typing-arguments

Typing arguments using the Generic base class in python are great, but they lack the ability to easily access the type arguments at runtime. This library provides a mixin class that can be used to make type arguments available in the class and its instances.

This is also true for classes based on pydantic.BaseModel for pydantic > 2.x.

Installation

Just use pip install typing-arguments to install the library.

Note: typing-arguments is tested on Python 3.10, 3.11, 3.12, 3.13 and 3.14. This is also ensured running all tests on all those versions using tox.

Usage

Quick Example

T1 = TypeVar("T1")
T2 = TypeVar("T2", bound="SomeBaseClass")


class Something(
    GenericArgumentsMixin,
    Generic[T1, T2],
):
    t1 = typing_arg(T1)
    t2 = typing_arg(T2)


ConcreteClass = Something[str, SomeBaseClassChild]
ConcreteClass.t1  # str
ConcreteClass.t2  # SomeBaseClassChild

Hint: You may also use this with pydantic models:

T1 = TypeVar("T1")
T2 = TypeVar("T2", bound="SomeBaseClass")


class SomethingModel(
    GenericArgumentsMixin,
    BaseModel,
    Generic[T1, T2],
):
    t1: ClassVar = typing_arg(T1)
    t2: ClassVar = typing_arg(T2)


ConcreteClassModel = SomethingModel[str, SomeBaseClassChild]
ConcreteClassModel.t1  # str
ConcreteClassModel.t2  # SomeBaseClassChild

Internally GenericArgumentsMixin will create a new attribute __typing_arguments__ inside the class and its instances. This attribute is a dictionary mapping the type variables to their concrete types. This is useful if you want to access the type arguments in a generic way.

The typing_arg function is a helper function to make the type arguments available in the class and its instances using a nicely named attribute. This is just a convenience function, as you can also access the type arguments directly from the __typing_arguments__ attribute.

Note: If you are using pydantic models you should use the ClassVar annotation to ensure pydantic will not try to catch and validate the type arguments as normal model fields.

You may also mix different generic base classes like so:

T1 = TypeVar("T1")
T2 = TypeVar("T2", bound="SomeBaseClass")


class Base1(
    GenericArgumentsMixin,
    Generic[T1],
):
    pass


class Base2(
    GenericArgumentsMixin,
    Generic[T2],
):
    t2 = typing_arg(T2)


class Something(
    Base1[str],
    Base2[SomeBaseClassChild],
):
    t1 = typing_arg(T1)


Something.t1  # str
Something.t2  # SomeBaseClassChild

In this example Base1 and Base2 are both generic base classes. Base1 only defines a type argument T1 and Base2 only defines a type argument T2. The Something class inherits from both Base1 and Base2. Note that Base1 does not define a simple accessor like t1 using typing_arg, while Base2 does. This is not a problem and can be later added by Something using typing_arg as well.

You may encounter issues using the typing_arg function when using type validator like mypy or your IDE. If so you might need to use cast to tell the type checker you are sure about what you are doing. For example:

T1 = TypeVar("T1", bound="SomeBaseClass")


class Something(
    GenericArgumentsMixin,
    Generic[T1],
):
    t1 = cast(type[SomeBaseClass], typing_arg(T1))

Note: You will still need to use ClassVar when using pydantic models. This might result in using the same type twice (inside ClassVar and cast).