KNOD: Domain Knowledge Distilled Tree Decoder for Automated Program Repair

This is the artifact of paper "KNOD: Domain Knowledge Distilled Tree Decoder for Automated Program Repair, Nan Jiang, Thibaud Lutellier, Yiling Lou, Lin Tan, Dan Goldwasser, and Xiangyu Zhang", ICSE 2023.

Dependency

• Python 3.10.4
• PyTorch >= 1.9.1
• Java 8
• Docker
• nvidia-docker
• Defects4J

To Run in Docker

• To run this artifact in docker, you need to have docker installed first.
• To use GPUs in docker, you need to install nvidia-docker.
• Pull the docker image, which contains a copy of the whole artifact as well as all the dependencies satisfied:

docker pull jiang719/knod:latest

• Start the docker image:

docker run -it --name knod --gpus all jiang719/knod:latest
cd /home/KNOD
bash setup.sh

Content

The file structure of the artifact is as follow:

• data:
  • defects4j_input: model input files of Defects4J benchmark
  • defects4j_output: generated patches for Defects4J benchmark
  • quixbugs_input: model input files of QuixBugs benchmark
  • quixbugs_output: generated patches for QuixBugs benchmark
  • vocabulary: vocabulary files for building model
  • example_training_ast.json: example training samples
  • example_validation_ast.json: example validation samples
  • jdk.json: static analysis of JDK-8
• javalang: modified javalang package by adding API to convert an AST object to source code
• javaparser: a Java tool to parse Java files and prepare input for models.
• results: KNOD's correct fixes on Defects4J v1.2, Defects4J v2.0 and QuixBugs benchmarks.
• src: source code to run KNOD to generate patches or train KNOD models.
• QuixBugs: temprary folders for validating QuixBugs bugs.
• preprint_for_artifact_evaluation.pdf: An early draft of the preprint for artifact evaluation purpose only (to be updated).

KNOD's patches

The correct/plausible patches KNOD generated for Defects4J v1.2, Defects4J v2.0 and QuixBugs benchmarks are under the results folder, which refers to result of RQ1 in the paper.

To train KNOD models

cd src
python train.py

There are several configurations you may want to modify: training_data_path is the path to the training data, validating_data_path is the path to the validation data, and save_dir is the folder to save the models (default set to ../data/models/).

Our training data is shared atvia Zenodo at https://doi.org/10.5281/zenodo.7570475.

To generate patches for Defects4J

cd src
python prepare_defects4j_input.py
python generate_defects4j_output.py

To generate patches for QuixBugs

cd src
python prepare_quixbugsj_input.py
python generate_quixbugs_output.py

Citation

If you find this code to be useful for your research, please consider citing:

@inproceedings{jiang@domain,
   author = {Jiang, Nan and Lutellier, Thibaud and Lou, Yiling and Tan, Lin and Goldwasser, Dan and Zhang, Xiangyu},
   title = {KNOD: Domain Knowledge Distilled Tree Decoder for Automated Program Repair},
   year = {2023},
   isbn = {9781665457019},
   publisher = {IEEE Press},
   url = {https://doi.org/10.1109/ICSE48619.2023.00111},
   doi = {10.1109/ICSE48619.2023.00111},
   booktitle = {Proceedings of the 45th International Conference on Software Engineering},
   pages = {1251–1263},
   numpages = {13},
   keywords = {abstract syntax tree, deep learning, automated program repair},
   location = {Melbourne, Victoria, Australia},
   series = {ICSE '23}
}