1. Summary
  2. Files
  3. Support
  4. Report Spam
  5. Create account
  6. Log in

Turtle

From turtle

Jump to: navigation, search
This is the documentation for the 1.2.0 version of Turtle, for the latest version see http://turtle.sourceforge.net

A C++ Mock_object library based on Boost with a focus on usability, simplicity and flexibility.

Contents

Motivation

Consider a (very) simple calculator class :

class calculator
{
public:
    int add( int a, int b );
};

Obviously writing unit tests for such a class is trivial, one of them could be :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    calculator c;
    BOOST_CHECK_EQUAL( 0, c.add( 0, 0 ) );
}

What now if the calculator class looks more like this :

class view
{
public:
    virtual void display( int result ) = 0;
};

class calculator
{
public:
    calculator( view& v );

    void add( int a, int b ); // the result will be sent to the view 'v'
};

Writing unit tests becomes a bit more tedious and requires some boiler-plate code, for instance :

class my_view : public view
{
public:
    my_view()
        : called( false )
    {}
    virtual void display( int result )
    {
        called = true;
        value = result;
    }
    bool called;
    int value;
};

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    my_view v;
    calculator c( v );
    c.add( 0, 0 );
    BOOST_REQUIRE( v.called );
    BOOST_CHECK_EQUAL( 0, v.value );
}

Mock objects main purpose is to alleviate the user from the burden of writing all this boiler-plate code.

Here is how the last test can be rewritten using a mock object :

MOCK_BASE_CLASS( mock_view, view ) // declare a 'mock_view' class implementing 'view'
{
    MOCK_METHOD( display, 1 )      // implement the 'display' method from 'view' (taking 1 argument)
};

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    mock_view v;
    calculator c( v );
    MOCK_EXPECT( v.display ).once().with( 0 ); // expect the 'display' method to be called once (and only once) with a parameter value equal to 0
    c.add( 0, 0 );
}

And all the checks are handled by the library.

Getting started

This section introduces most of the library features in a series of use cases built on the example from the motivation section.

For all the code examples the following is assumed :

#define BOOST_AUTO_TEST_MAIN
#include <boost/test/auto_unit_test.hpp>
#include <turtle/mock.hpp>

Create, expect, trigger, verify

A simple unit test with mock objects usually splits into several phases as illustrated by :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
   mock_view v;                               // create mock objects
   calculator c( v );                         // create object under test
   MOCK_EXPECT( v.display ).once().with( 0 ); // configure mock objects
   c.add( 0, 0 );                             // exercise object under test
}                                             // verify mock objects

Triggering the object under test in turn calls methods on the mock objects, and any unexpected call raises an error.

Mock objects are automatically verified during their destruction and an error is signalled if any unfulfilled expectation remains.

More sophisticated tests sometimes require more complex use cases and in particular might need to :

  • manually verify mock objects
  • manually reset mock objects

Here is an example highlighting the different possibilities :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
   mock_view v;
   calculator c( v );
   MOCK_EXPECT( v.display ).once().with( 0 );
   c.add( 0, 0 );
   MOCK_VERIFY( v.display );                  // verify all expectations are fulfilled for the 'display' method
   mock::verify( v );                         // verify all expectations are fulfilled for all methods of 'v'
   mock::verify();                            // verify all expectations are fulfilled for all existing mock objects
   MOCK_RESET( v.display );                   // reset all expectations for the 'display' method
   mock::reset( v );                          // reset all expectations for all methods of 'v'
   mock::reset();                             // reset all expectations for all existing mock objects
}                                             // automatically verify all expectations are fulfilled for all mock objects going out of scope

Note that all verifications upon destruction will be disabled if the mock objects are destroyed in the context of an exception being raised.

Expectation selection algorithm

A method can be configured with several expectations, for instance :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    mock_view v;
    calculator c( v );
    MOCK_EXPECT( v.display ).once().with( 0 ); // this call must occur once (and only once)
    MOCK_EXPECT( v.display ).with( 1 );        // this call can occur any number of times (including never)
    c.add( 0, 0 );
}

Each method call is then handled by processing the expectations in the order they have been defined :

  1. looking for a match with valid parameter constraints evaluated from left to right
  2. checking that the invocation count for this match is not exhausted

An error is raised if none can be found.

By default the relative order of the calls does not matter. It can however be enforced :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    mock_view v;
    calculator c( v );
    mock::sequence s;
    MOCK_EXPECT( v.display ).once().with( 0 ).in( s ); // add this expectation to the sequence
    MOCK_EXPECT( v.display ).with( 1 ).in( s );        // add this expectation to the sequence after the previous one
    c.add( 0, 0 );
    c.add( 1, 0 );
}

Therefore an error will be issued if the second expectation is matched before the first one has been exhausted.

An expectation can be part of several sequences :

BOOST_AUTO_TEST_CASE( zero_plus_zero_is_zero )
{
    mock_view v;
    calculator c( v );
    mock::sequence s1, s2;
    MOCK_EXPECT( v.display ).once().with( 0 ).in( s1 );
    MOCK_EXPECT( v.display ).once().with( 1 ).in( s2 );
    MOCK_EXPECT( v.display ).with( 2 ).in( s1 ).in( s2 ); // add this expectation to both sequences after the previous ones
    c.add( 0, 0 );
    c.add( 1, 0 );
    c.add( 1, 1 );
    c.add( 2, 0 );
}

Error diagnostic

During the execution of a test case, an error can happen for one of the following reasons :

  • unexpected call when no match can be found for the given arguments (typically logs an error and throws an exception)
  • sequence failure when an enforced call sequence has not been followed (typically logs an error and throws an exception)
  • verification failure if a remaining match has not been fulfilled upon manual verification (typically logs an error)
  • untriggered expectation if a remaining match has not been fulfilled when destroying the mock object (typically logs an error)
  • missing action if a method supposed to return something else than void has not been configured properly (typically logs an error and throws an exception)

The exact type of the exception thrown depends on the #Error_policy used.

An error log typically looks like :

unknown location(0): error in "zero_plus_zero_is_zero": unexpected call: v.mock_view::display( 0 )
v once().with( 1 )
v once().with( 2 )
. once().with( 3 )

On the first line is the description of what happened : here the display method of object v of class mock_view has been called with an actual value of 0.
The following lines list the set expectations with the check (the v character) meaning the expectation has been exhausted.
It therefore means that the two first expectations have been fulfilled by two calls, and then instead of 3 in the third call 0 has been erroneously passed on to the mock object.

Another common error looks like :

src/tests/turtle_test/Tutorial.cpp(73): error in "zero_plus_zero_is_zero": untriggered expectation: v.mock_view::display
v once().with( 1 )
v once().with( 2 )
. once().with( 3 )

The first line tells that a set expectation has not been fulfilled. The file and line number give the location where the corresponding expectation has been configured.
The following lines once again list the set expectations.
It means the two first calls correctly passed the expected values to the mock object, but then no third call happened.

Boost.Test integration

The only requirement when using Boost.Test is to include it before the mock library, for instance :

#include <boost/test/auto_unit_test.hpp>
#include <turtle/mock.hpp>

This allows the mock library to detect the test framework and to provide various specific error handling mechanisms, see #Error_policy.

Alternatively if for some reason Boost.Test cannot be included before the mock library, the following define can be used :

#define MOCK_USE_BOOST_TEST
#include <turtle/mock.hpp>

Customisation

This section explains how to customise different aspects of the library.

Logging

The library will perform logging lazily, e.g. only when actually needed, which is usually because an error happens but it depends on the #Error_policy used.
Parameters and #Constraints are serialized to report meaningful diagnostics of the failures.

By default the library attempts to serialize to an std::ostream and if this is not possible will use a '?'.
Any incomplete type is gracefully handled and yields a '?'.

If for some reason the serialization to an std::ostream shouldn't be used, it can be overridden by a serialization operator to a mock::stream, for instance to log user_type declared in user_namespace :

namespace user_namespace
{
    inline mock::stream& operator<<( mock::stream& s, const user_type& t )
    {
        return s << ...
    }
}

The operators are found using argument-dependent name lookup (e.g. ADL) which means it needs to be in the namespace of either one of its arguments. The easiest is to define them in the same namespace as the type being serialized. If this is not possible (for instance when serializing a type in namespace std because the C++ standard explicitly forbids adding definitions into the std namespace) a serialization operator to mock::stream can be in the mock namespace instead.

The serialization operators detection doesn't attempt to do conversions when finding a match (because this can sometimes yield an ambiguous resolution error).
As conversions can prove convenient, for instance when dealing with a base class which is derived to a lot of sub-classes, they can be activated by defining MOCK_USE_CONVERSIONS prior to including the library :

#define MOCK_USE_CONVERSIONS
#include <turtle/mock.hpp>

Be aware though that in this case the compiler can produce a compilation error when attempting to detect whether serialization operators exist or not.
It is always possible however to define a serialization operator to a mock::stream in order to bypass the detection.

In all custom implementations it is probably a good thing for most of the data to recursively rely on the same mechanism the library uses in order to log everything, for instance this is how std::pair is handled :

namespace mock
{
    template< typename T1, typename T2 >
    mock::stream& operator<<( mock::stream& s, const std::pair< T1, T2 >& p )
    {
        return s << '(' << mock::format( p.first ) << ',' << mock::format( p.second ) << ')';
    }
}

The interesting part is the call to mock::format which is merely a helper to enable the whole can-be-serialized-or-? logics.

Constraints

A constraint provides a means to validate a parameter received in a call to a mock object.

The library comes with a set of pre-defined constraints matching the most widely used cases, see #Adding_constraints.
From time to time however it is rather common to have the need to perform a custom validation.

Creating a constraint is as simple as writing a function, for instance :

bool custom_constraint( int actual )
{
   return actual == 42;
}

Any functor will actually do as long as its signature matches the requirement : take a type convertible from the actual type and return a boolean.

Using the custom constraint is also pretty trivial, for instance :

BOOST_AUTO_TEST_CASE( forty_one_plus_one_is_forty_two )
{
   mock_view v;
   calculator c( v );
   MOCK_EXPECT( v.display ).with( &custom_constraint );
   c.add( 41, 1 );
}

Simple enough, however this constraint isn't serializable and thus yields a rather uninformative '?' in the logs.
Just like a parameter, a constraint can be displayed in a readable form using its serialization operator, see #Logging.
Thus for a widely used constraint (for instance one shipped with the code of a library) it is likely better to define it like this :

struct custom_constraint
{
    friend bool operator==( int actual, const custom_constraint& )
    {
        return actual == 42;
    }

    friend std::ostream& operator<<( std::ostream& s, const custom_constraint& )
    {
        return s << "_ == 42";
    }
};

And of course the constraint is to be used in a slightly different manner :

BOOST_AUTO_TEST_CASE( forty_one_plus_one_is_forty_two )
{
   mock_view v;
   calculator c( v );
   MOCK_EXPECT( v.display ).with( custom_constraint() );
   c.add( 41, 1 );
}

Actually real world use cases sometimes need several other features as well :

  • a state
  • (template) parameters
  • an operator with one or several (template) signatures

Therefore a more realistic and complete example would be :

template< typename Expected >
struct near_constraint
{
    near_constraint( Expected expected, Expected threshold )
      : expected_( expected )
      , threshold_( threshold )
    {}

    template< typename Actual >
    bool operator()( Actual actual ) const
    {
        return std::abs( actual - boost::unwrap_ref( expected_ ) )
            < boost::unwrap_ref( threshold_ );
    }

    friend std::ostream& operator<<( std::ostream& s, const near_constraint& c )
    {
        return s << "near( " << mock::format( c.expected_ )
            << ", " << mock::format( c.threshold_ ) << " )";
    }

    Expected expected_, threshold_;
};

template< typename Expected >
mock::constraint< near_constraint< Expected > > near( Expected expected, Expected threshold )
{
    return near_constraint< Expected >( expected, threshold );
}

And it would be used like this :

BOOST_AUTO_TEST_CASE( forty_one_plus_one_is_forty_two_plus_or_minus_one )
{
   mock_view v;
   calculator c( v );
   MOCK_EXPECT( v.display ).with( near( 42, 1 ) );
   c.add( 41, 1 );
}

The purpose of the 'near' template function is to :

  • remove the burden of specifying the template parameter when instantiating near_constraint
  • wrap the constraint in a mock::constraint so that it plays nicely with the operators (e.g. !, && and ||)

The use of boost::unwrap_ref provides support for passing arguments as references with boost::ref and boost::cref and delaying their initialization, for instance :

BOOST_AUTO_TEST_CASE( forty_one_plus_one_is_forty_two_plus_or_minus_one )
{
   mock_view v;
   calculator c( v );
   int expected, threshold;
   MOCK_EXPECT( v.display ).with( near( boost::cref( expected ), boost::cref( threshold ) ) );
   expected = 42;
   threshold = 1;
   c.add( 41, 1 );
}

See #Adding_constraints for an explanation of how the library detects whether an argument is a functor or a value.

For more information about the serialization operator and the use of mock::format, refer to #Logging.

Number of arguments

The maximum number of arguments a mocked method can have is defined by MOCK_MAX_ARGS.
By default this value is set to 9, but if needed it can be changed to another value before including the library :

#define MOCK_MAX_ARGS 20
#include <turtle/mock.hpp>

This means methods with up to 20 arguments will then be accepted.

The mock object library uses several boost libraries and will adjust some of their constants if they haven't already been defined :

  • Boost.Function with BOOST_FUNCTION_MAX_ARGS required at MOCK_MAX_ARGS or higher
  • Boost.FunctionTypes with BOOST_FT_MAX_ARITY required at MOCK_MAX_ARGS + 1 or higher
  • Boost.Phoenix (when increasing MOCK_MAX_ARGS over 9) with PHOENIX_LIMIT required at MOCK_MAX_ARGS or higher

A compilation error will happen if one of those constants is already defined too low.

Error policy

Integrating the library with any given unit test framework can be done simply by defining a custom error policy.

The library provides two error policies :

  • default_error_policy
    • logs to std::cerr
    • throws mock::exception
  • boost_test_error_policy is automatically enabled if Boost.Test is detected, see #Boost.Test integration
    • logs using the logger from Boost.Test
    • throws mock::exception deriving from boost::execution_aborted via boost::enable_current_exception

A custom error policy needs to implement the following concept :

template< typename Result >
struct custom_policy
{
   static Result abort()
   {
       // ...
   }
   static void checkpoint( const char* file, int line )
   {
       // ...
   }
   template< typename Context >
   static void unexpected_call( const Context& context )
   {
       // ...
   }
   template< typename Context >
   static void expected_call( const Context& context, const char* file, int line )
   {
       // ...
   }
   template< typename Context >
   static void missing_action( const Context& context, const char* file, int line )
   {
       // ...
   }
   template< typename Context >
   static void sequence_failed( const Context& context, const char* file, int line )
   {
       // ...
   }
   template< typename Context >
   static void verification_failed( const Context& context, const char* file, int line )
   {
       // ...
   }
   template< typename Context >
   static void untriggered_expectation( const Context& context, const char* file, int line )
   {
       // ...
   }
};

The context, which stands for "something serializable to an std::ostream", is actually built only if an attempt to serialize it is made, thus enabling lazy serialization of all elements (e.g. constraints and parameters).
File and line show were the expectation has been configured.

The policy can then be activated by defining MOCK_ERROR_POLICY prior to including the library :

#define MOCK_ERROR_POLICY custom_policy
#include <turtle/mock.hpp>


Reference

This section describes the library syntax exhaustively.

All source code snippets assume the following prerequisite :

#include <turtle/mock.hpp>

Creation

Mock objects can be assigned and copied around freely, unless they derive from a type which disables it.
Copies of a mock object share the same internal state, meaning setting an expectation on one of them will impact all of them. Thus it is possible to let an object under test copy a mock object and still be able to set, verify or reset expectations.

The library defines a set of convenient macros for creating mock objects of different kinds :

  • classes
  • functors
  • functions

It is worth noting that creating a mock object creates a new object and does not magically replace existing ones. For instance creating a mock function will not replace an already existing function with the same name and signature.

Creating a mock object mainly does two things under the hood :

  • define an object
  • declare an identifier for manipulating the object

Most of the time the identifier will be identical to the object name, but in case of ambiguity (for instance overloaded methods) a different identifier will have to be specified.

Class

Synopsis :

MOCK_CLASS( class )
{
};

template< typename T >
MOCK_CLASS( class )
{
};

struct class : mock::object
{
};

template< typename T >
struct mock_class : mock::object
{
};

MOCK_BASE_CLASS( class, base_class )
{
};

template< typename T >
MOCK_BASE_CLASS( class, base_class< T > )
{
};

struct mock_class : base_class, mock::object
{
};

template< typename T >
struct mock_class : base_class< T >, mock::object
{
};

Example :

MOCK_CLASS( mock_class )
{
};

BOOST_AUTO_TEST_CASE( demonstrates_instantiating_a_mock_class )
{
    mock_class c;
}

Example :

template< typename T >
MOCK_CLASS( mock_class )
{
};

BOOST_AUTO_TEST_CASE( demonstrates_instantiating_a_template_mock_class )
{
    mock_class< int > c;
}

Example :

class base_class
{
};

MOCK_BASE_CLASS( mock_class, base_class )
{
};

BOOST_AUTO_TEST_CASE( demonstrates_instantiating_a_derived_mock_class )
{
    mock_class c;
}

Example :

template< typename T >
class base_class
{
};

template< typename T >
MOCK_BASE_CLASS( mock_class, base_class< T > )
{
};

BOOST_AUTO_TEST_CASE( demonstrates_instantiating_a_template_derived_mock_class )
{
    mock_class< int > c;
}

Using the macro or deriving from mock::object is optional but provides the additional following features :

  • #Verify and #Reset can be done on the object as a whole in order to verify/reset all the expectations for all its methods at once
  • logs involving the object are enhanced because configuring an expectation for a method will set the class name for all the other methods as well

Member function

Synopsis :

MOCK_METHOD( method_name, arity )                                          // only works in the context of a [derived [#Class]] and generates both const and non-const methods

MOCK_METHOD_EXT( method_name, arity, signature, identifier )               // generates both const and non-const methods
MOCK_CONST_METHOD_EXT( method_name, arity, signature, identifier )         // generates only the const version of the method
MOCK_NON_CONST_METHOD_EXT( method_name, arity, signature, identifier )     // generates only the non-const version of the method

MOCK_METHOD_EXT_TPL( method_name, arity, signature, identifier )           // must be used if the signature uses a template parameter of the class
MOCK_CONST_METHOD_EXT_TPL( method_name, arity, signature, identifier )     // must be used if the signature uses a template parameter of the class
MOCK_NON_CONST_METHOD_EXT_TPL( method_name, arity, signature, identifier ) // must be used if the signature uses a template parameter of the class

Example :

class base_class
{
    virtual void method( int ) = 0;
};

MOCK_BASE_CLASS( mock_class, base_class )
{
    MOCK_METHOD( method, 1 )              // only possible when referring unambiguously to a base class method
};

Example :

class base_class
{
    virtual void method( int, const std::string& ) = 0;
    virtual void method( float ) = 0;
};

MOCK_BASE_CLASS( mock_class, base_class )
{
    MOCK_METHOD_EXT( method, 2, void( int, const std::string& ), identifier_1 ) // MOCK_METHOD cannot be used because of overloading
    MOCK_METHOD_EXT( method, 1, void( float ), identifier_2 )                   // the identifier must differ from the previous one in order to fully disambiguate methods
};

Example :

class base_class
{
    virtual void method( float ) = 0;
    virtual void method( float ) const = 0;
};

MOCK_BASE_CLASS( mock_class, base_class )
{
    MOCK_METHOD( method, 1 )             // this generates both const and non-const versions
};

Example :

class base_class
{
    virtual void method( float ) = 0;
    virtual void method( float ) const = 0;
};

MOCK_BASE_CLASS( mock_class, base_class )
{
    MOCK_CONST_METHOD_EXT( method, 1, void( float ), identifier_1 )     // this generates only the const version
    MOCK_NON_CONST_METHOD_EXT( method, 1, void( float ), identifier_2 ) // this generates only the non-const version, with a different identifier
};

Example :

class base_class
{
    virtual void method( float ) = 0;
};

struct mock_class : base_class
{
    typedef base_class base_type; // this is required for MOCK_METHOD to work when not using MOCK_BASE_CLASS

    MOCK_METHOD( method, 1 )
};

Example :

MOCK_CLASS( mock_class )
{
    MOCK_NON_CONST_METHOD_EXT( operator=, 1, mock_class&( const mock_class& ), assignment ) // operators need a custom identifier
};

Note that a #Conversion operator requires special care.

Example :

template< typename T >
MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT_TPL( method, 1, void( const T& ), method ) // the _TPL variants must be used if the signature includes a template type
};

Static member function

Synopsis :

MOCK_STATIC_METHOD( method, arity, signature, identifier )
MOCK_STATIC_METHOD_TPL( method, arity, signature, identifier ) // must be used if the signature uses a template parameter of the class

Example :

MOCK_CLASS( mock_class )
{
    MOCK_STATIC_METHOD( method, 1, float( int ), method )
};

A static object is used behind the scene in order to keep track of the expectations of a mock static method, therefore to ensure all tests run in isolation it is strongly suggested to manually #Verify and #Reset the static method at the end of each test.

Constructor

Synopsis :

MOCK_CONSTRUCTOR( class, arity, signature, identifier )
MOCK_CONSTRUCTOR_TPL( class, arity, signature, identifier ) // must be used if the signature uses a template parameter of the class

Example :

MOCK_CLASS( mock_class )
{
    MOCK_CONSTRUCTOR( mock_class, 2, ( int, const std::string& ), identifier )
};

Example :

template< typename T >
MOCK_CLASS( mock_class )
{
    MOCK_CONSTRUCTOR( mock_class, 2, ( int, const std::string& ), identifier )
    MOCK_CONSTRUCTOR_TPL( mock_class, 2, ( T, const std::string& ), identifier )
};

Destructor

Synopsis :

MOCK_DESTRUCTOR( class, identifier )

Example :

MOCK_CLASS( mock_class )
{
    MOCK_DESTRUCTOR( mock_class, destructor )
};

When mocking a destructor it is strongly suggested to manually #Verify the expectation at the end of the test, because the automatic verification will not be triggered if the mock object is not destroyed.

Conversion operator

Synopsis :

MOCK_CONVERSION_OPERATOR( type, identifier )
MOCK_CONST_CONVERSION_OPERATOR( type, identifier )
MOCK_NON_CONST_CONVERSION_OPERATOR( type, identifier )

Example :

MOCK_CLASS( mock_class )
{
    MOCK_CONVERSION_OPERATOR( int, conversion_to_int )
    MOCK_CONST_CONVERSION_OPERATOR( const std::string&, conversion_to_string )
};

Functor

Synopsis :

MOCK_FUNCTOR( functor, signature );

Example :

BOOST_AUTO_TEST_CASE( demonstrates_instantiating_a_mock_functor )
{
   MOCK_FUNCTOR( f, void( int ) );
}

Function

Synopsis :

MOCK_FUNCTION( function, arity, signature, identifier )

Example :

MOCK_FUNCTION( mock_function, 1, float( int ), mock_function )

A static object is used behind the scene in order to keep track of the expectations of a mock free function, therefore to ensure all tests run in isolation it is strongly suggested to :

  • manually verify the free function at the end of a test using mock::verify() or mock::verify( identifier )
  • manually reset the free function at the end of a test using mock::reset() or mock::reset( identifier )

Configuration

Setting up an invocation

Synopsis :

MOCK_EXPECT( <<scope>> identifier );
MOCK_EXPECT( <<scope>> identifier ).once();
MOCK_EXPECT( <<scope>> identifier ).never();
MOCK_EXPECT( <<scope>> identifier ).exactly( count );
MOCK_EXPECT( <<scope>> identifier ).at_least( min );
MOCK_EXPECT( <<scope>> identifier ).at_most( max );
MOCK_EXPECT( <<scope>> identifier ).between( min, max ); // throws std::invalid_argument if 'min' > 'max'

Example :

MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT( method, 2, void( int, const std::string& ), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_setting_up_invocations_on_a_mock_method )
{
   mock_class c;
   MOCK_EXPECT( c.method ).once(); // can only be called once
   MOCK_EXPECT( c.method );        // can be called an unlimited number of times
}

Example :

BOOST_AUTO_TEST_CASE( demonstrates_setting_up_an_invocation_on_a_mock_functor )
{
   MOCK_FUNCTOR( f, void( int, const std::string& ) );
   MOCK_EXPECT( f );
}

Example :

MOCK_FUNCTION( free_function, 1, float( int ), free_function )

BOOST_AUTO_TEST_CASE( demonstrates_setting_up_an_invocation_on_a_mock_function )
{
    MOCK_EXPECT( free_function ).once();
}

Example :

MOCK_CLASS( mock_class )
{
    MOCK_STATIC_METHOD( method, 1, float( int ), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_setting_up_an_invocation_on_a_mock_static_method )
{
    MOCK_EXPECT( mock_class::method ).once();
}

Adding constraints

Synopsis :

MOCK_EXPECT( <<scope>> identifier ).with( parameter1_constraint, parameter2_constraint, ... );

Constraints :

Expression Effect Description Explanation
mock::any true does not perform any verification
expected expected( actual ) calls expected as a functor returning a bool, throws std::invalid_argument if !expected when passing expected directly as a shortcut :
  • mock::call is implied for :
    • a function
    • a function pointer
    • an instance of a type with a result_type member typedef (support for standard library, Boost.Bind, Boost.Function functors)
    • an instance of a type with a sig member (support for Boost.Lambda functors)
    • an instance of a type with a result member (support for Boost.Phoenix functors)
  • mock::equal is implied for anything else
actual == expected compares actual to expected using operator ==
mock::equal( expected ) actual == expected compares actual to expected using operator ==
mock::less( expected ) actual < expected compares actual to expected using operator <
mock::greater( expected ) actual > expected compares actual to expected using operator >
mock::less_equal( expected ) actual <= expected compares actual to expected using operator <=
mock::greater_equal( expected ) actual >= expected compares actual to expected using operator >=
mock::call( expected ) expected( actual ) calls expected as a functor returning a bool and accepting actual as parameter
mock::same( expected ) &actual == &expected compares actual to expected by comparing their pointers
mock::assign( expected ) actual = expected, true assigns expected to actual using operator = The switch to one form or another is made depending on whichever is the most relevant based on types involved.

These constraints have side effects and they may modify data in unexpected ways. For instance they may be called again after their expectations have already been exhausted because of the way the expectation selection algorithm works, see #Expectation_selection_algorithm. Therefore it is probably a good idea to use an action instead, see #Configuring actions

*actual = expected, true assigns expected to *actual using operator =
mock::retrieve( expected ) expected = actual, true retrieves actual into expected using operator =
expected = &actual, true retrieves actual address into expected using operator =
mock::contain( expected ) actual.find( expected_ ) != std::string::npos checks whether expected is contained in the std::string actual
mock::affirm actual uses actual as a boolean
mock::negate  !actual negates actual using operator !
mock::evaluate actual() evaluates actual as a functor returning a bool and taking no argument

Note : all constraints taking a parameter support the use of boost::ref and boost::cref in order to delay its initialization.

Example :

MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT( method, 2, void( int, const std::string& ), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_adding_builtin_constraints )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( mock::equal( 3 ), mock::equal( "some string" ) );
   MOCK_EXPECT( c.method ).with( 3, "some string" );                               // equivalent to the previous one using shortcuts
}

Example using a function pointer :

bool custom_constraint( int actual )
{
   return actual == 42;
}

BOOST_AUTO_TEST_CASE( demonstrates_adding_a_custom_constraint_with_a_free_function )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( &custom_constraint );
}

Example using a standard library functor :

bool custom_constraint( int expected, int actual )
{
   return expected == actual;
}

BOOST_AUTO_TEST_CASE( demonstrates_adding_a_custom_constraint_with_a_standard_library_functor )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( std::bind1st( std::ptr_fun( &custom_constraint ), 42 ) ); // std::ptr_fun creates an std::unary_function
}

Example using Boost.Bind :

bool custom_constraint( int expected, int actual )
{
   return expected == actual;
}

BOOST_AUTO_TEST_CASE( demonstrates_adding_a_custom_constraint_with_boost_bind )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( boost::bind( &custom_constraint, 42, _1 ) ) );
}

Example using Boost.Lambda :

BOOST_AUTO_TEST_CASE( demonstrates_adding_a_custom_constraint_with_boost_lambda )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( boost::lambda::_1 == 42 );
}

Example using Boost.Phoenix :

BOOST_AUTO_TEST_CASE( demonstrates_adding_a_custom_constraint_with_boost_phoenix )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( boost::phoenix::arg_names::arg1 == 42 );
   MOCK_EXPECT( c.method ).with( boost::phoenix::arg_names::_1 == 42 );
}

Example combining constraints using &&, || and ! :

BOOST_AUTO_TEST_CASE( demonstrates_combining_constraints )
{
   mock_class c;
   MOCK_EXPECT( c.method ).with( mock::less( 4 ) && mock::greater( 2 ), ! mock::equal( "" ) );
}

Enforcing expectations order

Synopsis :

mock::sequence s;
MOCK_EXPECT( <<scope>> identifier1 ).in( s );
MOCK_EXPECT( <<scope>> identifier2 ).in( s );

Example :

MOCK_CLASS( mock_class1 )
{
    MOCK_METHOD_EXT( method1, 0, void(), method1 )
};
MOCK_CLASS( mock_class2 )
{
    MOCK_METHOD_EXT( method2, 0, void(), method2 )
};

BOOST_AUTO_TEST_CASE( demonstrates_enforcing_expectations_order )
{
   mock_class1 c1;
   mock_class2 c2;
   mock::sequence s;
   MOCK_EXPECT( c1.method1 ).in( s );
   MOCK_EXPECT( c2.method2 ).in( s );
}

Example of setting several sequences by chaining calls to in :

BOOST_AUTO_TEST_CASE( demonstrates_enforcing_several_expectations_orders )
{
   mock_class1 c1;
   mock::sequence s1, s2;
   MOCK_EXPECT( c1.method1 ).in( s1 ).in( s2 );
}

Configuring actions

Synopsis :

MOCK_EXPECT( <<scope>> identifier ).returns( value );
MOCK_EXPECT( <<scope>> identifier ).throws( exception );
MOCK_EXPECT( <<scope>> identifier ).calls( functor );    // throws std::invalid_argument if !functor

Example :

MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT( method, 0, int( int ), method )
};

int function( int i )
{
    return i;
}

BOOST_AUTO_TEST_CASE( demonstrates_configuring_actions )
{
   mock_class c;
   MOCK_EXPECT( c.method ).returns( 42 );
   MOCK_EXPECT( c.method ).throws( std::runtime_error( "error !" ) );
   MOCK_EXPECT( c.method ).calls( &function );                        // forwards 'method' parameter to 'function'
   MOCK_EXPECT( c.method ).calls( boost::bind( &function, 42 ) );     // drops 'method' parameter and binds 42 as parameter to 'function'
}

Summary

Synopsis :

MOCK_EXPECT( <<scope>> identifier ).invocation( arguments ).with( constraints ).in( sequence ).action( value );

Example :

MOCK_CLASS( mock_class )
{
   MOCK_METHOD_EXT( method, 0, int( int ), method )
   MOCK_METHOD_EXT( method, 0, void( const std::string&, float ), method2 )
};

BOOST_AUTO_TEST_CASE( demonstrates_configuring_mock_objects )
{
   mock_class c;
   mock::sequence s;
   MOCK_EXPECT( c.method ).once().with( 0 ).in( s ).returns( 42 );
   MOCK_EXPECT( c.method2 ).never().with( "ok", mock::any );
   MOCK_EXPECT( c.method2 ).at_least( 2 ).in( s ).throws( std::runtime_error() );
}

Verification

Synopsis :

MOCK_VERIFY( <<scope>> identifier );
mock::verify( object );              // verifies all expectations set for all methods of 'object' which must be an instance of a class created using MOCK_CLASS or MOCK_BASE_CLASS, or inherit mock::object
mock::verify();                      // verifies all existing mock objects, functions and functors

These calls all return a boolean indicating whether the verification was successful or not, however usually simply calling them is enough because a failing verification delegates to the #Error_policy in order to log a message.

Note that each mock object verifies itself automatically upon destruction, which is usually sufficient for the most common use cases.
However using mock functions or mock static methods means the associated underlying objects will not be destroyed before exiting the test application, thus it is strongly suggested to verify (and possibly #Reset) them at the end of each test case (for instance using a fixture) to ensure that each test runs in isolation.

Example :

MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT( method, 0, void(), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_verifying_a_mock_method )
{
   mock_class c;
   MOCK_VERIFY( c.method ); // logs an error and returns false if not all expectations are met
   mock::verify( c );       // verifies all expectations set for all methods of 'c'
   mock::verify();          // verifies all existing mock objects, functions and functors
}

Example :

BOOST_AUTO_TEST_CASE( demonstrates_verifying_a_mock_functor )
{
   MOCK_FUNCTOR( f, void( int ) );
   MOCK_VERIFY( f );               // logs an error and returns false if not all expectations are met
   mock::verify( f );              // behaves the same as MOCK_VERIFY
   mock::verify();                 // verifies all existing mock objects, functions and functors
}

Example :

MOCK_FUNCTION( f, 1, void( int ), f )

BOOST_AUTO_TEST_CASE( demonstrates_verifying_a_mock_function )
{
   MOCK_VERIFY( f ); // logs an error and returns false if not all expectations are met
   mock::verify();   // verifies all existing mock objects, functions and functors
}

Example :

MOCK_CLASS( mock_class )
{
    MOCK_STATIC_METHOD( method, 0, void(), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_verifying_a_static_mock_method )
{
   mock_class c;
   MOCK_VERIFY( c::method ); // logs an error and returns false if not all expectations are met
   mock::verify();           // verifies all existing mock objects, functions and functors
}

Reset

Synopsis :

MOCK_RESET( <<scope>> identifier );
mock::reset( object );              // resets all expectations set for all methods of 'object' which must be an instance of a class created using MOCK_CLASS or MOCK_BASE_CLASS, or inherit mock::object
mock::reset();                      // resets all existing mock objects, functions and functors

Example :

MOCK_CLASS( mock_class )
{
    MOCK_METHOD_EXT( method, 0, void(), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_resetting_a_mock_method )
{
   mock_class c;
   MOCK_RESET( c.method ); // resets all expectations set for 'c.method'
   mock::reset( c );       // resets all expectations set on 'c'
   mock::reset();          // resets all existing mock objects, functions and functors
}

Example :

BOOST_AUTO_TEST_CASE( demonstrates_resetting_a_mock_functor )
{
   MOCK_FUNCTOR( f, void( int ) );
   MOCK_RESET( f );                // resets all expectations set for 'f'
   mock::reset();                  // resets all existing mock objects, functions and functors
}

Example :

MOCK_FUNCTION( f, 1, void( int ), f )

BOOST_AUTO_TEST_CASE( demonstrates_resetting_a_mock_function )
{
   MOCK_RESET( f ); // resets all expectations set for 'f'
   mock::reset();   // resets all existing mock objects, functions and functors
}

Example :

MOCK_CLASS( mock_class )
{
    MOCK_STATIC_METHOD( method, 0, void(), method )
};

BOOST_AUTO_TEST_CASE( demonstrates_resetting_a_static_mock_method )
{
   mock_class c;
   MOCK_RESET( c::method ); // resets all expectations set for 'c::method'
   mock::reset();           // resets all existing mock objects, functions and functors
}

Rationale

This section explains some of the design and implementation choices.

General design forces

The general idea behind the library design is to be able to write tests quickly and easily as well as to get the best possible diagnostic upon error (both compile and runtime errors).
The chainable syntax has been chosen in order to be as intuitive as possible, with the most simple form covering the most general use cases.
Several design choices follow the same motivation :

  • expectations are automatically verified upon destruction
  • both const and non-const versions of a method are mocked by default
  • the shortcuts for adding constraints cover 95% of the use cases
  • non-serializable types do not yield compilation errors but are logged as '?' by default
  • etc..

At the same time customizing any aspect of the library should require minimum effort, for instance :

  • custom constraints can be any functors, including free functions
  • customizing the logging of a type is done by defining a serialization operator
  • etc..

Exceptions thrown should not extend std::exception

By design the exceptions thrown upon error should not inherit from std::exception, for instance consider the following test case based on the example from the #Motivation section :

BOOST_AUTO_TEST_CASE( overflow_throws )
{
   mock_view v;
   calculator c( v );
   BOOST_CHECK_THROW( c.add( std::numeric_limits< int >::max(), 1 ), std::exception );
}

Any call to 'v' will be unexpected and yield an exception, which if it were an std::exception would erroneously make the test succeed whereas it is supposed to pass only if the operation overflows (thus not triggering 'v').

The library interface is based on many macros

Despite being often considered harmful they also provide a number of advantages :

  • they pack a lot of code and hide implementation details (MOCK_BASE_CLASS, MOCK_METHOD)
  • they make the interface homogeneous (MOCK_FUNCTOR, MOCK_CLASS)
  • line number and file can be added for logging purposes (MOCK_EXPECT)

Limitations

This section lists the library known limitations.

No support for unicode logging

There is no support for unicode logging mainly because Boost.Test does not support it either.

Litteral 0 cannot be used as null pointer in constraints

The following code does not compile :

class base
{
public:
    virtual void method( int i* ) = 0;
};

MOCK_BASE_CLASS( mock_base, base )
{
    MOCK_METHOD( method, 1 )
};
mock_base m;
MOCK_EXPECT( m.method ).with( mock::equal( 0 ) ); // this fails
MOCK_EXPECT( m.method ).with( 0 );                // this fails too !

This is due to the fact that the library uses templates pretty heavily, and the litteral 0 is considered as an int when instantiating a template function.

A workaround is :

MOCK_EXPECT( m.method ).with( mock::equal< int* >( 0 ) ); // this compiles

However a somewhat better solution would be :

MOCK_EXPECT( m.method ).with( mock::negate );

Template methods cannot be mocked

Given the following client code :

class concept
{
public:
    template< typename T >
    void method( T t )
    {}
};

template< typename T >
class client
{
public:
    client( T t )             // T is supposed to model the previous concept
    {
        t.method( 42 );
        t.method( "string" );
    }
};

Writing a mock object modeling 'concept' requires to list all the possible versions of 'method' :

MOCK_CLASS( mock_concept )
{
   MOCK_METHOD_EXT( method, 1, void( int ), method_int )
   MOCK_METHOD_EXT( method, 1, void( const char* ), method_string )
};

While still somewhat possible, mocking a template method is indeed a bit cumbersome.

A private pure virtual method cannot be mocked using MOCK_METHOD

The following code does not compile :

class base
{
private:
    virtual void method() = 0;
};
MOCK_BASE_CLASS( mock_base, base )
{
    MOCK_METHOD( method, 0 )       // this fails to compile because 'method' is not visible
};

The workaround would be to use MOCK_METHOD_EXT :

MOCK_BASE_CLASS( mock_base, base )
{
    MOCK_METHOD_EXT( method, 0, void(), method )
};

Commas are not allowed in templates in MOCK_BASE_CLASS

The following code does not compile :

template< typename T1, typename T2 >
struct my_base_class
{};
MOCK_BASE_CLASS( my_mock, my_base_class< int, int > )
{};

The problem is that the pre-processor believes the macro to be called with 3 arguments because of the comma between the types used to instantiate the template.

One workaround is :

typedef my_base_class< int, int > my_base_type;

MOCK_BASE_CLASS( my_mock, my_base_type )
{};

Of course this is not always possible, as in :

template< typename T1, typename T2 >
MOCK_BASE_CLASS( my_mock, my_base_type< T1, T2 > )
{};

Another workaround would make use of Boost.PP :

template< typename T1, typename T2 >
MOCK_BASE_CLASS( my_mock, my_base_type< T1 BOOST_PP_COMMA() T2 > )
{};

Actually BOOST_PP_COMMA implementation is quite trivial, being only :

#define BOOST_PP_COMMA() ,

Finally another workaround but be to not use the macro at all :

template< typename T1, typename T2 >
struct my_mock : my_base_type< T1, T2 >, mock::object
{};

The extra features provided by MOCK_BASE_CLASS are not usable with template base classes anyway because there is no equivalent to MOCK_METHOD for them.

Warning C4505: '...' : unreferenced local function has been removed

Example :

warning C4505: 'base::[thunk]: __thiscall base::`vcall'{0,{flat}}' } : unreferenced local function has been removed

This seems to be a random bug with some versions of the Microsoft Visual Studio compiler, see https://connect.microsoft.com/VisualStudio/feedback/ViewFeedback.aspx?FeedbackID=324427

The only known workaround is to disable the warning with a pragma :

#pragma warning( disable: 4505 )

Warning C4301: '...': overriding virtual function only differs from '...' by const/volatile qualifier

Example :

warning C4301: '`anonymous-namespace'::base::method': overriding virtual function only differs from '`anonymous-namespace'::base::method' by const/volatile qualifier

The following code produces this warning with some versions of the Microsoft Visual Studio compiler :

class base
{
public:
    virtual void method( const int ) = 0;
};

MOCK_BASE_CLASS( mock_base, base )
{
    MOCK_METHOD( method, 1 )                                // this produces the warning
    MOCK_METHOD_EXT( method, 1, void( const int ), method ) // this produces the warning too !
};

The problem is that the 'const' is actually not part of the function signature and therefore cannot be introspected.

The first workaround would be to remove the 'const' all together.
This is more sensible than it first sounds, after all the 'const' is useless in this situation, indeed :

class derived : public base
{
public:
    virtual void method( const int );
};

void derived::method( int )           // this compiles, links and is valid C++
{}

Otherwise another workaround would be to provide a proxy method :

MOCK_BASE_CLASS( mock_base, base )
{
    void method( const int i )
    {
        method_stub( i );
    }
    MOCK_METHOD_EXT( method_stub, 1, void( int ), method )
};

warning C4267: 'argument' : conversion from 'size_t' to 'unsigned int', possible loss of data

Compiling under Microsoft Visual Studio with the /Wp64 flag produces this warning at various locations in the library code.
This is actually a bug in the compiler, for more information see incorrect-warning-c4267.

Tips

This section highlights not-so-obvious features of the library gathered from real use cases.

Waiting for an asynchronous call

Problem :

namespace
{
    class base_class
    {
    public:
        virtual void method() = 0;
    };

    class my_class
    {
    public:
        explicit my_class( base_class& );

        void flush(); // repetitively calling this method will in turn call base_class::method at some point
    };
}

Solution :

#include <turtle/mock.hpp>
#include <boost/lambda/lambda.hpp>
#include <boost/thread.hpp>

namespace
{
    template< typename F >
    void wait( bool& condition, F flush, int timeout = 100, int sleep = 100 )
    {
        while( !condition && timeout > 0 )
        {
            --timeout;
            if( sleep > 0 )
                boost::this_thread::sleep( boost::posix_time::milliseconds( sleep ) );
            flush();
        }
    }
    MOCK_BASE_CLASS( mock_base_class, base_class )
    {
        MOCK_METHOD( method, 0 )
    };
}

BOOST_AUTO_TEST_CASE( method_is_called )
{
    mock_base_class mock;
    my_class c( mock );
    bool done = false;
    MOCK_EXPECT( mock.method ).once().calls( boost::lambda::var( done ) = true );
    wait( done, boost::bind( &my_class::flush, &c ) );
}

Retrieving an argument to use in a later constraint

Problem :

namespace
{
    class base_class
    {
    public:
        virtual void method( int value ) = 0;
    };

    class my_class
    {
    public:
        explicit my_class( base_class& );

        void process(); // the processing will call 'method' two times with the same value, but we don't know what value beforehand
    };
}

Solution :

#include <turtle/mock.hpp>

namespace
{
    MOCK_BASE_CLASS( mock_base_class, base_class )
    {
        MOCK_METHOD( method, 1 )
    };
}

BOOST_AUTO_TEST_CASE( method_is_called_two_times_with_the_same_value )
{
    mock_base_class mock;
    my_class c( mock );
    int value;
    MOCK_EXPECT( mock.method ).once().with( mock::retrieve( value ) );
    MOCK_EXPECT( mock.method ).once().with( boost::cref( value ) );
    c.process();
}

Comparing strings properly

Problem :

#include <turtle/mock.hpp>

namespace
{
    MOCK_CLASS( mock_class )
    {
        MOCK_METHOD_EXT( method, 1, void( const char* ), method )
    };
}

BOOST_AUTO_TEST_CASE( comparing_char_pointers_does_not_compare_strings )
{
    mock_class mock;
    MOCK_EXPECT( mock.method ).with( "some string" ); // this compiles and runs, but the expectation fails because pointers and not strings are being compared
    mock.method( "some string" );
}

Solution :

BOOST_AUTO_TEST_CASE( comparing_char_pointers_does_not_compare_strings )
{
    mock_class mock;
    MOCK_EXPECT( mock.method ).with( std::string( "some string" ) ); // this makes the test pass as the expectation now compares strings
    mock.method( "some string" );
}

Invoking a functor received as parameter

#include <turtle/mock.hpp>
#include <boost/function.hpp>
#include <boost/bind/apply.hpp>
#include <boost/bind.hpp>

namespace
{
    MOCK_CLASS( mock_class )
    {
        MOCK_METHOD_EXT( method, 1, void( boost::function< void( int ) > ), method )
    };
}

BOOST_AUTO_TEST_CASE( how_to_invoke_a_functor_passed_as_parameter_of_a_mock_method )
{
    mock_class mock;
    MOCK_EXPECT( mock.method ).calls( boost::bind( boost::apply< void >(), _1, 42 ) ); // whenever 'method' is called, invoke the functor with 42
    MOCK_FUNCTOR( f, void( int ) );                                                    // create a mock functor to verify this
    MOCK_EXPECT( f ).once().with( 42 );                                                // expect it to be called with 42
    mock.method( f );                                                                  // call 'method' with the mock functor
}                                                                                      // et voilĂ  !

Acknowledgments

Many thanks to Adrien Gervaise and Silvin Lubecki !

References

Personal tools