DefectGuard: An Integrated, Efficient and Effective Tool for JIT Defect Prediction

About

• DefectGuard is a python package
• Basic functionalities:
  • Mining commits from Git repositories
  • Post-processing, training, inferencing JITDP model via CLI or import library
• DefectGuard had been integrated into VSC (extension), Jenkins & GitHub Action (via command)

Installation

via Docker (RECOMMENDED)

Please checkout #2 in TROUBLESHOOT.md if you do not have GPU(s) on your machine

docker compose up --build -d
docker exec -it defectguard /bin/bash

Inside docker container:

# This setup pyszz and defectguard
bash scripts/setup.sh

If you want Docker container to access GPU(s), please download nvidia-container-toolkit

Note: download this outside of the container

Install the nvidia-container-toolkit package as per official documentation at Github.

We also provide a quick-run script for Debian-based OS

From scratch

• SrcML

DefectGuard requires PySZZ for mining data functionality. SrcML is required by PySZZ. Please install it before mining data.

# Install libarchive13 libcurl4 libxml2
sudo apt-get install libarchive13 libcurl4 libxml2

# Install libssl
wget http://archive.ubuntu.com/ubuntu/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_amd64.deb && \
    dpkg -i libssl1.1_1.1.1f-1ubuntu2_amd64.deb && \
    rm -rf libssl1.1_1.1.1f-1ubuntu2_amd64.deb

# Install SrcML
wget http://131.123.42.38/lmcrs/v1.0.0/srcml_1.0.0-1_ubuntu20.04.deb && \
    dpkg -i srcml_1.0.0-1_ubuntu20.04.deb && \
    rm -rf srcml_1.0.0-1_ubuntu20.04.deb

• PySZZ

DefectGuard requires this external tool for mining data functionality. Please install it before mining data.

git clone https://github.com/grosa1/pyszz_v2.git

• Dependencies

Please checkout #2 in TROUBLESHOOT.md if you do not have GPU(s) on your machine

Check out this requirements.txt or this cpu-only-requirements.txt

pip install -r requirements.txt

• Setup DefectGuard

python setup.py develop

Basic usages

Mining commits from Git repositories

defectguard mining \
    -repo_name <project_name> \
    -repo_path <path/to/project> \
    -repo_language <main_language_of_project> \
    -pyszz_path <path/to/project/pyszz_v2>

Example

Training

defectguard training \
    -model <model_name> \
    -repo_name <project_name> \
    -repo_language <main_language_of_project> \
    -epochs <epochs>

Example

Fine-tuning

defectguard training \
    -model <model_name> \
    -from_pretrain \
    -repo_name <project_name> \
    -repo_language <main_language_of_project> \
    -epochs <epochs>

Example

Evaluating

defectguard evaluating \
    -model <model_name> \
    -repo_name <project_name> \
    -repo_language <main_language_of_project>

Example

Evaluating from our pretrain models

defectguard evaluating \
    -model <model_name> \
    -from_pretrain \
    -repo_name <project_name> \
    -repo_language <main_language_of_project>

Inference

Comming Soon

Integrate into CLI-like Continuous Integration

Comming Soon

Output of DefectGuard

Folder's structure

.
├── dg_cache
│   ├── dataset // default folder for saving dataset
│   ├── save // default folder for saving extracted data
│   ├── repo // default folder for cloning github repository

Extracted Data's folder structure

A sample structure of extracted data:

.
├── save
|   ├── repo_name
|   |   ├── commit_ids.pkl
|   |   ├── etracted_info.json // the config for Extractor
|   |   ├── repo_bug_fix.json // the bug_fix file for running PySZZ
|   |   ├── repo_commits_{num}.pkl // files storing commits information
|   |   ├── repo_features.pkl // files storing commits features

Processed Data's folder structure

A sample structure of processed data:

.
├── dataset
|   ├── repo_name
|   |   ├── commits
|   |   |   ├── cc2vec.pkl
|   |   |   ├── deepjit.pkl
|   |   |   ├── simcom.pkl
|   |   |   ├── dict.pkl
|   |   ├── features
|   |   |   ├── feature.csv

Repository's structure

In case this tool is run on mode="local", please follow this repository's structure paths:

.
├── repo_path
|   ├── repo_owner
|   |   ├── repo_name
|   |   |   ├── .git
|   |   |   ├── other repo content

Troubleshoot

Find here: https://github.com/manhtdd/DefectGuard-the-Package/blob/main/TROUBLESHOOT.md