siringa (meaning syringe in Italian) is a minimalist, idiomatic dependency injection and inversion of control library for Python, implemented in Hy, a homoiconic Lisp dialect for Python.

To get started, take a look to the documentation, API, tutorial and examples.


  • Simple, idiomatic and versatile programmatic API.
  • Annotation based dependency injection that is PEP 3017 and PEP 0484 friendly.
  • First-class decorator driven dependency injection and registering.
  • Ability to create multiple dependency containers.
  • Hierarchical dependency containers based on inheritance.
  • Dependency inference based on pattern-matching techniques.
  • First-class support for dependency mocking for better testing.
  • Detects cyclic dependencies (work in progress).
  • Small and (almost) dependency-free library.
  • Works with CPython 3+.

Design philosophy

  • Code instrumentation should be non-intrusive and idiomatic.
  • Explicitness over implicitness: dependencies and injections much be explicitly defined.
  • Python idiomatic: embrace decorators and type annotations.
  • Minimalism: less enables more.
  • Uniformity: there is only one way to declare and consume dependencies.
  • Predictability: developer intentions must persist based on explicitly defined intention.
  • Domain agnostic: do not enforce any domain-specific pattern.


Using pip package manager:

pip install --upgrade siringa

Or install the latest sources from Github:

pip install -e git+git://github.com/h2non/siringa.git#egg=siringa


Importing siringa

import siringa

Instrumenting dependencies

siringa embraces type hints/arguments annotation Python syntax for dependency inference and pattern matching.

def task(x, y, logger: '!Logger'):
    logger.info('task called with arguments: {}, {}'.format(x, y))
    return x * y

You can optionally annotate dependencies via siringa type annotations:

from siringa import A

def task(x, y, logger: A('Logger')):
    logger.info('task called with arguments: {}, {}'.format(x, y))
    return x * y

Finally, for a DRYer approach you can simply annotate dependencies with ! annotation flag.

In this case, the argument name expression will be used for dependency inference.

from siringa import A

def task(x, y, Logger: '!'):
    Logger.info('task called with arguments: {}, {}'.format(x, y))
    return x * y

Registering dependencies

siringa allows you to rely on decorators for idiomatic dependencies registering.

Dependency name is dynamically inferred at registration time based on class or function name.

class Logger(object):
    logger = logging.getLogger('siringa')

    def info(msg, *args, **kw):
        logger.info(msg, *args, **kw)

However, you can define a custom dependency name by simply passing a string as first argument:

class Logger(object):

Finally, you can register dependencies with a traditional function call, such as:

class Logger(object):

siringa.register('MyCustomLogger', Logger)

class compute(x, y):
    return x * y

siringa.register('multiply', compute)


siringa wraps callable object in the transparent and frictionless way abstracting things for developers.

You can invoke or instantiate any dependency injection instrumented object as you do traditionally in raw Python code and siringa will do the rest for you inferring and pattern-matching required dependencies accordingly for you.

Below is an example of how simple it is:

# Call our previously declared function in this tutorial.
# Here, siringa will transparently inject required dependencies accordingly,
# respecting the invokation arguments and order.
task(2, 2) # => 4

Let’s demostrate this with a featured example:

import siringa

def mul(x, y):
    return x * y

def mul2(x, mul: '!mul'):
    return mul(x, 2)

def pow2(x):
    return x ** 2

def compute(x, pow: '!pow2', mul: '!mul2'):
    return pow(mul(x))

compute(2) # => 16

You can also use the invocation API in case that the target object was not properly instrumented as dependency:

def mul2(x):
    return x * 2

# Note that the function was not instrumented yet!
def compute(x, mul: '!mul2'):
    return mul(x)

siringa.invoke(compute, 2)

Create a new dependency container

siringa provides a built-in global dependency container for usability purposes, but you can create as much containers as you want.

In the siringa idioms, this means creating a new dependency layer which provides its own container and dependency injection API, pretty much as the global package API.

You can create a new dependencies layer such as:

layer = siringa.Layer('app')

# Then you can use the standard API
layer.register('print', print)

# Then you can use the standard API
def mul2(x, print: '!'):
    print('Argument:', x)
    return x * 2


A dependency layer can inherit from a parent dependency layer.

This is particularly useful in order to create a hierarchy of dependency layers where you can consume and inject dependencies from a parent container.

parent = siringa.Layer('parent')
child = siringa.Layer('child', parent)

# Register a sample dependency within parent
def mul2(x):
    return x * 2

# Verify that the dependency is injectable from child layer
parent.is_injectable('mul2') # True
child.is_injectable('mul2') # True

def compute(x, mul: '!mul2'):
    return mul(x)

compute(2) # => 2

Mocking dependencies

siringa allows you to define mocks for dependencies, which is particularly useful during testing:

class DB(object):
    def query(self, sql):
        return ['john', 'mike']

class DBMock(object):
    def query(self, sql):
        return ['foo', 'bar']

def run(sql, db: '!DB'):
    return db().query(sql)

# Test mock call
assert run('SELECT name FROM foo') == ['foo', 'bar']

# Once done, clear all the mocks

# Or alternatively clear all the registed mocks within the container

# Test read call
assert run('SELECT name FROM foo') == ['john', 'mike']

Indices and tables