of the need to investigate specific and cross project bug localization. To the best of our knowledge, we are not aware of any dataset that addresses this concern. In this paper, we present BuGL, a large-scale cross-language dataset. BuGL constitutes of more than 10,000 bug reports drawn from open-source projects written in four programming languages, namely C, C++, Java, and Python. The dataset consists of information which includes Bug Reports and Pull-Requests. BuGL aims to unfold new research opportunities in the area of bug localization.