Adding Typings

Warning

This section is still a work in progress.

Adding type hints to functions and parameters

Manim is currently in the process of adding type hints into the library. In this section, you will find information about the standards used and some general guidelines.

If you’ve never used type hints before, this is a good place to get started: https://realpython.com/python-type-checking/#hello-types.

Typing standards

Manim uses mypy to type check its codebase. You will find a list of configuration values in the mypy.ini configuration file. To be able to use the newest typing features not available in the lowest supported Python version, make use of typing_extensions.

To be able to use the new Union syntax (|) and builtins subscripting, use the from __future__ import annotations import.

Typing guidelines

  • Manim has a dedicated typing module where type aliases are provided. Most of them may seem redundant, in particular the ones related to numpy. This is in anticipation of the support for shape type hinting (related issue). Besides the pending shape support, using the correct type aliases will help users understand which shape should be used.

  • For typings of generic collections, check out this link.

  • Always use a type hint of None for functions that does not return a value (this also applies to __init__), e.g.:

def height(self, value) -> None:
    self.scale_to_fit_height(value)
  • For variables representing paths, use the StrPath or StrOrBytesPath type alias defined in the typing module.

  • *args and **kwargs shouldn’t be left untyped (in most cases you can use Any).

  • Following PEP 484, use float instead of int | float.

  • Use x | y instead of Union[x, y]

  • Mobjects have the typehint Mobject, e.g.:

def match_color(self, mobject: "Mobject"):
    """Match the color with the color of another :class:`~.Mobject`."""
    return self.set_color(mobject.get_color())
  • Always parametrize generics (list[int] instead of list, type[Any] instead of type, etc.). This also applies to callables.

rate_func: Callable[[float], float] = lambda t: smooth(1 - t)
  • Use TypeVar when you want to “link” type hints as being the same type. Consider Mobject.copy, which returns a new instance of the same class. It would be type-hinted as:

T = TypeVar("T")


def copy(self: T) -> T: ...
  • Use typing.Iterable whenever the function works with any iterable, not a specific type.

  • Prefer numpy.typing.NDArray over numpy.ndarray

import numpy as np

if TYPE_CHECKING:
    import numpy.typing as npt


def foo() -> npt.NDArray[float]:
    return np.array([1, 0, 1])
  • If a method returns self, use typing_extensions.Self.

if TYPE_CHECKING:
    from typing_extensions import Self


class CustomMobject:
    def set_color(self, color: ManimColor) -> Self:
        ...
        return self
  • If the function returns a container of a specific length each time, consider using tuple instead of list.

def foo() -> tuple[float, float, float]:
    return (0, 0, 0)
  • If a function works with a parameter as long as said parameter has a __getitem__, __iter___ and __len__ method, the typehint of the parameter should be collections.abc.Mapping. If it also supports __setitem__ and/or __delitem__, it should be marked as collections.abc.MutableMapping.

  • Typehinting something as object means that only attributes available on every Python object should be accessed, like __str__ and so on. On the other hand, literally any attribute can be accessed on a variable with the Any typehint - it’s more freeing than the object typehint, and makes mypy stop typechecking the variable. Note that whenever possible, try to keep typehints as specific as possible.

  • If objects are imported purely for type hint purposes, keep it under an if typing.TYPE_CHECKING guard, to prevent them from being imported at runtime (helps library performance). Do not forget to use the from __future__ import annotations import to avoid having runtime NameError exceptions.

from typing import TYPE_CHECKING

if TYPE_CHECKING:
    from manim.typing import Vector3D
# type stuff with Vector3D

Missing Sections for typehints are: