Skip to main content

Easy to use mocking, stubbing and spying framework.

Project description





python, mocking, testing, unittest, unittest2


Chai provides a very easy to use api for mocking, stubbing and spying your python objects, patterned after the Mocha library for Ruby.


The following is an example of a simple test case which is mocking out a get method on the CustomObject. The Chai api allows use of call chains to make the code short, clean, and very readable. It also does away with the standard setup-and-replay work flow, giving you more flexibility in how you write your cases.

from chai import Chai

class CustomObject (object):
    def get(self, arg):

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        self.expect(obj.get).args('name').returns('My Name')
        self.assert_equals(obj.get('name'), 'My Name')
        self.expect(obj.get).args('name').returns('Your Name')
        self.assert_equals(obj.get('name'), 'Your Name')

    def test_mock_get_with_at_most(self):
        obj = CustomObject()
        self.expect(obj.get).args('name').returns('My Name').at_most(2)
        self.assert_equals(obj.get('name'), 'My Name')
        self.assert_equals(obj.get('name'), 'My Name')
        self.assert_equals(obj.get('name'), 'My Name') # this one will fail

if __name__ == '__main__':
    import unittest2


All of the features are available by extending the Chai class, itself a subclass of unittest.TestCase. If unittest2 is available Chai will use that, else it will fall back to unittest. Chai also aliases all of the assert* methods to lower-case with undersores. For example, assertNotEquals can also be referenced as assert_not_equals.

Additionally, Chai loads in all assertions, comparators and mocking methods into the module in which a Chai subclass is declared. This is done to cut down on the verbosity of typing self. everywhere that you want to run a test. The references are loaded into the subclass’ module during setUp, so you’re sure any method you call will be a reference to the class and module in which a particular test method is currently being executed. Methods and comparators you define locally in a test case will be globally available when you’re running that particular case as well.

class ProtocolInterface(object):
    def _private_call(self, arg):
    def get_result(self, arg):
        return 'ok'

class TestCase(Chai):
    def assert_complicated_state(self, obj):
        return True  # ..or.. raise AssertionError()

    def test_mock_get(self):
        obj = ProtocolInterface()
        data = object()
        assert_equals('ok', obj.get_result(data))

As of 0.3.0, the Chai API has significantly changed such that the default behavior of an expectation is least specific. This supports rapid iterative testing with minimal pain and verbosity. An example of the differences:

class CustomObject (object):
    def get(self, arg):
        return arg

class TestCase(Chai):
    def test_0_2_0(self):
        obj = CustomObject()
        assert_equals( None, obj.get(5) )
        assert_equals( None, obj.get(5) )
        assert_equals( None, obj.get(5) )
        expect(obj.get).any_args().returns( 'test' ).times(2)
        assert_equals( 'test', obj.get(5) )
        assert_equals( 'test', obj.get(5) )
        assert_raises( UnexpectedCall, obj.get )

    def test_0_3_0(self):
        obj = CustomObject()
        assert_equals( None, obj.get() )
        expect(obj.get).returns( 'test' )
        assert_equals( 'test', obj.get(5) )
        assert_equals( 'test', obj.get(5) )


The simplest mock is to stub a method. This replaces the original method with a subclass of chai.Stub, the main instrumentation class. All additional stub and expect calls will re-use this stub, and the stub is responsible for re-installing the original reference when Chai.tearDown is run.

Stubbing is used for situations when you want to assert that a method is never called, and is equivalent to expect(target).times(0):

class CustomObject (object):
    def get(self, arg):
    def prop(self):

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_raises( UnexpectedCall, obj.get )

In this example, we can reference obj.get directly because get is a bound method and provides all of the context we need to refer back to obj and stub the method accordingly. There are cases where this is insufficient, such as module imports, special Python types, and when module attributes are imported from another (like os and posix). If the object can’t be stubbed with a reference, UnsupportedStub will be raised and you can use the verbose reference instead.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        stub(obj, 'get')
        assert_raises( UnexpectedCall, obj.get )

Stubbing an unbound method will apply that stub to all future instances of that class.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_raises( UnexpectedCall, obj.get )

Unbound methods can also be stubbed by attribute name instead of by reference.

class TestCase(Chai):
    def test_mock_get(self):
        stub(CustomObject, 'get')
        obj = CustomObject()
        assert_raises( UnexpectedCall, obj.get )

Some methods cannot be stubbed because it is impossible to call setattr on the object, typically because it’s a C extension. A good example of this is the datetime.datetime class. In that situation, it is best to mock out the entire module (see below).

Finally, Chai supports stubbing of properties on classes. In all cases, the stub will be applied to a class and individually to each of the 3 property methods. Because the stub is on the class, all instances need to be addressed when you write expectations. The first interface is via the named attribute method which can be used on both classes and instances.

class TestCase(Chai):
    def test_prop_attr(self):
        obj = CustomObject()
        stub( obj, 'prop' )
        assert_raises( UnexpectedCall, lambda: obj.prop )
        stub( stub( obj, 'prop' ).setter )

Using the class, you can directly refer to all 3 methods of the property. To refer to the getter you use the property directly, and for the methods you use its associated attribute name. You can stub in any order and it will still resolve correctly.

class TestCase(Chai):
  def test_prop_attr(self):
    stub( CustomObject.prop.setter )
    stub( CustomObject.prop )
    stub( CustomObject.prop.deleter )
    assert_raises( UnexpectedCall, lambda: CustomObject().prop )

Python 3

Unbound methods can only be stubbed by attribute in Python 3 as unbound methods do not have a reference to the class they’re defined in, and appear as module functions.


PyPy does not support stubs on the setter and deleter methods of properties. Additionally, it does not support spies on methods such as ExampleClass.__hash__ because it is represented as an unbound method, rather than CPython’s method-wrapper, and unbound methods do not support spies.

Expectations and Spies

Expectations are individual test cases that can be applied to a stub. They are expected to be run in order (unless otherwise noted). They are greedy, in that so long as an expectation has not been met and the arguments match, the arguments will be processed by that expectation. This mostly applies to the “at_least” and “any_order” expectations, which (may) stay open throughout the test and will handle any matching call.

Expectations will automatically create a stub if it’s not already applied, so no separate call to stub is necessary. The arguments and edge cases regarding what can and cannot have expectations applied are identical to stubs. The expect call will return a new chai.Expectation object which can then be used to modify the expectation. Without any modifiers, an expectation will expect at least one call with any arguments and return None.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_equals(None, obj.get())
        assert_equals(None, obj.get())

As noted above, Chai will by default perform a greedy match, closing out an implied at_least_once() on every expectation when a new expectation is defined. The expectation will be immediately closed if it has already been satisfied when a new expectation is created.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_equals(1, obj.get())
        assert_equals(2, obj.get())
        assert_equals(2, obj.get())
        assert_equals(3, obj.get())

Modifiers can be applied to the expectation. Each modifier will return a reference to the expectation for easy chaining. In this example, we’re going to match a parameter and change the behavior depending on the argument. This also shows the ability to incrementally add expectations throughout the test.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_equals('hello', obj.get('foo') )
        assert_equals('hello', obj.get('foo') )
        expect(obj.get).args('bar').raises( ValueError )
        assert_raises(ValueError, obj.get, 'bar')

It is very common to need to run expectations on the constructor for an object, possibly including returning a mock object. Chai makes this very simple.

def method():
    obj = CustomObject('state')
    return obj

class TestCase(Chai):
    def test_method(self):
        obj = mock()
        expect( CustomObject ).args('state').returns( obj )
        expect( )
        assert_equals( obj, method() )

The arguments modifier supports several matching functions. For simplicity in covering the common cases, the args modifier assumes an equals test for instances and a logical or of [instanceof, equals] test for types. All rules that apply to positional arguments also apply to keyword arguments.

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject()
        assert_raises( UnexpectedCall, obj.get, 3 )
        assert_equals( 42, obj.get(3.14) )

        assert_equals( 'yes', obj.get('no') )

        assert_raises( UnexpectedCall, obj.get, [] )
        assert_equals( 'yes', obj.get(list) )

Lastly, Chai 1.0.0 supports spies. These are an extension of expectations and support most of the same features. The modifiers returns and raises raise UnsupportedModifier because the spy passes arguments and returns or raises the results of the stubbed function. You can make use of side_effect to inject code just before the spied-on function is executed, however the return value will be ignored. This behavior is especially useful when testing race conditions. Additionally, there are a few types of stubs which are not (currently) supported by spies:

  • properties

  • unbound methods

Spies can be used just like any expectation.

class TestCase(Chai):
    def test_spy(self):
        class Spy(object):
            def __init__(self, val):
                self._val = val
            def set(self, val):
                self._val = val
            def get(self):
                return self._val

        # Spy on the constructor
        obj = Spy(3)
        assert_equals(3, obj._val)

        # Spy on the set function
        assert_equals(5, obj._val)

        # Spy on the get function
        assert_equals(5, obj.get())

        # Spy on the hash function
        spy(Spy, '__hash__')
        dict()[obj] = "I spy with my little eye"


Expectations expose the following public methods for changing their behavior.

args(*args, **kwargs)

Add a test to the expectation for matching arguments.


Any arguments are accepted.


Add a return value to the expectation when it is matched and executed.


When the expectation is run it will raise this exception. Accepts type or instance.


An integer that defines a hard limit on the minimum and maximum number of times the expectation should be executed.


Sets a minimum number of times the expectation should run and removes any maximum.


Equivalent to at_least(1).


Sets a maximum number of times the expectation should run. Does not affect the minimum.


Equivalent to at_most(1).


Equivalent to times(1), also the default for any expectation.


The expectation can be called at any time, independent of when it was defined. Can be combined with at_least_once to force it to respond to all matching calls throughout the test.

side_effect(callable, *args, **kwargs)

Called with a function argument. When the expectation passes a test, the function will be executed. The side effect will be executed even if the expectation is configured to raise an exception. If the side effect is defined with arguments, then those arguments will be passed in when it’s called, otherwise the arguments passed in to the expectation will be passed in.


[Spies Only] Called with a function argument. When the expectation passes a test, the function will be executed and passed the return value from the function as an argument.


Will remove the stub after the expectation has been met. This is useful in cases where you need to mock core methods such as open, but immediately return its original behavior after the mocked call has run.

Argument Comparators

Argument comparators are defined as classes in chai.comparators, but loaded into the Chai class for convenience (and by extension, a subclass’ module). Chai handles the common case of a type object by using the is_a comparator, else defaults to the equals comparator. Users can create subclasses of Comparator and use those for custom argument processing.

Comparators can also be used inside data structures. For example:

expect( area ).args( {'pi':almost_equals(3.14), 'radius':is_a(int,long,float)} )

The default comparator, uses standard Python equals operator

almost_equals(float, places)

Identical to assertAlmostEquals, will match an argument to the comparator value to a most places digits beyond the decimal point.


Matches parameters with defined length. Must be either an integer or a set of integers that implements the in function.


Match an argument of a given type. Supports same arguments as builtin function isinstance.


Matches an argument using the Python is comparator.


Matches an argument if any of the comparators in the argument list are met. Uses automatic comparator generation for instances and types in the list.


Matches an argument if all of the comparators in the argument list are met. Uses automatic comparator generation for instances and types in the list.


Matches an argument if the supplied comparator does not match.


Matches an argument using a regular expression. Standard re rules apply.


Matches an argument if the callable returns True. The callable must take one argument, the parameter being checked.


Matches any argument.


Matches if the argument is in the in_list.


Matches if the argument contains the object using the Python in function.


Matches if the argument contains all of the same items as in container. Insists that the argument is the same type as container. Useful when you need to assert a few values in a list or dictionary, but the exact contents are not known or can vary.


A variable match against the first time that the argument is called. In the case of multiple calls, the second one must match the previous value of name. After your tests have run, you can check the value against expected arguments through var(name).value. This is really useful when you’re testing a deep stack and it’s simpler to assert that “value A was used in method call X”. Variables can also be used to capture an argument and return it.

expect( encode ).args( var('src'), 'gzip' ).returns( var('src') )

A note of caution If you are using the func comparator to produce side effects, be aware that it may be called more than once even if the expectation you’re defining only occurs once. This is due to the way Stub.__call__ processes the expectations and determines when to process arguments through an expectation.

Context Manager

An expectation can act as a context manager, which is very useful in complex mocking situations. The context will always be the return value for the expectation. For example:

def get_cursor(cname):
    return db.Connection( 'host:port' ).collection( cname ).cursor()

def test_get_cursor():
    with expect( db.Connection ).any_args().returns( mock() ) as connection:
        with expect( connection.collection ).args( 'collection' ).returns( mock() ) as collection:
            expect( collection.cursor ).returns( 'cursor' )

    assert_equals( 'cursor', get_cursor('collection') )


Sometimes you need a mock object which can be used to stub and expect anything. Chai exposes this through the mock method which can be called in one of two ways.

Without any arguments, Chai.mock() will return a chai.Mock object that can be used for any purpose. If called with arguments, it behaves like stub and expect, creating a Mock object and setting it as the attribute on another object.

Any request for an attribute from a Mock will return a new Mock object, but setattr behaves as expected so it can store state as well. The dynamic function will act like a stub, raising UnexpectedCall if no expectation is defined.

class CustomObject(object):
    def __init__(self, handle):
        _handle = handle
    def do(self, arg):

class TestCase(Chai):
    def test_mock_get(self):
        obj = CustomObject( mock() )
        expect( ).args('it').returns('ok')
        assert_raises( UnexpectedCall, obj._handle.do_it_again )

The stub and expect methods handle Mock objects as arguments by mocking the __call__ method, which can also act in place of __init__.

# module
from collections import deque

class CustomObject(object):
    def __init__(self):
        self._stack = deque()

# module
import custom
from custom import CustomObject

class TestCase(Chai):
    def test_mock_get(self):
        mock( custom, 'deque' )
        expect( custom.deque ).returns( 'stack' )

        obj = CustomObject()
        assert_equals('stack', obj._stack)

Here we can see how to mock an entire module, in this case replacing the deque import in with a Mock.

Mock objects, because of the getattr implementation, can also support nested attributes.

class TestCase(Chai):
    def test_mock(self):
      m = mock() = 42
      expect( ).returns( 'hello' )
      assert_equals( 'hello', )
      assert_equals( 42, )

In addition to implementing __call__, Mock objects implement __nonzero__, the container and context manager interfaces are defined. Nonzero will always return True; other methods will raise UnexpectedCall. The __getattr__ method cannot be itself stubbed.


You can install Chai either via the Python Package Index (PyPI) or from source.

To install using pip,:

$ pip install chai

Downloading and installing from source

Download the latest version of Chai from

You can install it by doing the following,:

$ tar xvfz chai-*.*.*.tar.gz
$ cd chai-*.*.*.tar.gz
$ python install # as root

Using the development version

You can clone the repository by doing the following:

$ git clone git://


Use nose to run the test suite.

$ nosetests

Bug tracker

If you have any suggestions, bug reports or annoyances please report them to our issue tracker at


This software is licensed under the New BSD License. See the LICENSE.txt file in the top distribution directory for the full license text.


Special thank you to the following people for contributions to Chai

Project details

Download files

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

Source Distribution

chai-1.1.2.tar.gz (107.8 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page