A simple Inversion of Control container
If you search for Inversion of Control containers for Python you often encounter the argument “Python is dynamic and does not need those things that static languages need”. This is partly true.
Dependency Injection and Inversion of Control is a pattern and not a language feature. It not only makes your code easier to test, but also way more readable. The dependencies are clearly noted in the constructor and your IDEs will give you autocompletion support. If you need to test a class, it is clear where and how to pass in the mocks.
For further information watch Google’s Clean Code Talks
Pymple does currently not support:
This library is a Python 3.4+ library. On Python 3.4 however the typings package is required.
Install it via pip for Python 3:
sudo pip3 install pymple
Pymple knows two types of parameters:
- Singletons: A singleton is a callable that is executed once and the result is saved so future calls to the build method will return the same instance
- Factories: A factory is callable that is executed again every time it is accessed
By default Pymple tries to resolved a singleton based on the annotated type, e.g.:
from pymple import Container class A: def __init__(self): pass class B: def __init__(self, param: A): self.a = A container = Container() b = container.resolve(B) isinstance(b.a, A) == True
Overriding The Default Behavior
However you can also override it by defining it explicitly:
container = Container() ccontainer.register(B, lambda c: B('hi')) b = container.resolve(B) b.a == 'hi'
The first passed in variable to the lambda is the container instance itself, so you can also resolve other classes on it:
container = Container() ccontainer.register(B, lambda c: B(c.resolve(A))) b = container.resolve(B) isinstance(b.a, A) == True
If you want to register a factory instead of a singleton, simple pass False as the second parameter:
container = Container() ccontainer.register(B, lambda c: B('hi'), False) b = container.resolve(B) c = container.resolve(B) b != c
Sometimes a type interface uses an abstract class as type annotation. In that case you can simply define an alias:
container = Container() ccontainer.alias(ConcreteClass, AbstractClass) clazz = container.resolve(AbstractClass) isinstance(clazz, ConcreteClass) == True