ghtest

Coverage-Guided Test Generator for Python

ghtest is a test generation tool that automatically creates regression tests for your Python code. It combines static analysis with coverage-guided feedback to generate (hopefully) high-quality tests that capture the current behavior of your application.

Key Features

• Coverage-Guided Generation: Iteratively generates tests to target missed branches and edge cases.
• VCR Integration: Automatically records and replays external API interactions (like GitHub API calls) using vcrpy, making tests deterministic and offline-capable.
• Scenario Support: Detects CRUD patterns and generates stateful scenario tests (e.g., create -> read -> delete).
• Smart Parameter Suggestion: Analyzes function signatures and type hints to suggest realistic test inputs.
• Portable Tests: Generates standalone pytest files with relative paths, ready to be committed to your repository.
• Safe Execution: Includes an interactive mode to prevent accidental execution of destructive functions (like delete_repo) during test generation. Nevertheless, it is stringly recommended to run ghtest in a safe environment.

Installation

pip install ghtest

Usage

Run ghtest on your source directory:

ghtest src/my_project

This will:

1. Scan your code for functions.
2. Generate initial tests.
3. Run them to capture behavior (recording cassettes).
4. Analyze coverage to find missed branches.
5. Generate targeted tests to fill coverage gaps.
6. Output ready-to-run test files in tests/.

Options

Example

Generate tests for the current directory, asking for permission before running anything risky:

ghtest . --interactive --verbose

How It Works

1. Static Analysis: ghtest parses your code to understand function signatures, types, and dependencies.
2. Parameter Suggestion: It suggests input values based on types and common patterns.
3. Execution & Capture: It runs the functions with the suggested inputs.
   • Return values and exceptions are captured.
   • File I/O is tracked.
   • Network requests are recorded into VCR cassettes.
4. Coverage Feedback: It measures code coverage and identifies unexecuted branches (e.g., if x > 5:). It then solves for inputs that would trigger those branches and generates new tests.
5. Code Generation: Finally, it writes out standard Python test files that assert the captured behavior matches the replay.

License

MIT