KBL: A Golden Keywords-based Query Reformulation Approach for Bug Localization

Supplementary Results Display

• The detailed statistical significance results mentioned in the paper are shown in the significance_res.csv file.
• Other attempts to improve the classifier are also shown in the other_attempts.csv file.

Getting Started

Clone the repository by using the following command.

git clone https://github.com/Caiby0927/KBL.git

If you don't have git, please install git first using following commands.

$ sudo apt-get update $ sudo apt-get install git

Install Python

We use python 3.8

$ sudo add-apt-repository ppa:fkrull/deadsnakes

$ sudo apt-get update

$ sudo apt-get install python3.8 python

$ sudo apt-get install python-pip

Install python libraries

We have 5 dependencies below:

lightgbm >= 3.3.2
pandas >= 1.1.4
numpy >= 1.13.3
scipy >= 0.19.1
nltk >= 3.7

Execution

Step1: Download the bug reports from the bug tracking systems of Eclipse foundations The bug reports of Eclipse foundation could be downloaded from https://bugs.eclipse.org/

Step2: Obtain the preliminary keywords for bug reports After getting the bug report, you can use the code in the GA repository (also written by us) to get the preliminary keywords corresponding to each bug report.

Step3: Refine the preliminary keywords With the preliminary keywords, we can use the keyword_refinement.java, which is also in the GA repository, to refine them. After refinement, we can get the golden keywords for each bug report.

Step4: Extract features for terms in the bug report To build the keywords classifier, we first need to extract features from each term in the bug report. We can use the scripts call bug_semantic.py, code_semantic.py and camel_case.py to extract all 61 features for the term.

Step5: Build the keywords classifier Before training the classifier, we need to label the terms which are used for training and testing. Here, we use the golden keywords to label each term. Then, we can use the script train.py to build the keywords classifier.

Step6: Predict keywords for new bug report After building the classifier, we can input the new bug report and use the trained classifier to predict the keywords of the bug report with the script call eval.py.