applipy-inject

Applipy Inject

pip install applipy_inject

Library that provides a dependency injector that works with type hinting.

Basic usage

from applipy_inject import Injector

injector = Injector()

# bind dictionary instance
injector.bind(dict, {'v': 143, 'k': 'bar'})

class A:
    def __init__(self, v: int):
        self.v = v

class B(A):
    def __init__(self, d: dict):
        super().__init__(d['v'])
        self.d = d

# bind subtype B to type A
injector.bind(A, B)

class C:
    def __init__(self, a: A):
        self.a = a

# bind type C
injector.bind(C)

# get an instace of type C
c = injector.get(C)

# the instance is initialized correctly, resolving the dependecy chain
print(c.a.v)

Output:

143

bind(...)

The bind() function lets bind an object to a type. The meaning on the binding depends on the type of the object. In general the object can be one of:

• instance: an instance of a type
• type: a type
• provider: a callable object that has type annotations for its arguments and a return type annotation.

The bind function can be used in multiple ways:

bind(type, instance)

Bind an instance to a type. The instance must be an instance of the type or of a subtype of the type.

injector.bind(int, 5)

bind((typeA, typeB, ...), instance)

Bind an instance to multiple types. The instance must be an instance of all the types or of a subtype of all the types.

injector.bind((str, object), 'hello')

bind(type)

The type is bound as a "provider" of its own type and dependecy annotations are taken from the __init__ function.

injector.bind(A)

bind(type, subtype)

Similar to bind(type) but subtype is bound to the specified type.

injector.bind(A, B)

bind((typeA, typeB, ...), subtype)

Similar to bind(type, subtype) but subtype is bound to the specified types.

injector.bind((A, B), B)
# or
injector.bind((A, B), C)

bind(provider)

The provider will be bound to the type it returns (as per the type annotation).

def provide_A(s: str) -> A:
    return A(int(s))

injector.bind(provide_A)

bind(type, provider)

Similar to bind(provider) but the provider is bound to the specified type. The annotated return type of the provider must be a subtype of the specified type.

def provide_B(v: int, k: str) -> B:
    return B({'v': v, 'k': k})

injector.bind(A, provide_B)

bind((typeA, typeB, ...), provider)

Similar to bind(type, provider) but the provider is bound to all the specified types. The annotated return type of the provider must be a subtype of all the specified types.

def provide_B(v: int, k: str) -> B:
    return B({'v': v, 'k': k})

injector.bind((A, B), provide_B)

Common optional parameters:

• name: defaults to None. Lets the user give a name to the binding. This allows to make multiple bindings to the same type and be able to select which one you want to get by using the name.

• singleton: defaults to True. Define whether the injector should instantiate or call the provider only once and inject always the same instance or return a new result every time. It does not applipy to bound instances.

injector.bind(provide_A, name='foo', singleton=False)

get(...)

Get an instance registered to a given type.

Similar to bind(), there is an optional name parameter that tells the injector the name of the instance to get for that type.

injector.get(A, name='foo')

get_all(...)

The Injector allows to bind multiple objects to the same type and name. get_all() retrieves all instances for a given type and name combination as a list, instead of just one as get() does.

Similar to bind(), there is an optional name parameter that tells the injector the name of the instance to get for that type.

injector.get_all(A, name='foo')

Named dependencies

Dependencies can be given names so that different providers can depend on different instances of the same type. This can be achieved by annotating the dependency type with the dependency name:

from typing import Annotated
from applipy_inject import name


def provide_A(s: Annotated[str, name('foo_num')]) -> A:
    return A(int(s))


injector.bind(str, '13', name='foo_num')
injector.bind(provide_A)

Utility functions

with_names(provider, names)

Give names to the arguments of an existing provider or class. The injector will use this to set the value for name with doing get().

names can be:

• dict, where the keys are the names of the arguments and the values are the names for their type.
• str, all arguments will be named with the value

injector.bind(with_names(provide_B, {'k': 'name_for_k'}))

Note that v won't have a name

injector.bind(with_names(provide_B, 'name_for_all'))

Both v and k will have name name_for_all

It can also be used on classes:

injector.bind(with_names(C, 'app'))

named(...)

Similar to with_names() but it is a decorator that is used when defining a provider or class __init__.

class Z:
    @named({'a': 'conf'})
    def __init__(self, a: dict, b: str):
        ...


@named('foo')
def provide_int(x: int, b: str) -> int:
    ...

This is equivalent to using typing.Annotated as follows:

from typing import Annotated
from applipy_inject import name

class Z:
    def __init__(self, a: Annotated[dict, name('conf')], b: str):
        ...


def provide_int(x: Annotated[int, name('foo')], b: Annotated[str, name('foo')]) -> int:
    ...