dataclasses¤

The dataclasses extension adds support for dataclasses from the standard library. It works both statically and dynamically. When used statically, it re-creates the __init__ methods and their signatures (as Griffe objects), that would otherwise be created at runtime. When used dynamically, it does nothing since __init__ methods are created by the library and can be inspected normally.

Example:

from dataclasses import dataclass


@dataclass
class Room:
    uid: int
    name: str
    capacity: int = 10
    available: bool = True

With the dataclasses extension enabled, the Griffe object for the Room class will get an __init__ method with the following signature:

def __init__(self, uid: int, name: str, capacity: int = 10, available: bool = True) -> None:
    ...

Additional metadata like ClassVar, the init and kw_only parameters, or the KW_ONLY sentinel are also recognized and will update the __init__ method signature accordingly.

This extension is enabled by default. It is always added last. If you need to give it a higher priority, you can explictly enable it to change its position in the list of extensions (it will run only once):

$ griffe dump -e dataclasses,other my_package

import griffe

my_package = griffe.load("my_package", extensions=griffe.load_extensions("dataclasses", "other"))

plugins:
- mkdocstrings:
    handlers:
      python:
        options:
          extensions:
          - dataclasses
          - other