This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

A fixture lifecycle management library for your tests

Project Description

Why do we need this library

We are not satisfied with classical xUnit way of setup and teardown. We prefer concise approach of py.test over the verbosity of standard unittest.

We found ourselves copying and pasting the same boilerplate code from one test to another or creating extensive structure of test class hierarchy.

py.test fixtures, injected in test functions as parameter names, is different approach for fixture management. It’s neither worse nor better, but we found it to be not as flexible as we need.

Some questions, that we wanted to solve often, looked like:

  • I have a py.test fixture which creates a new user with default set of properties. Is there a way I can create a user with different name by the same fixture?
  • Is there a way to create two users in one test case with the same fixture?
  • Is there an easy recipe to create a user first, and then, say, a todo item for this particular user in another, separate, fixture?

Sure enough, we can handle or work around all these issues somehow with xUnit setups and teardowns or py.test fixtures, but we wanted something more flexible, easy and convenient to use. That’s why we created resources library.

How do we use it

First, we define functions which we call “resource makers”. These makers are responsible for creating and destroying resources. It’s like setup and teardown in one callable.

from resources import resources

def user(email='', password='password', username='Joe'):
    user = User.objects.create(email=email, password=password, username=username)
        yield user

The flow is simple: we create, we yield, we destroy.

We get a number of resource makers, and we group them into modules, like tests/, tests/, etc.

Then, in a test file, where we plan to use resources, we import the same global object, load resource modules we need, and activate them in tests.

from resources import resources

def test_user_properties():
    with resources.user_ctx() as user:
        assert user.username == 'Joe'

This is where a little bit of magic happens. Once you define and register the resource maker with name foo, a context manager foo_ctx is created for your convenience. This context manager creates a new resource instance with the corresponding maker function, and destroys the object the way you defined, once the code flow abandons a wrapping “with”-context.

When it shines

At this point and maybe not so exciting. Yeah, everyone can write the code like this, the difference is that we actually did it :-). We also have a bunch of nifty features making the whole stuff more interesting.

Feature 1. Customizeable resources

Contexts are better than py.test fixtures, because they are customizeable. Provide everything you need to context manager, and it will be passed to resource maker function as an arguments.

def test_user_properties():
    with resources.user_ctx(name='Mary') as user:
        assert user.username == 'Mary'

Feature 2. Global object scope and dependent resources

We need to have access to resources at different stages of our tests: to get access to object’s properties and methods, to initiate another, dependent fixture instance, and finally to tear down everything.

As soon as you enter the context with resources.foo_ctx() a variable will be created and will be available from everywhere, including your test function, and other resource makers.

The latter fact is especially important, because it’s the way we manage dependent resources. Yet we need some conventions, which resource is created first, and so on.

def todo_item(content='Foo'):
    item = TodoItem.objects.create(user=resources.user, content=content)

We agreed that we create user resource first, and todo item afterwards, and created a new resource maker, taking advantage of this convention.

We use it like this:

def test_todo_item_properties():
    with resources.user_ctx(), resources.todo_item_ctx():
        assert resources.todo_item.content == 'Foo'

By the way, if you are still stuck with python2.6, several context managers in the same “with” expression aren’t available for you yet. Use contextlib.nested to avoid deep indentation.

Feature 3. Several resources of the same class, and tuneable resource names

Sometimes we need to create a couple of resources of the same type, instead of just one instance. It’s not a problem, if you don’t want to use global namespace to get access to them. Otherwise you must create a unique identifier for every resource.

Actually, it’s trivial. All you should do is provide a special _name attribute to context manager constructor. This attribute won’t be passed to your resource maker function.

def test_a_couple_of_users():
    with resources.user_ctx(username='Adam', _name='adam'), \
         resurces.user_ctx(username='Eve', _name='eve'):
        assert resources.adam.username == 'Adam'
        assert resources.eve.username == 'Eve'

Feature 4. Function decorators

Context manager can work as a decorator too. When we use it like this, an extra argument will be passed to the function.

def test_user_properties(user):
    assert user.username == 'Joe'

We should say that usually it works, but to make it work along with py.test which performs deep introspection of function signatures, we made in with some “dirty hacks” inside, and you may find out that in some cases the chain of decorators dies with a misleading exception. We’d recommend to use context managers instead of decorators, wherever possible.

Feature 5. Resource managers

Yes, we do use setup and teardown methods too. If every function in your test suite uses the same set of resources, it would be counterproductive to write the same chain of decorators or context managers over and over again.

In this case we use another concept: resource managers. Every resource maker foo creates the resources.foo_mgr instance, having start and stop methods. The start method accepts all arguments which the foo_ctx function does, including special _name argument. The stop method has only one optional _name argument, and is used to destroy previously created instance.

Here is a py.test example

def setup_function(func):

def test_user_properties():
    assert resources.user.username == 'Mary'

def teardown_function(func):

Feature 6. Built-in console and debugger

Sometimes it’s nice to take a look on what’s going on within test function and get access at some point to python console or debugger.

Usually you probably do something like

import pdb; pdb.set_trace()

Or, if you need to get shell and have IPython installed

from IPython import embed; embed()

As it happens often, we added to resources two functions, launching either debugger or python console inside your test function.

from resources import resources

def test_something():
    resources.pdb()    # to launch debugger  # to launch Python REPL

If you install IPython and ipdb (pip install IPython ipdb), you get more friendly versions of consoles, otherwise resources fall back to built-in python console and debugger.

Launch py.test with -s switch to be able to fall into interactive console.

It’s especially cool that resources object is autocomplete-friendly and it works well in IPython

In [1]: resources.
resources.john            resources.pdb             resources.register_mod
resources.mary            resources.register_func

In [1]: resources.mary
Out[1]: {'name': 'Mary Moe'}

In [2]: resources.user_mgr.start()
Out[2]: {'name': 'John Doe'}

In [3]: resources.todo
resources.todo_item_ctx  resources.todo_item_mgr

In [3]: resources.todo_item_mgr.start()
Out[3]: {'text': 'Do something', 'user': {'name': 'John Doe'}}

In [4]: resources.todo
resources.todo_item      resources.todo_item_ctx  resources.todo_item_mgr

In [4]: resources.todo_item
Out[4]: {'text': 'Do something', 'user': {'name': 'John Doe'}}

Feature 7. Globally accessible storage of constants

This feature is not something unique to resources module. Pretty much every object can act this way, but it is handy to have a convention about the way you store your test-related constants.

It may work like this.

resources.TEST_DIRECTORY = '/tmp/foo'
resources.DOMAIN_NAME = ''
resources.SECRET_KEY = 'foobar'

And then, in the test file.

from resources import resources
resoures.register_mod('<a resource module name here>')

def test_constants():
    assert resources.TEST_DIRECTORY == '/tmp/foo'
    assert resources.DOMAIN_NAME == ''
    assert resources.SECRET_KEY == 'foobar'


The resources library works for us in py.test environment. We don’t see any reasons why it shouldn’t work the same way with nose or classic unitttests. It works for python versions 2.6, 2.7 and 3.3.

Please bear in mind that the library is not thread safe, as we are happy with single threaded tests at this time.

And after all… Seven extra features to improve your test suites for free! What are you waiting for? It’s already improved the quailty of our lives in Doist Inc, and we do hope it will do the same for your projects.

Release History

Release History

This version
History Node


History Node


Download Files

Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
python-resources-0.3.tar.gz (8.3 kB) Copy SHA256 Checksum SHA256 Source Mar 22, 2013

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting