Dependency injection library for python.

Dependency injection enables us to write better code, regardless of doing so manually or with the help of some library. It helps specially to keep our code inside the Law of Demeter, by programming against a known interface without having to worry about how a concrete implementation is composed.

Python is a highly dynamic language with an "open class" implementation for user types, thus the need for a full blown dependency injection framework is not specially needed. For medium to large applications though there is still the issue of how to actually implement dependency injection in the code using only Python's standard syntax/library.

This library is designed to be very lightweight and flexible as to allow its use in a variety of scenarios, including their use to aid with unit testing. It doesn't form a framework but just a set of utilities to keep the dependency injection needs in a project under control by applying it only where it makes sense, with minimum overhead and a lean learning curve.

from http.client import HTTPSConnection
from di import injector

# Create the decorator setting up the dependencies (at configuration time)
inject = injector({
  HTTPSConnection: HTTPSConnection('localhost', '8080')  
})

# Apply the decorator to our app logic to inject what we have configured (at runtime)
@inject
def fetch_it(id, http=HTTPSConnection):
  http.request("GET","/?id={0}".format(id))
  return http.getresponse()

# Call the logic without worrying about dependencies :)
print fetch_it(100).status

# Override the dependency if we have some specific use case
print fetch_it(100, http=HTTPSConnect('google.com', '80')).status

import hashlib
from di import injector, Key, DependencyMap

# Setup the dependency map
dm = DependencyMap()

# Define a custom Key to map a dependency when it's not a class
HashDep = Key('hash')

# Build a dependency programatically but only the first time it's used
@dm.singleton(HashDep)
def hash(deps):
  return lambda x: hashlib.md5(x).hexdigest()

# Create the decorator and bind it to the dependency map
inject = injector(dm)

# Define our logic defining what should be injected by default
@inject
def hasher(subject, hash=HashDep)
  return hash(subject)

print hasher('foobarbaz')

• DI at method level
• DI using descriptor protocol
• DI using a metaclass

Those are the python interpreters being validated via travis.ci upon every change in the repository.

• python 2.6
• python 2.7
• python 3.3
• python 3.4
• pypy

pip install di-py

pip install git+ssh://[email protected]/telefonicaid/di-py.git@master

Generate RPM from source code

git clone [email protected]/telefonicaid/di-py.git
cd di-py
python setup.py bdist_rpm

See LICENSE

Use the GitHub's pull request and issue tracker to provide patches or report problems with the library. All new functionality must be covered by unit tests before it can be included in the repository.

The master branch always has the cutting edge version of the code, if you are using it in your project it would be wise to create a fork of the repository or target a specific tag/commit for your dependencies.

• Iván -DrSlump- Montes
• Juan de Bravo
• Sergi Sorribas
• Tomas Montserrat