antidote 0.9.0

Dependency injection.

Antidotes is a dependency injection micro-framework for Python 3.6+. It is designed on two core ideas:

• Keep dependency declaration close to the actual code as it’s deeply related. Dependency injection is about removing the responsibility of building dependencies from their clients. Not separating how a dependency is built from its implementation.

• It should help creating maintainable code in a straightforward way and offer effortless integration.

It provides the following features:

• Ease of use

  • injection anywhere you need through a decorator @inject, be it static methods, functions, etc.. By default it will only rely on type hints, but it supports a lot more !

  • no **kwargs arguments hiding actual arguments and fully mypy typed, helping you and your IDE.

  • documented, see https://antidote.readthedocs.io/en/stable. If you don’t find what you need, open an issue ;)

  • thread-safe, cycle detection

  • magic is frowned upon and avoided as much as possible. But it is used when it doesn’t hurt understandability and improves readability.

• Flexibility

  • A rich ecosystem of dependencies out of the box: services, configuration, factories, interface/implementation, tags.

  • All of those are implemented on top of the core implementation. If Antidote doesn’t provide what you need, there’s a good chance you can implement it yourself quickly.

• Maintainability

  • The different kind of dependencies are designed to be easy to track back. Finding where a dependency is defined is easy.

  • Overriding dependencies will raise an exception (no duplicates) and one can freeze the dependency space. Once frozen, no new dependencies can be defined.

• Testability

  • @inject lets you override any injections by passing explicitly the arguments.

  • Override dependencies locally within a context manager.

• Performance

  • Antidote has two implementations: the pure Python one which is the reference and the Cython one which is heavily tuned for fast injection. Injection is roughly 10x times faster than with the pure Python. It allows using injections without impact on most functions. See injection benchmark

Installation

To install Antidote, simply run this command:

pip install antidote

Quick Start

How does injection looks like ? Here is a simple example:

from antidote import inject, Service, Constants, const

class Conf(Constants):
    # A constant is defined by const. When needed, the associated value will
    # be passed on to get() to retrieve, only once, lazily the real value.
    DB_HOST = const[str]('host')
    # The type is not required, but it provides Mypy the necessary information
    # on the real dependency value, once retrieved through get().
    DB_HOST_UNTYPED = const('host')

    def __init__(self):
        self._data = {'host': 'localhost:6789'}

    def get(self, key):
        return self._data[key]

# Declare Database as a dependency that can be injected
class Database(Service):
    # Antidotes configuration for this Service. By default `__init__()` is wired (injected).
    #
    __antidote__ = Service.Conf().with_wiring(
        # Specifying that arguments named 'host' should be injected with the
        # dependency Conf.DB_HOST.
        dependencies=dict(host=Conf.DB_HOST))

    # You could have wired yourself `__init__()` by applying:
    # @inject(dependencies=dict(host=Conf.DB_HOST))
    def __init__(self, host: str):
        self._host = host

# Inject dependencies in f(), by default only type annotations are used. But
# arguments name, explicit mapping, etc.. can also be used.
@inject
def f(db: Database = None):
    # Defaulting to None allows for MyPy compatibility but isn't required to work.
    assert db is not None
    pass

f()  # Service will be automatically injected if not provided
f(Database('localhost:6789'))  # but you can still use the function normally

Want more ? Here is an over-engineered example to showcase a lot more features:

"""
Simple example where a MovieDB interface is defined which can be used
to retrieve the best movies. In our case the implementation uses IMDB
to dot it.
"""
from antidote import Constants, factory, Implementation, inject, world, const

class MovieDB:
    """ Interface """

    def get_best_movies(self):
        pass

class ImdbAPI:
    """ Class from an external library. """

    def __init__(self, *args, **kwargs):
        pass

# Defining a singleton. Can only be overridden in tests.
world.singletons.add('conf_path', '/some/file')

class Conf(Constants):
    IMDB_HOST = const[str]('imdb.host')
    # Constants will by default automatically enforce the cast to int,
    # float and str. Can be removed or extended to support Enums.
    IMDB_PORT = const[int]('imdb.port')
    IMDB_API_KEY = const[str]('imdb.api_key')

    # `use_names=True` specifies that Antidote can use the argument names
    # when type hints are not present or too generic (builtins typically).
    __antidote__ = Constants.Conf().with_wiring(use_names=True)

    def __init__(self, conf_path: str):
        """ Load configuration from `conf_path` """
        self._raw_conf = {
            'imdb': {
                'host': 'dummy_host',
                'api_key': 'dummy_api_key',
                'port': '80'
            }
        }

    def get(self, key):
        """
        self.get('a.b') <=> self._raw_conf['a']['b']
        """
        from functools import reduce
        return reduce(dict.get, key.split('.'), self._raw_conf)  # type: ignore

# ImdbAPI will be provided by this factory, as defined by the return type annotation.
# The dependencies arguments specifies what must be injected
@factory(dependencies=(Conf.IMDB_HOST, Conf.IMDB_PORT, Conf.IMDB_API_KEY))
def imdb_factory(host: str, port: int, api_key: str) -> ImdbAPI:
    # Here host = Conf().get('imdb.host')
    return ImdbAPI(host=host, port=port, api_key=api_key)

# Implementation tells Antidote that this class should be used as an implementation of
# the interface MovieDB
class IMDBMovieDB(MovieDB, Implementation):
    # As ImdbAPI is provided by imdb_factory, Antidote requires it to be explicitly
    # specified. This ensures that can always track back where dependencies are
    # coming from.
    __antidote__ = Implementation.Conf().with_wiring(
        dependencies=dict(imdb_api=ImdbAPI @ imdb_factory))

    def __init__(self, imdb_api: ImdbAPI):
        self._imdb_api = imdb_api

    def get_best_movies(self):
        pass

@inject
def f(movie_db: MovieDB = None):
    assert movie_db is not None
    pass

# You can also retrieve dependencies by hand
world.get(Conf.IMDB_HOST)
# To help Mypy to for type checking you can provide a type hint. Contrary to const,
# it won't enforce anything, like typing.cast().
world.get[str](Conf.IMDB_HOST)
# To avoid repetition, if the dependency is the type itself, you may omit it:
world.get[IMDBMovieDB]()

# If you need to handle multiple different api_keys for some reason you can
# specify them in the dependency itself:
world.get[ImdbAPI](ImdbAPI @ imdb_factory.with_kwargs(api_key='XX'))
# As imdb_factory returns a singleton, by default, this will also be the case
# here. Using the same API key, will return the same instance. This avoids boilerplate
# code when the same instance is needed with different arguments. The same works
# with a Service. In the previous example you could have
# used `Database.with_kwargs(host='something')`

# Like before you can call f() without any arguments:
f()

That looks all good, but what about testability ?

# You can still explicitly pass the arguments to override
# injection.
conf = Conf('/path')
f(IMDBMovieDB(imdb_factory(
    # The class attributes will retrieve the actual value when called on a instance.
    # Hence this is equivalent to conf.get('imdb.host'), making your tests easier.
    host=conf.IMDB_HOST,
    port=conf.IMDB_PORT,
    api_key=conf.IMDB_API_KEY,  # <=> conf.get('imdb.api_key')
)))

# When testing you can also override locally some dependencies:
with world.test.clone(overridable=True, keep_singletons=True):
    world.test.override.singleton({
        Conf.IMDB_HOST: 'other host'
    })
    f()

If you ever need to debug your dependency injections, Antidote also provides a tool to have a quick summary of what is actually going on. This would be especially helpful if you encounter cyclic dependencies for example.

world.debug(f)
# will output:
"""
f
└── Static link: MovieDB -> IMDBMovieDB
    └── IMDBMovieDB
        └── ImdbAPI @ imdb_factory
            └── imdb_factory
                ├── Const: Conf.IMDB_API_KEY
                │   └── Lazy: Conf()  #0BjHAQ
                │       └── Singleton 'conf_path' -> '/...'
                ├── Const: Conf.IMDB_HOST
                │   └── Lazy: Conf()  #0BjHAQ
                │       └── Singleton 'conf_path' -> '/...'
                └── Const: Conf.IMDB_PORT
                    └── Lazy: Conf()  #0BjHAQ
                        └── Singleton 'conf_path' -> '/...'
"""

# For example suppose we don't have the singleton `'conf_path'`
with world.test.clone(keep_singletons=False):
    world.debug(f)
    # As you can see, 'conf_path` is not found. Hence when Conf will be instantiated
    # it will fail.
    """
    f
    └── Static link: MovieDB -> IMDBMovieDB
        └── IMDBMovieDB
            └── ImdbAPI @ imdb_factory
                └── imdb_factory
                    ├── Const: Conf.IMDB_API_KEY
                    │   └── Lazy: Conf()  #0BjHAQ
                    │       └── /!\\ Unknown: 'conf_path'
                    ├── Const: Conf.IMDB_HOST
                    │   └── Lazy: Conf()  #0BjHAQ
                    │       └── /!\\ Unknown: 'conf_path'
                    └── Const: Conf.IMDB_PORT
                        └── Lazy: Conf()  #0BjHAQ
                            └── /!\\ Unknown: 'conf_path'
    """

Hooked ? Check out the documentation ! There are still features not presented here !

Cython

The cython implementation is roughly 10x faster than the Python one and strictly follows the same API than the pure Python implementation. This implies that you cannot depend on it in your own Cython code if any. It may be moved to another language.

If you encounter any inconsistencies, please open an issue ! You can avoid the pre-compiled wheels from PyPI with the following:

pip install --no-binary antidote

Mypy

Antidote passes the strict Mypy check and exposes its type information (PEP 561). Unfortunately static typing for decorators is limited to simple cases, hence Antidote @inject will just return the same signature from Mypys point of view. The best way, currently that I know of, is to define arguments as optional as shown below:

from antidote import inject, Service

class MyService(Service):
    pass

@inject
def f(my_service: MyService = None) -> MyService:
    # We never expect it to be None, but it Mypy will now
    # understand that my_service may not be provided.
    assert my_service is not None
    return my_service


s: MyService = f()

# You can also overload the function, if you want a more accurate type definition:
from typing import overload

@overload
def g(my_service: MyService) -> MyService: ...

@overload
def g() -> MyService: ...

@inject
def g(my_service: MyService = None) -> MyService:
    assert my_service is not None
    return my_service


s2: MyService = g()

Note that any of this is only necessary if you’re calling _explicitly_ the function, if only instantiate MyService through Antidote for example, you won’t need this for its __init__() function typically. You could also use a Protocol to define a different signature, but it’s more complex.

Issues / Feature Requests / Questions

Feel free to open an issue on Github for questions, requests or issues ! ;)

How to Contribute

1. Check for open issues or open a fresh issue to start a discussion around a feature or a bug.

2. Fork the repo on GitHub. Run the tests to confirm they all pass on your machine. If you cannot find why it fails, open an issue.

3. Start making your changes to the master branch.

4. Writes tests which shows that your code is working as intended. (This also means 100% coverage.)

5. Send a pull request.

Be sure to merge the latest from “upstream” before making a pull request!

If you have any issue during development or just want some feedback, don’t hesitate to open a pull request and ask for help !

Pull requests will not be accepted if:

• classes and non trivial functions have not docstrings documenting their behavior.

• tests do not cover all of code changes (100% coverage).