LLM4ModuleSum

This repository contains the code for the paper "Commenting Higher-level Code Unit: Full Code, Reduced Code, or Hierarchical Code Summarization".

Overview

The goal of this project is to explore different approaches to generating comments for higher-level code units such as modules or classes. This includes working with:

• Full Code: Using the complete code for generating summaries.
• Reduced Code: Leveraging compressed or abstracted versions of the code.
• Hierarchical Code Summarization: Creating summaries at different levels of granularity.

Table of Contents

• Installation
• Usage
• Project Structure
• Functions and Scripts
• License

Installation

1. Clone the repository:

   git clone https://github.com/wssun/LLM4ModuleSum.git

2. Navigate into the project directory:

   cd LLM4ModuleSum

3. Install the required packages:

   pip install -r requirements.txt

Usage

Data Processing

The scripts located at the root directory are primarily for data processing tasks, including:

• caseslect_file.py: Selects specific cases from the dataset.
• compress.py: Compresses the Java code by removing unnecessary comments and whitespace.
• slicebycommunity.py: Slices the code based on community detection.
• slicebyfunction.py: Slices the code by individual functions.
• understand.py: Analyzes and processes the code using the 'understand' tool.

Comment Generation

The scripts used to interact with large language models (LLMs) for comment generation are organized in two levels:

File Level (Located in file_level/)

• run_all.py: Runs the full process for generating comments on all files.
• run_compression.py: Applies compression techniques before generating comments.
• run_fuc.py: Focuses on function-level summarization.
• run_fuc_summary.py: Generates summaries for each function within the code files.
• run_point.py: Extracts key points from the code for summarization.
• run_point_summary.py: Generates summaries based on key points extracted from the code.

Module Level (Located in module_level/)

• find_package.py: Identifies and processes package-level information in the code.
• module_summary.py: Generates summaries at the module level.
• module_summary_prompt.py: Provides prompts for LLMs to generate module-level summaries.

Example Command

To process files and generate comments:

python file_level/run_all.py

Project Structure

LLM4ModuleSum/
│
├── data/                              # Directory containing JSON and CSV files
├── file_level/                        # File-level comment generation scripts
│   ├── run_all.py
│   ├── run_compression.py
│   ├── run_fuc.py
│   ├── run_fuc_summary.py
│   ├── run_point.py
│   └── run_point_summary.py
├── module_level/                      # Module-level comment generation scripts
│   ├── find_package.py
│   ├── module_summary.py
│   └── module_summary_prompt.py
├── caseslect_file.py                  # Script to select specific cases
├── compress.py                        # Script to compress Java code
├── slicebycommunity.py                # Script to slice code by community
├── slicebyfunction.py                 # Script to slice code by function
├── understand.py                      # Code analysis script using 'understand' tool
├── requirements.txt                   # Required Python packages
└── README.md                          # Readme file

Functions and Scripts

Data Processing Scripts

• caseslect_file.py: Selects specific cases from the dataset.
• compress.py: Compresses the Java code by removing unnecessary comments and whitespace.
• slicebycommunity.py: Slices the code based on community detection.
• slicebyfunction.py: Slices the code by individual functions.
• understand.py: Analyzes and processes the code using the 'understand' tool.

File-Level Comment Generation

• run_all.py: Runs the full process for generating comments on all files.
• run_compression.py: Applies compression techniques before generating comments.
• run_fuc.py: Focuses on function-level summarization.
• run_fuc_summary.py: Generates summaries for each function within the code files.
• run_point.py: Extracts key points from the code for summarization.
• run_point_summary.py: Generates summaries based on key points extracted from the code.

Module-Level Comment Generation

• find_package.py: Identifies and processes module-level information in the code.
• module_summary.py: Generates summaries at the module level.
• module_summary_prompt.py: Provides prompts for LLMs to generate module-level summaries.

License

This project is licensed under the MIT License. See the LICENSE file for details.