Griffe

Warnings

• breaking Version 2.0.0 removed several previously deprecated public APIs, including `ExportedName`, `infer_docstring_style`, `parse_auto`, `parse_google`, `parse_numpy`, `parse_sphinx`, `assert_git_repo`, `get_latest_tag`, `get_repo_root`, and `tmp_worktree`.
  Fix: Review the 2.0.0 changelog for a complete list of removed APIs. Replace usage of these deprecated functions/objects with their current equivalents, or remove dependencies on them.
• deprecated The signature of the `on_alias` event changed in version 1.14.0. It changed from `on_alias(self, *, node: AST | ObjectNode, alias: Alias, agent: Visitor | Inspector, **kwargs)` (an analysis event) to `on_alias(self, *, alias: Alias, loader: GriffeLoader, **kwargs)` (a load event).
  Fix: Update custom handlers or extensions that interact with the `on_alias` event to match the new signature `on_alias(self, *, alias: Alias, loader: GriffeLoader, **kwargs)`.
• gotcha Internal packages and modules, particularly those starting with an underscore (e.g., `_griffe`), may be refactored or moved without prior deprecation. Version 1.11.1 moved the private `_griffe` package under `griffe._internal`.
  Fix: Avoid importing from or relying on undocumented internal modules or packages. Stick to the public API exposed directly under the `griffe` top-level module to ensure forward compatibility. If you must use internals, pin your `griffe` version tightly.

Install

• pip install griffe Standard installation
• pip install griffe[pypi] With PyPI index support (for `griffe check`)

Imports

• griffe

  import griffe

  Griffe exposes its public API directly in the top-level 'griffe' module. Importing from internal or private modules (e.g., '_internal' or those starting with an underscore) is not guaranteed to be stable and can break across minor versions, as seen with the '_griffe' package refactoring in 1.11.1.

Quickstart

This quickstart demonstrates how to use `griffe.load` to extract API information from a Python module. It creates a temporary Python file, loads its structure into Griffe's data models, and then prints basic information about the extracted function and class.

import griffe
import os

# Create a dummy Python file for demonstration
with open('my_dummy_module.py', 'w') as f:
    f.write('def my_function(param1: str, param2: int = 0) -> str:\n    """A dummy function."""\n    return f"{param1}-{param2}"\n\nclass MyClass:\n    """A dummy class."""\n    def __init__(self, name: str):\n        self.name = name\n\n    def greet(self) -> str:\n        """Greets by name."""\n        return f"Hello, {self.name}"\n')

# Load the module's API data
# Griffe will try to find sources and fall back to introspection if not found
# Using the current directory as a search path
package = griffe.load("my_dummy_module", search_paths=['.'])

# Access a module, class, or function
print(f"Package name: {package.name}")

my_function = package["my_function"]
print(f"Function: {my_function.name}, Parameters: {[p.name for p in my_function.parameters]}")

my_class = package["MyClass"]
print(f"Class: {my_class.name}, Methods: {[m.name for m in my_class.members.values() if m.is_function]}")

# Clean up the dummy file
os.remove('my_dummy_module.py')