• Corpus ID: 237635019

Broccoli: Bug localization with the help of text search engines

@article{Ledel2021BroccoliBL,
  title={Broccoli: Bug localization with the help of text search engines},
  author={Benjamin Ledel and Steffen Herbold},
  journal={ArXiv},
  year={2021},
  volume={abs/2109.11902}
}

• Benjamin Ledel, S. Herbold
• Published 24 September 2021
• Computer Science
• ArXiv

Bug localization is a tedious activity in the bug fixing process in which a software developer tries to locate bugs in the source code described in a bug report. Since this process is time-consuming and sometimes requires additional knowledge about the software project, current literature proposes several information retrieval techniques which can aid the bug localization process. However, recent research questioned the state-of-the-art approaches, since they are barely adopted in practical scenarios. In this paper, we introduce a new bug localization approach, called Broccoli, which uses techniques from previous approaches and extends them with a search engine to improve the localization. Primarily, we utilize a search engine to find similar source code locations and prior bug reports that are comparable to the new bug report. We combine these search engine techniques with information retrieval techniques used by previous approaches. In a case study, we evaluate the performance of our search engine approach against seven bug localization algorithms on 82 open source projects in two data sets. In the first data set, we are not able to show a significant difference between Broccoli and Locus. However, the search engine approach has the highest Mean Reciprocal Rank (MRR) and Mean Average Precision (MAP) value on average and exceeds the other algorithm by 7% 77% (MAP) and 6% 108% (MRR). In the second data set, the MAP and MRR value of Broccoli is also higher than the state-of-the-art on average with a better 9% 107% MAP value and 32% 173% better MRR value. However, we are not able to show a significant difference between Broccoli and BRTracer+ in the statistical analysis regarding the MRR metric. Additionally, we identified several issues regarding the version matching in current evaluation methods for bug localization, which leads to the bug localization algorithm not being able to predict the correct files for the bug report. Therefore, we propose a new version matching strategy, called Time-Aware matching, to overcome this issues. Our experiment indicates that the Time-Aware strategy is more accurate than the state-of-the-art, since it can reduce the number of non-detectable files from 22.6% to 13.7% and non-detectable bug reports from 10.8% to 0.3%. These changes in the version matching lead to a significant increases of the MAP and MRR value for Broccoli.