FlexMock is a simple, but flexible, mock object library for Ruby unit testing.

You can install FlexMock with the following command.

 $ gem install flexmock

Only significant changes (new APIs, deprecated APIs or backward-compatible changes) are documented here, a.k.a. minor or major version bumps. If you want a detailed changelog, go over the commit log in github (it's pretty low-traffic)

3.0.0:

• Added keyword argument and explicit proc support. Keyword argument support is Ruby 3.0+ only, getting the support to work on Ruby 2.7 deemed to be too complicated. See the release notes for 2.4 below for information about how migration can be done smoothly

• with now expects all expected keyword arguments to be explicitly given in a natural way, for instance:

  mock.should_receive(:m).with("some", "args", with: 42)
  

  The values given to the arguments can be any matcher valid for the positional arguments

• note that not giving any keyword arguments to with is interpreted as a negative (no keyword arguments are expected), and will fail if some arguments are given. Call with_any_kw_args after the with if you do not desire validation of keyword arguments:

  mock.should_receive(:m).with("some", "args").with_any_kw_args
  

• for more complex matches, pass a match object to the with_kw_args method. For instance, to match only some keyword arguments, do

  mock.should_receive(:m).with("some", "args").with_kw_args(hsh(with: 42))
  

• this release also makes matching procs explicit. Instead of passing Proc at the end of with as in 2.x, call with_block or with_no_block. If neither are called, flexmock won't validate either way (ignore whether or not a block was given). The default is to not match blocks, that is working

• The default is to assume that blocks are optional (i.e. flexmock will match either way). Nonetheless, the method with_optional_block is implemented to help migration from flexmock 2.4.0 (but is a no-op).

2.4.0:

• forward-compatible implementation of with_kw_args, with_any_kw_args, with_block and with_no_block. The objective of this release is to ensure that any test changes needed to handle Ruby 3 (along with flexmock 3) can run on ruby 2.7 and flexmock 2.4
• the default behavior of flexmock 2 regarding proc matching, that is that one needs to match them explicitly, is unchanged. Use with_optional_block instead of passing optional_proc to with, to match optionally (IMPORTANT the explicit with methods that match blocks are called block, not proc)

2.3.0:

• implemented validation of call arity for partial mocks. By setting FlexMock.partials_verify_signatures = true flexmock will verify on partials that the number of arguments, and the keyword arguments passed to the mocked call match the existing method's signature

2.2.0:

• #new_instances now mocks the #initialize method instead of mocking after the allocation was done. This allows to do mock methods called by #initialize itself. Behaviour when the allocator is explicitely provided is left unchanged, which means that the old behaviour is still available by passing :new to new_instances.

2.1.0:

• added #and_iterates to fix some shortcomings of #and_yield without breaking backward compatibility
• strict partial mocks (and "based mocks" if FlexMock.partials_are_based is set to true) are now based on the object's singleton class, instead of its class.

We have a data acquisition class (TemperatureSampler) that reads a temperature sensor and returns an average of 3 readings. We don't have a real temperature to use for testing, so we mock one up with a mock object that responds to the read_temperature message.

Here's the complete example:

  require 'test/unit'
  require 'flexmock/test_unit'

  class TemperatureSampler
    def initialize(sensor)
      @sensor = sensor
    end

    def average_temp
      total = (0...3).collect {
        @sensor.read_temperature
      }.inject { |i, s| i + s }
      total / 3.0
    end
  end

  class TestTemperatureSampler < Test::Unit::TestCase
    def test_sensor_can_average_three_temperature_readings
      sensor = flexmock("temp")
      sensor.should_receive(:read_temperature).times(3).
        and_return(10, 12, 14)

      sampler = TemperatureSampler.new(sensor)
      assert_equal 12, sampler.average_temp
    end
  end

You can find an extended example of FlexMock in Google Example.

FlexMock integrates nicely with Minitest. Just require the 'flexmock/minitest' file at the top of your test file. The flexmock method will be available for mock creation, and any created mocks will be automatically validated and closed at the end of the individual test. It works with both tests unit-style (subclasses of Minitest::Test) and spec-style.

Your test case will look something like this:

  require 'flexmock/minitest'

  class TestDog < Minitest::Test
    def test_dog_wags
      tail_mock = flexmock(:wag => :happy)
      assert_equal :happy, tail_mock.wag
    end
  end

NOTE: If you don't want to automatically extend every Minitest::Test with the flexmock methods and overhead, then require the 'flexmock' file and explicitly include the FlexMock::Minitest module in each test case class where you wish to use mock objects.

FlexMock integrates nicely with Test::Unit. Just require the 'flexmock/test_unit' file at the top of your test file. The flexmock method will be available for mock creation, and any created mocks will be automatically validated and closed at the end of the individual test.

Your test case will look something like this:

  require 'flexmock/test_unit'

  class TestDog < Test::Unit::TestCase
    def test_dog_wags
      tail_mock = flexmock(:wag => :happy)
      assert_equal :happy, tail_mock.wag
    end
  end

NOTE: If you don't want to automatically extend every TestCase with the flexmock methods and overhead, then require the 'flexmock' file and explicitly include the FlexMock::TestCase module in each test case class where you wish to use mock objects. FlexMock versions prior to 0.6.0 required the explicit include.

FlexMock also supports integration with the RSpec behavior specification framework. Starting with version 2.11.0 of RSpec, you will be able to say:

  RSpec.configure do |config|
    config.mock_with :flexmock
  end

  describe "Using FlexMock with RSpec" do
    it "should be able to create a mock" do
      m = flexmock(:foo => :bar)
      m.foo.should === :bar
    end
  end

NOTE: I often can't remember the proper RSpec configuration for flexmock without looking it up. If you are the same, you can put require 'flexmock/rspec/configure' in your spec helper to auto-configure RSpec to use flexmock.

NOTE: Older versions of RSpec used the Spec::Runner for configuration. If you are running with a very old RSpec, you may need the following:

  # Configuration for RSpec prior to RSpec 2.x
  Spec::Runner.configure do |config|
    config.mock_with :flexmock
  end

The flexmock method is used to create mocks in various configurations. Here's a quick rundown of the most common options. See FlexMock::MockContainer#flexmock for more details.

• mock = flexmock("joe")

  Create a mock object named "joe" (the name is used in reporting errors).

• mock = flexmock(:foo => :bar, :baz => :froz)

  Create a mock object and define two mocked methods (:foo and :baz) that return the values :bar and :froz respectively. This is useful when creating mock objects with just a few methods and simple return values.

• mock = flexmock("joe", :foo => :bar, :bar => :froz)

  You can combine the mock name and an expectation hash in the same call to flexmock.

• mock = flexmock("joe", :on, User)

  This defines a strict mock that is based on the User class. Strict mocks prevent you from mocking or stubbing methods that are not instance methods of the restricting class (i.e. User in our example). This helps prevent tests from becoming stale with incorrectly mocked objects when the method names change.

  Use the explicitly modifier to should_receive to override the strict mock restrictions.

• partial_mock = flexmock(real_object)

  If you you give flexmock a real object in the argument list, it will treat that real object as a base for a partial mock object. The return value partial_mock may be used to set expectations. The real_object should be used in the reference portion of the test.

• partial_mock = flexmock(real_object, :on, class_object)

• partial_mock = flexmock(real_object, :strict)

  Partial mocks can also take a restricting base, so that you cannot mock methods not in the class (without the explicitly modifier). Since partials already have a class, you can use the :strict keyword to mean the same thing as :on, real_object.class.

• partial_mock = flexmock(real_object, "name", :foo => :baz)

  Names and expectation hashes may be used with partial mocks as well.

• partial_mock = flexmock(:base, real_string_object)

  Since Strings (and Symbols for that matter) are used for mock names, FlexMock will not recognize them as the base for a partial mock. To force a string to be used as a partial mock base, proceed the string object in the calling sequence with :base.

• partial_mock = flexmock(:safe, real_object) { |mock| mock.should_receive(...) }

  When mocking real objects (i.e. "partial mocks"), FlexMock will add a handful of mock related methods to the actual object (see below for list of method names). If one or more of these added methods collide with an existing method on the partial mock, then there are problems.

  FlexMock offers a "safe" mode for partial mocks that does not add these methods. Indicate safe mode by passing the symbol :safe as the first argument of flexmock. A block is required when using safe mode (the partial_mock returned in safe mode does not have a should_receive method).

  The methods added to partial mocks in non-safe mode are:

  • should_receive
  • new_instances
  • flexmock_get
  • flexmock_teardown
  • flexmock_verify
  • flexmock_received?
  • flexmock_calls

• mock = flexmock(...) { |mock| mock.should_receive(...) }

  If a block is given to any of the flexmock forms, the mock object will be passed to the block as an argument. Code in the block can set the desired expectations for the mock object.

• mock_model = flexmock(:model, YourModel, ...) { |mock| mock.should_receive(...) }

  When given :model, flexmock() will return a pure mock (not a partial mock) that will have some ActiveRecord specific methods defined. YourModel should be the class of an ActiveRecord model. These predefined methods make it a bit easier to mock out ActiveRecord model objects in a Rails application. Other that the predefined mocked methods, the mock returned is a standard FlexMock mock object.

  The predefined mocked methods are:

  • id -- returns a unique ID for each mocked model.
  • to_params -- returns a stringified version of the id.
  • new_record? -- returns false.
  • errors -- returns an empty (mocked) errors object.
  • is_a?(other) -- returns true if other == YourModel.
  • instance_of?(class) -- returns true if class == YourModel
  • kind_of?(class) -- returns true if class is YourModel or one of its ancestors
  • class -- returns YourModel.

• mock = flexmock(... :on, class_object, ...)

NOTE: Versions of FlexMock prior to 0.6.0 used flexstub to create partial mocks. The flexmock method now assumes all the functionality that was spread out between two different methods. flexstub is deprecated, but still available for backward compatibility.

Once a mock is created, you need to define what that mock should expect to see. Expectation declarators are used to specify these expectations placed upon received method calls. A basic expectation, created with the should_receive method, just establishes the fact that a method may (or may not) be called on the mock object. Refinements to that expectation may be additionally declared. FlexMock always starts with the most general expectation and adds constraints to that.

For example, the following code:

    mock.should_receive(:average).and_return(12)

Means that the mock will now accept method calls to an average method. The expectation will accept any arguments and may be called any number of times (including zero times). Strictly speaking, the and_return part of the declaration isn't exactly a constraint, but it does specify what value the mock will return when the expectation is matched.

If you want to be more specific, you need to add additional constraints to your expectation. Here are some examples:

    mock.should_receive(:average).with(12).once

    mock.should_receive(:average).with(Integer).
        at_least.twice.at_most.times(10).
        and_return { rand }

Expectation are always matched in order of declaration. That means if you have a general expectation before a more specific expectation, the general expectation will have an opportunity to match first, effectively hiding the second expectation.

For example:

    mock.should_receive(:average)              # Matches any call to average
    mock.should_receive(:average).with(1).once # Fails because it never matches

In the example, the second expectation will never be triggered because all calls to average will be handled by the first expectation. Since the second expectation is require to match one time, this test will fail.

Reversing the order of the expections so that the more specific expectation comes first will fix that problem.

If an expectation has a count requirement (e.g. once or times), then once it has matched its expected number of times, it will let other expectations have a chance to match.

For example:

    mock.should_receive(:average).once.and_return(1)
    mock.should_receive(:average).once.and_return(2)
    mock.should_receive(:average).and_return(3)

In the example, the first time average is called, the first expectation is matched an average will return 1. The second time average is called, the second expectation matches and 2 is returned. For all calls to average after that, the third expectation returning 3 will be used.

Occasionally it is useful define a set of expecations in a setup method of a test and override those expectations in specific tests. If you mark an expectation with the by_default marker, that expectation will be used only if there are no non-default expectations on that method name. See "by_default" below.

The following methods may be used to create and refine expectations on a mock object. See theFlexMock::Expectation for more details.

• should_receive(method_name)

  Declares that a message named method_name will be sent to the mock object. Constraints on this expected message (called expectations) may be chained to the should_receive call.

• should_receive(method_name1, method_name2, ...)

  Define a number of expected messages that have the same constraints.

• should_receive(meth1 => result1, meth2 => result2, ...)

  Define a number of expected messages that have the same constrants, but return different values.

• should_receive(...).explicitly

  If a mock has a base class, use the explicitly modifier to override the restriction on method names imposed by the base class. The explicitly modifier must come immediately after the should_receive call and before any other expectation declarators.

  If a mock does not have a base class, this method has no effect.

• should_expect { |recorder| ... }

  Creates a mock recording object that will translate received method calls into mock expectations. The recorder is passed to a block supplied with the should_expect method. See examples below.

• with(arglist)

  Declares that this expectation matches messages that match the given argument list. The === operator is used on a argument by argument basis to determine matching. This means that most literal values match literally, class values match any instance of a class and regular expression match any matching string (after a to_s conversion). See argument validators (below) for details on argument validation options.

• with_any_args

  Declares that this expectation matches the message with any argument (default)

• with_no_args

  Declares that this expectation matches messages with no arguments

• zero_or_more_times

  Declares that the expected message is may be sent zero or more times (default, equivalent to at_least.never).

• once

  Declares that the expected message is only sent once. at_least / at_most modifiers are allowed.

• twice

  Declares that the expected message is only sent twice. at_least / at_most modifiers are allowed.

• never

  Declares that the expected message is never sent. at_least / at_most modifiers are allowed.

• times(n)

  Declares that the expected message is sent n times. at_least / at_most modifiers are allowed.

• at_least

  Modifies the immediately following message count constraint so that it means the message is sent at least that number of times. E.g. at_least.once means the message is sent at least once during the test, but may be sent more often. Both at_least and at_most may be specified on the same expectation.

• at_most

  Similar to at_least, but puts an upper limit on the number of messages. Both at_least and at_most may be specified on the same expectation.

• ordered

  Declares that the expected message is ordered and is expected to be received in a certain position in a sequence of messages. The message should arrive after and previously declared ordered messages and prior to any following declared ordered messages. Unordered messages are ignored when considering the message order.

  Normally ordering is performed only against calls in the same mock object. If the "globally" adjective is used, then ordering is performed against the other globally ordered method calls.

• ordered(group)

  Declare that the expected message belongs to an order group. Methods within an order group may be received in any order. Ordered messages outside the group must be received either before or after all of the grouped messages.

  For example, in the following, messages flip and flop may be received in any order (because they are in the same group), but must occur strictly after start but before end. The message any_time may be received at any time because it is not ordered.

    m = flexmock()
    m.should_receive(:any_time)
    m.should_receive(:start).ordered
    m.should_receive(:flip).ordered(:flip_flop_group)
    m.should_receive(:flop).ordered(:flip_flop_group)
    m.should_receive(:end).ordered

Normally ordering is performed only against calls in the same mock object. If the "globally" adjective is used, then ordering is performed against the other globally ordered method calls.

• globally.ordered

• globally.ordered(group_name)

  When modified by the "globally" adjective, the mock call will be ordered against other globally ordered methods in any of the mock objects in the same container (i.e. same test). All the options of the per-mock ordering are available in the globally ordered method calls.

• by_default

  Marks the expectation as a default. Default expectations act as normal as long as there are no non-default expectations for the same method name. As soon as a non-default expectation is defined, all default expectations for that method name are ignored.

  Default expectations allow you to setup a set of default behaviors for various methods in the setup of a test suite, and then override only the methods that need special handling in any given test.

Action expectations are used to specify what the mock should do when the expectation is matched. The actions themselves do not take part in determining whether a given expectation matches or not.

• and_return(value)

  Declares that the expected message will return the given value.

• and_return(value1, value2, ...)

  Declares that the expected message will return a series of values. Each invocation of the message will return the next value in the series. The last value will be repeatably returned if the number of matching calls exceeds the number of values.

• and_return { |*args, **kw, &block| code ... }

  Declares that the expected message will return the yielded value of the block. The block will receive all the arguments in the message.

• returns( ... )

  Alias for and_return.

• and_return_undefined

  Declares that the expected message will return a self-preserving undefined object (see FlexMock::Undefined for details).

• returns_undefined

  Alias for and_returns_undefined

• and_raise(exception, *args)

  Declares that the expected message will raise the specified exception. If exception is an exception class, then the raised exception will be constructed from the class with new given the supplied arguments. If exception is an instance of an exception class, then it will be raised directly.

• raises( ... )

  Alias for and_raise.

• and_throw(symbol)

• and_throw(symbol, value)

  Declares that the expected messsage will throw the specified symbol. If an optional value is included, then it will be the value returned from the corresponding catch statement.

• throws( ... )

  Alias for and_throw.

• and_yield(values, ...)

  Declares that the mocked method will receive a block, and the mock will call that block with the values given. Not providing a block will be an error. Providing more than one and_yield clause one a single expectation will mean that subsquent mock method calls will yield the values provided by the additional and_yield clause.

• yields( ... )

  Alias for and_yield( ... ).

• and_iterates(value1, value2>, ...)

  Declares that the mocked method will receive a block, and the mock will iterate over the values given, calling the block once for each value. Not providing a block will be an error. Providing more than one and_iterates or and_yield clause one a single expectation will mean that subsquent mock method calls will yield the values provided by the additional and_iterates/and_yield clause.

• pass_thru

• pass_thru { |value| .... }

  Declares that the expected message will allow the method to be passed to the original method definition in the partial mock object. pass_thru is also allowed on regular mocks, but since there is no original method to be called, pass_thru will always return the undefined object.

  If a block is supplied to pass_thru, the value returned from the original method will be passed to the block and the value of the block will be returned. This allows you to mock methods on the returned value.

    Dog.should_receive(:new).pass_thru { |dog|
      flexmock(dog, :wag => true)
    }

• mock

  Expectation constraints always return the expectation so that the constraints can be chained. If you wish to do a one-liner and assign the mock to a variable, the mock method on an expectation will return the original mock object.

    m = flexmock.should_receive(:hello).once.and_return("World").mock

NOTE: Using mock when specifying a Demeter mock chain will return the last mock of the chain, which might not be what you expect.

The values passed to the with declarator determine the criteria for matching expectations. The first expectation found that matches the arguments in a mock method call will be used to validate that mock method call.

The following rules are used for argument matching:

• A with parameter that is a class object will match any actual argument that is an instance of that class.

  Examples:

     with(Integer)     will match    f(3)

• A regular expression will match any actual argument that matches the regular expression. Non-string actual arguments are converted to strings via to_s before applying the regular expression.

  Examples:

    with(/^src/)      will match    f("src_object")
    with(/^3\./)      will match    f(3.1415972)

• Most other objects will match based on equal values.

  Examples:

      with(3)         will match    f(3)
      with("hello")   will match    f("hello")

• If you wish to override the default matching behavior and force matching by equality, you can use the FlexMock.eq convenience method. This is mostly used when you wish to match class objects, since the default matching behavior for class objects is to match instances, not themselves.

  Examples:

      with(eq(Integer))             will match       f(Integer)
      with(eq(Integer))             will NOT match   f(3)

Note: If you do not use the FlexMock::TestCase Test Unit integration module, or the FlexMock::ArgumentTypes module, you will have to fully qualify the eq method. This is true of all the special argument matches (eq, on, any, hsh and ducktype).

      with(FlexMock.eq(Integer))
      with(FlexMock.on { code })
      with(FlexMock.any)
      with(FlexMock.hsh(:tag => 3))
      with(FlexMock.ducktype(:wag, :bark))

• If you wish to match a hash on some of its values, the FlexMock.hsh(...) method will work. Only specify the hash values you are interested in, the others will be ignored.

      with(hsh(:run => true))  will match    f(:run => true, :stop => false)

• If you wish to match any object that responds to a certain set of methods, use the FlexMock.ducktype method.

      with(ducktype(:to_str))     will match   f("string")
      with(ducktype(:wag, :bark)) will match   f(dog)
                                  (assuming dog implements wag and bark)

• If you wish to match anything, then use the FlexMock.any method in the with argument list.

  Examples (assumes either the FlexMock::TestCase or FlexMock::ArgumentTypes mix-ins has been included):

      with(any)             will match       f(3)
      with(any)             will match       f("hello")
      with(any)             will match       f(Integer)
      with(any)             will match       f(nil)

• If you wish to specify a complex matching criteria, use the FlexMock.on(&block) with the logic contained in the block.

  Examples (assumes FlexMock::ArgumentTypes has been included):

      with(on { |arg| (arg % 2) == 0 } )

will match any even integer.

• By default, flexmock will ignore given blocks, that is it will assume that blocks are optional.

• If you wish to verify that a method call received a block, use with_block

  Example:

      m.should_receive(:foo).with(Integer).with_block.and_return(:got_block)
      m.should_receive(:foo).with(Integer).and_return(:no_block)

will cause the mock to return the following:

     m.foo(1) { } => returns :got_block
     m.foo(1)     => returns :no_block

• If you wish to verify that a method call does not receive a block, use with_no_block

  Example:

      m.should_receive(:foo).with(Integer).with_no_block.and_return(:no_block)
      m.should_receive(:foo).with(Integer).and_return(:got_block)

will cause the mock to return the following:

     m.foo(1) { } => returns :got_block
     m.foo(1)     => returns :no_block

Sometimes it is useful to mock the behavior of one or two methods in an existing object without changing the behavior of the rest of the object. If you pass a real object to the flexmock method, it will allow you to use that real object in your test and will just mock out the one or two methods that you specify.

For example, suppose that a Dog object uses a Woofer object to bark. The code for Dog looks like this (we will leave the code for Woofer to your imagination):

  class Dog
    def initialize
      @woofer = Woofer.new
    end
    def bark
      @woofer.woof
    end
    def wag
      :happy
    end
  end

Now we want to test Dog, but using a real Woofer object in the test is a real pain (why? ... well because Woofer plays a sound file of a dog barking, and that's really annoying during testing).

So, how can we create a Dog object with mocked Woofer? All we need to do is allow FlexMock to replace the bark method.

Here's the test code:

  class TestDogBarking < Test::Unit::TestCase
    include FlexMock::TestCase

    # Setup the tests by mocking the +new+ method of
    # Woofer and return a mock woofer.
    def setup
      @dog = Dog.new
      flexmock(@dog, :bark => :grrr)
    end

    def test_dog
      assert_equal :grrr, @dog.bark   # Mocked Method
      assert_equal :happy, @dog.wag    # Normal Method
    end
  end

The nice thing about this technique is that after the test is over, the mocked out methods are returned to their normal state. Outside the test everything is back to normal.

NOTE: In previous versions of FlexMock, partial mocking was called "stubs" and the flexstub method was used to create the partial mocks. Although partial mocks were often used as stubs, the terminology was not quite correct. The current version of FlexMock uses the flexmock method to create both regular stubs and partial stubs. A version of the flexstub method is included for backwards compatibility. See Martin Fowler's article Mocks Aren't Stubs for a better understanding of the difference between mocks and stubs.

This partial mocking technique was inspired by the Stuba library in the Mocha project.

FlexMock supports spy-like mocks as well as the traditional mocks.

  # In Test::Unit / MiniTest
  class TestDogBarking < Test::Unit::TestCase
    def test_dog
      dog = flexmock(:on, Dog)
      dog.bark("loud")
      assert_spy_called dog, :bark, "loud"
    end
  end

  # In RSpec
  describe Dog do
    let(:dog) { flexmock(:on, Dog) }
    it "barks loudly" do
      dog.bark("loud")
      dog.should have_received(:bark).with("loud")
    end
  end

Since spies are verified after the code under test is run, they fit very nicely with the Given/When/Then technique of specification. Here is the above RSpec example using the rspec-given gem:

  require 'rspec/given'

  describe Dog do
    Given(:dog) { flexmock(:on, Dog) }

    context "when barking loudly" do
      When { dog.bark("loud") }
      Then { dog.should have_received(:bark).with("loud") }
    end
  end

NOTE: You can only spy on methods that are mocked or stubbed. That's not a problem with regular mocks, but normal methods on partial objects will not be recorded.

You can get around this limitation by stubbing the method in question on the normal mock, and then specifying pass_thru. Assuming :bark is a normal method on a Dog object, then the following allows for spying on :bark.

   dog = Dog.new
   flexmock(dog).should_receive(:bark).pass_thru
   # ...
   dog.should have_received(:bark)

FlexMock provied a custom assertion method for use with Test::Unit and MiniTest for asserting that mocked methods are actually called.

• assert_spy_called mock, options_hash, method_name, args...

  This will assert that the method called method_name has been called at least once on the given mock object. If arguments are given, then the method must be called with actual argument that match the given argument matchers.

  All the argument matchers defined in the "Argument Validation" section above are allowed in the assert_spy_called method.

  The options hash is optional. If omitted, all options will have their default values. See below for spy option definitions.

• assert_spy_not_called mock, options_hash, method_name, args...

  Same as assert_spy_called, except with the sense of the test reversed.

Spy Options

• times: n

  Specify the number of times a matching method should have been invoked. nil (or omitted) means any number of times.

• with_block: true/false/nil

  Is a block required on the invocation? true means the method must be invoked with a block. false means the method must have been invoked without a block. nil means that the presence of a block does not matter. Default is nil.

• and: [proc1, proc2...]

  Additional validations to be run on each matching method call. The list of arguments for each call is passed to the procs. This allows additional validations on supplied arguments. Default is no additional validations.

• on: n

  Only apply the additional validations on the n'th invocation of the matching method. Default is apply additional validations to all invocations.

Examples:

    dog = flexmock(:on, Dog)

    dog.wag(:tail)
    dog.wag(:head)
    dog.bark(5)
    dog.bark(6)

    assert_spy_called dog, :wag, :tail
    assert_spy_called dog, :wag, :head
    assert_spy_called dog, {times: 2}, :wag

    assert_spy_not_called dog, :bark
    assert_spy_not_called dog, {times: 3}, :wag

    is_even = proc { |n| assert_equal 0, n%2 }
    assert_spy_called dog, { and: is_even, on: 2 }, :bark, Integer

FlexMock also provides an RSpec matcher that can be used to specify spy behavior.

• mock.should have_received(method_name).modifier1.modifier2...

  Specifies that the method named method_name should have been received by the mock object with the given arguments.

  Just like should_receive, have_received will accept a number of modifiers that modify its behavior.

Modifiers for have_received

• with(args)

  If a with modifier is given, only messages with matching arguments are considered. args can be any of the argument matches mentioned in the "Argument Validation" section above. If with is not given, then the arguments are not considered when finding matching calls.

• times(n)

  If a times modifier is given, then there must be exactly n calls for that method name on the mock. If the times clause is not given, then there must be at least one call matching the method name (and arguments if they are considered).

  • never is an alias for times(0),
  • once is an alias for times(1), and
  • twice is an alias for times(2).

• and { |args| code }

  If an and modifier is given, then the supplied block will be run as additional validations on any matching call. Arguments to the matching call will be supplied to the block. If multiple and modifiers are given, all the blocks will be run. The additional validations are run on all the matching calls unless an on modifier is supplied.

• on(n)

  If an on modifier is given, then the additional validations supplied by and will only be run on the n'th invocation of the matching method.

Examples:

    dog = flexmock(:on, Dog)

    dog.wag(:tail)
    dog.wag(:head)

    dog.should have_received(:wag).with(:tail)
    dog.should have_received(:wag).with(:head)
    dog.should have_received(:wag).twice

    dog.should_not have_received(:bark)
    dog.should_not have_received(:wag).times(3)

    dog.bark(3)
    dog.bark(6)
    dog.should have_received(:bark).with(Integer).and { |arg|
      (arg % 3).should == 0
    }
    dog.should have_received(:bark).with(Integer).and { |arg|
      arg.should == 6
    }.on(2)

In the previous example we mocked out the bark method of a Dog object to avoid invoking the Woofer object. Perhaps a better technique would be to mock the Woofer object directly. But Dog uses Woofer explicitly so we cannot just pass in a mock object for Dog to use.

But wait, we can add mock behavior to any existing object, and classes are objects in Ruby. So why don't we just mock out the Woofer class object to return mocks for us.

  class TestDogBarking < Test::Unit::TestCase
    include FlexMock::TestCase

    # Setup the tests by mocking the `new` method of
    # Woofer and return a mock woofer.
    def setup
      flexmock(Woofer).should_receive(:new).
         and_return(flexmock(:woof => :grrr))
      @dog = Dog.new
    end

    def test_dog
      assert_equal :grrrr, @dog.bark  # Calls woof on mock object
                                      # returned by Woofer.new
    end
  end

Sometimes returning a single mock object is not enough. Occasionally you want to mock every instance object created by a class. FlexMock makes this very easy.

  class TestDogBarking < Test::Unit::TestCase
    include FlexMock::TestCase

    # Setup the tests by mocking Woofer to always
    # return partial mocks.
    def setup
      flexmock(Woofer).new_instances.should_receive(:woof => :grrr)
    end

    def test_dog
      assert_equal :grrrr, Dog.new.bark  # All dog objects
      assert_equal :grrrr, Dog.new.bark  # are mocked.
    end
  end

Note that FlexMock adds the mock expectations after the original new method has completed. If the original version of new yields the newly created instance to a block, that block will get an non-mocked version of the object.

Note that new_instances will accept a block if you wish to mock several methods at the same time. E.g.

      flexmock(Woofer).new_instances do |m|
        m.should_receive(:woof).twice.and_return(:grrr)
        m.should_receive(:wag).at_least.once.and_return(:happy)
      end

Sometimes you want to setup a bunch of default expectations that are pretty much for a number of different tests. Then in the individual tests, you would like to override the default behavior on just that one method you are testing at the moment. You can do that by using the by_default modifier.

In your test setup you might have:

  def setup
    @mock_dog = flexmock("Fido")
    @mock_dog.should_receive(:tail => :a_tail, :bark => "woof").by_default
  end

The behaviors for :tail and :bark are good for most of the tests, but perhaps you wish to verify that :bark is called exactly once in a given test. Since :bark by default has no count expectations, you can override the default in the given test.

  def test_something_where_bark_must_be_called_once
    @mock_dog.should_receive(:bark => "woof").once

    # At this point, the default for :bark is ignored,
    # and the "woof" value will be returned.

    # However, the default for :tail (which returns :a_tail)
    # is still active.
  end

By setting defaults, your individual tests don't have to concern themselves with details of all the default setup. But the details of the overrides are right there in the body of the test.

The Law of Demeter says that you should only invoke methods on objects to which you have a direct connection, e.g. parameters, instance variables, and local variables. You can usually detect Law of Demeter violations by the excessive number of periods in an expression. For example:

     car.chassis.axle.universal_joint.cog.turn

The Law of Demeter has a very big impact on mocking. If you need to mock the "turn" method on "cog", you first have to mock chassis, axle, and universal_joint.

    # Manually mocking a Law of Demeter violation
    cog = flexmock("cog")
    cog.should_receive(:turn).once.and_return(:ok)
    joint = flexmock("gear", :cog => cog)
    axle = flexmock("axle", :universal_joint => joint)
    chassis = flexmock("chassis", :axle => axle)
    car = flexmock("car", :chassis => chassis)

Yuck!

The best course of action is to avoid Law of Demeter violations. Then your mocking exploits will be very simple. However, sometimes you have to deal with code that already has a Demeter chain of method calls. So for those cases where you can't avoid it, FlexMock will allow you to easily mock Demeter method chains.

Here's an example of Demeter chain mocking:

    # Demeter chain mocking using the short form.
    car = flexmock("car")
    car.should_receive( "chassis.axle.universal_joint.cog.turn" => :ok).once

You can also use the long form:

    # Demeter chain mocking using the long form.
    car = flexmock("car")
    car.should_receive("chassis.axle.universal_joint.cog.turn").once.
      and_return(:ok)

That's it. Anywhere FlexMock accepts a method name for mocking, you can use a demeter chain and FlexMock will attempt to do the right thing.

But beware, there are a few limitations.

The all the methods in the chain, except for the last one, will mocked to return a mock object. That mock object, in turn, will be mocked so as to respond to the next method in the chain, returning the following mock. And so on. If you try to manually mock out any of the chained methods, you could easily interfer with the mocking specified by the Demeter chain. FlexMock will attempt to catch problems when it can, but there are certainly scenarios where it cannot detect the problem beforehand.

Refer to the following documents for examples of using FlexMock:

• RSpec Examples
• Test::Unit / MiniTest Examples

Copyright 2003-2013 by Jim Weirich (jim.weirich@gmail.com). Copyright 2014- by Sylvain Joyeux (sylvain.joyeux@m4x.org)

Licensed under the MIT license

• Author -- Jim Weirich jim.weirich@gmail.com and Sylvain Joyeux sylvain.joyeux@m4x.org
• Requires -- Ruby 2.0 or later

If you like the spy capability of FlexMock, you should check out the rspec-given gem that allows you to use Given/When/Then statements in you specifications.

This software is provided "as is" and without any express or implied warranties, including, without limitation, the implied warranties of merchantibility and fitness for a particular purpose.