RegMiner 2.0

A Mining Approach towards Automating Regression Dataset Construction from Code Evolution History. This project is powerd by SE lab in Fudan University

Regressions4J

We also provide UI-free tools to help users replay test results for each regression bug. Get that support in the regs4J project.

Tutorial

Env requirements for RegMiner:

1. OS: MacOS/Ubuntu/CentOS

2. JDK: 11

3. Python: 3.0+

Attention! We have discovered that RegMiner does not run well on Windows OS, and currently, we don't plan to fix it.

Env requirements for mining projects:

1. JDK LTS: 1.7,1.8,11,17
2. Maven or Gradle
3. Others*: The specific environment required by a particular project. For example, MongoDB may be necessary for a project related to MongoDB middleware.

Easy Start:

Steps are as follows:

(1) Prepare the data.

1. Create a workspace for Miner. mkdir miner_space

2. Create a directory for maintaining the source code of the mined project.

  cd miner_space
  mkdir meta_projects

3. Prepare source code of the mined project. In the example, we use uniVocity/univocity-parsers.

  cd meta_projects
  git clone https://github.com/uniVocity/univocity-parsers.git 

(2) Generate the configuration.

1. Modify the configuration related to the database in scripts/gen_config.py.

sql_enable=0

This modification will disable the MySQL functionality of RegMiner. We currently do not provide the SQL table structure. If you need it, please contact Xuezhi.

3. Generate the configuration file.

cd /xxx/xxxx/RegMiner/scripts
rm env.properties
python gen_config.py
mv env.properties ../

These commands will generate the necessary configuration for running Regminer. The script will automatically detect the JDK installed on the system.

3. Load RegMiner Project to IDEA.
4. Config debug params.

-ws /xxx/xxx/miner_space/ -pj univocity-parsers -cfg env.properties -t bfc

-t bfc means just mine bfc; -t bfc&bic means search for regressions.

5. Run RegMiner in miner/src/main/java/org/regminer/miner/start/MinerCli.java, you can find progress info in logs/app.log

Automate batch mining.

Automated mining of target project sets.

1. Build the JAR package for RegMiner and place it in the scripts directory, and move scripts directory to miner_space.
2. Run python gen_config.py under scripts directory.
3. Prepare a list of project names in file projects.in under scripts directory, with one project per line.
4. Clone the source code of these projects with .git files into miner_space/meta_projects.
5. Run the run_multi_projects.py under scripts directory to mine the projects in the list.

You can see the comments in the scripts for more details.

Paper

1. Xuezhi Song, Yun Lin*, Siang Hwee Ng, Yijian Wu, Xin Peng, Jin Song Dong and Hong Mei. RegMiner: Towards Constructing a Large Regression Dataset from Code Evolution History [⚡ CCF Prototype Competition Award (1st Prize)]

2. Xuezhi Song, Yun Lin*, Yijian Wu, Yifan Zhang, Xin Peng, Jin Song Dong, and Hong Mei. RegMiner: Mining Replicable Regression Dataset from Code Repositories

3. Xuezhi Song, Yijian Wu*, Junming Cao, Bihuan Chen, Yun Lin, Zhengjie Lu, Dingji Wang, Xin Peng. BugMiner: Automating Precise Bug Dataset Construction by Code Evolution History Mining

Acknowledgments

We express our great thanks for JetBrains support on RegMiner.