This is the official repository of paper "Fault Localization from the Semantic Code Search Perspective".

In this work, we present CosFL, a novel Fault Localization (FL) approach inspired by semantic Code Search (CS). CosFL proposes to regard FL tasks as a two-step process: the query generation for generating a natural language query to describe the problematic software functionalities, and the suspicious code search which employs CS to directly match buggy program elements from the entire code repository. Evaluation results on 835 real bugs show that CosFL significantly outperforms other state-of-the-art approaches, potentially paving new pathways for the advancement of FL techniques.

We evaluate CosFL on two Java benchmarks, they can be installed at:

Defects4J: https://github.com/rjust/defects4j

GrowingBugs: https://github.com/liuhuigmail/GrowingBugRepository

You can easily install the required Python environment through the commands:

conda create -n CosFL python=3.9.18
conda activate CosFL
pip install -r requirements.txt

Don't forget to install Java 1.8:

sudo apt-get install openjdk-8-jdk -y

Then, adopt the files under BenchmarkMode directory to support program instrumentation for the benchmark commands.

Finally, modify the config files in Config directory. You should focus on these properties:

• agent_lib: set as the path of the Java agent lib/classtracer-1.0.jar
• D4J_exec: set to your own Defects4J execution file path
• GB_exec: set to your own GrowingBugs execution file path
• api_key: set to your own api key for corresponding model

You can now directly run CosFL to localize all bugs with:

python run_all.py

For localizing a single bug, an example command is:

python run.py --config default.yaml --version d4j1.4.0 --project Chart --bugID 1 --subproj 

You can see the list of all bugs in projects.py.

After running CosFL, the debugging results will be put under DebugResult directory.

To evaluate the performance of CosFL, simply run:

python Evaluation/evaluate.py