pybughunt 0.2.0

A Python library for detecting logical and syntactical errors in Python code

PyBugHunt

Advanced Python Code Error Detection and Analysis

PyBugHunt is a sophisticated Python library designed to detect, analyze, and suggest fixes for both syntactical and logical errors in Python code. Leveraging a combination of static code analysis techniques and advanced transformer-based machine learning models, PyBugHunt offers developers a powerful tool to improve code quality and reduce debugging time.

---

Table of Contents

• Features
• Technology Stack
• Project Structure
• Installation
• Usage
  • Command Line Interface
  • Python API
• Error Detection Capabilities
• Machine Learning Approach
• Development
• License
• Contributing

---

Features

PyBugHunt offers comprehensive error detection capabilities:

• Robust Syntax Error Detection and Analysis
• Intelligent Logical Error Detection using both static analysis and machine learning.
• Transformer-Based Models:
  • CodeBERT: For classifying code as correct or containing a logical error.
  • T5 (Text-to-Text Transfer Transformer): For generating natural language descriptions of the detected errors.
• Fix Suggestion System
• Flexible Integration Options (CLI and Python API)
• Customization and Training of models.

---

Technology Stack

PyBugHunt utilizes a wide range of technologies and libraries:

Core Technologies

• Python 3.8+
• Abstract Syntax Tree (AST)
• Python Standard Library

Machine Learning

• PyTorch
• Hugging Face Transformers (for CodeBERT and T5)
• scikit-learn
• NumPy

Static Analysis

• Astroid
• PyLint

---

Project Structure

pybughunt/
├── src/
│   └── pybughunt/
│       ├── __init__.py
│       ├── cli.py
│       ├── detector.py
│       ├── logic_analyzer.py
│       ├── syntax_analyzer.py
│       └── models/
│           ├── __init__.py
│           ├── model_loader.py
│           ├── model_trainer.py
│           └── models.py  # New file for transformer model definitions
├── tests/
├── .gitignore
├── README.md
├── pyproject.toml
└── setup.py

---

Installation

From Source

# Clone the repository
git clone https://github.com/Preksha-7/pybughunt.git
cd pybughunt

# Install in development mode
pip install -e .

Dependencies

All dependencies will be automatically installed. The main dependencies are listed in pyproject.toml and setup.py.

---

Usage

Command Line Interface

Analyze a file with the default static analysis:

python -m pybughunt.cli analyze src/pybughunt/sample_buggy.py --model_type static

Analyze a file using a specific machine learning model:

python -m pybughunt.cli analyze src/pybughunt/sample_buggy.py --model_type codebert --model_path path/to/saved_codebert_model

Train a new model:

python -m pybughunt.cli train --dataset /path/to/python/files --output my_model --model_type codebert

Python API

from pybughunt import CodeErrorDetector

# Initialize the detector with a specific model
detector = CodeErrorDetector(model_type='codebert', model_path='path/to/saved_codebert_model')

code = '''
def incorrect_factorial(n):
    if n == 0:
        return 1
    else:
        return incorrect_factorial(n-1) # Missing multiplication with n
'''

results = detector.analyze(code)
print(results)

---

Error Detection Capabilities

Syntax Errors: Missing delimiters, indentation issues, invalid syntax, etc.

Logical Errors:

• Static Analysis: Infinite loops, unused variables, off-by-one errors, division by zero, unreachable code.
• Machine Learning: More subtle logical errors detected by the trained transformer models.

---

Machine Learning Approach

PyBugHunt now includes transformer-based models for more advanced logical error detection:

CodeBERT (microsoft/codebert-base): A model pre-trained on a large corpus of code, used for classifying code snippets as either correct or containing a logical error.

T5 (t5-small): A sequence-to-sequence model that can be trained to generate a natural language description of the error in a piece of code.

These models can be trained on your own dataset using the train command in the CLI.

---

Development

The project is structured to be modular and extensible. You can add new error detection patterns to the logic_analyzer.py or experiment with different models in the models/ directory.

---

License

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

---

Contributing

Contributions are welcome! Please feel free to submit a pull request.