Metaprogramming

This page documents library components that provide metaprogramming sorts of functionality. For the most part they are useful for putting design by contract checks into code or doing various kinds of clever things with templates.

For example, you might have a templated function that is templated on a type T and you want to make sure that T is either a char or wchar_t type. You could place the following into your code and it would cause the compile to error out when T was set to something other than char or wchar_t.
COMPILE_TIME_ASSERT((is_same_type<T,char>::value || is_same_type<T,wchar_t>::value));

[top]

ASSERT_ARE_NOT_SAME_TYPE



This is a macro function for debugging. Its form is ASSERT_ARE_NOT_SAME_TYPE(type1, type2). If type1 and type2 are the same type then the compile will fail. This is sometimes useful in validating template arguments.


More Details...
#include <dlib/assert.h>
[top]

ASSERT_ARE_SAME_TYPE



This is a macro function for debugging. Its form is ASSERT_ARE_SAME_TYPE(type1, type2). If type1 and type2 are not the same type then the compile will fail. This is sometimes useful in validating template arguments.


More Details...
#include <dlib/assert.h>
[top]

assign_zero_if_built_in_scalar_type



This function assigns its argument the value of 0 if it is a built in scalar type according to the is_built_in_scalar_type template. If it isn't a built in scalar type then it does nothing.

This function is useful for suppressing compiler warnings about uninitialized types inside of templates that are designed to accept the built in types as well as user defined classes.


More Details...
#include <dlib/algs.h>
[top]

basic_type



This is a template that takes a type and strips off any const, volatile, or reference qualifiers and gives you back the basic underlying type.

For example, promote<const int&>::type == int


More Details...
#include <dlib/algs.h>
[top]

COMPILE_TIME_ASSERT



This is a macro function for debugging. Its form is COMPILE_TIME_ASSERT(condition that should be true). The condition must be a compile time constant and if it is false then the compile will fail.


More Details...
#include <dlib/assert.h>
[top]

DLIB_ASSERT



This is a macro function for debugging. Its form is

DLIB_ASSERT(condition that should be true, error message)
or you can omit the error message and call it like:
DLIB_ASSERT(condition that should be true)
If the condition is false DLIB_ASSERT throws an exception of type dlib::fatal_error with fatal_error::type == EBROKEN_ASSERT. An error message detailing the nature of the problem is stored in the member variable info which is of type std::string. Look in the following file for more details. The exception classes are defined here.

This macro is only enabled if _DEBUG, DEBUG or ENABLE_ASSERTS is defined. Also, if this macro is enabled then ENABLE_ASSERTS will be defined even if you didn't define it.

Note that when this macro fails and throws an exception it also calls the global C function dlib_assert_breakpoint(). This behavior makes it easy to set a debugging tool to break when DLIB_ASSERT fails by setting a breakpoint on dlib_assert_breakpoint().


More Details...
#include <dlib/assert.h>
[top]

DLIB_ASSERT_HAS_STANDARD_LAYOUT



This macro is meant to cause a compiler error if a type doesn't have a simple memory layout (like a C struct). In particular, types with simple layouts are ones which can be copied via memcpy().

This was called a POD type in C++03 and in C++0x we are looking to check if it is a "standard layout type". Once we can use C++0x we can change this macro to something that uses the std::is_standard_layout type_traits class. See: http://www2.research.att.com/~bs/C++0xFAQ.html#PODs
More Details...
#include <dlib/assert.h>
[top]

DLIB_CASSERT



This is a macro function that is identical to the DLIB_ASSERT macro except that it is always enabled. Even if _DEBUG, DEBUG and ENABLE_ASSERTS are not defined.

Note that when this macro fails and throws an exception it also calls the global C function dlib_assert_breakpoint(). This behavior makes it easy to set a debugging tool to break when DLIB_CASSERT fails by setting a breakpoint on dlib_assert_breakpoint().


More Details...
#include <dlib/assert.h>
[top]

DLIB_MAKE_HAS_MEMBER_FUNCTION_TEST



The DLIB_MAKE_HAS_MEMBER_FUNCTION_TEST() macro is used to define traits templates that tell you if a class has a certain member function. For example, to make a test to see if a class has a public method with the signature void print(int) you would say:

DLIB_MAKE_HAS_MEMBER_FUNCTION_TEST(has_print, void, print, (int))
Then you can check if a class, T, has this method by looking at the boolean value:
has_print<T>::value
which will be true if the member function is in the T class.
More Details...
#include <dlib/algs.h>
[top]

DLIB_STACK_TRACE



This is a preprocessor macro that allows you to tag a function so that dlib will keep track of it in a function call stack. That is, you will be able to see a stack trace by calling get_stack_trace if you put this macro at the top of your functions.

This macro is only enabled if DLIB_ENABLE_STACK_TRACE is defined. If it isn't defined then this macro doesn't do anything. Also note that when this macro is defined it will cause DLIB_ASSERT and DLIB_CASSERT to include a stack trace in their error messages.


More Details...
#include <dlib/assert.h>
[top]

DLIB_STACK_TRACE_NAMED



This is a preprocessor macro just like DLIB_STACK_TRACE except that it allows you to supply your own string to use as the function name in the stack trace instead of the one deduced by DLIB_STACK_TRACE.

This macro is only enabled if DLIB_ENABLE_STACK_TRACE is defined.


More Details...
#include <dlib/assert.h>
[top]

enable_if



This is a family of templates from the Boost C++ libraries that makes it somewhat easier to control template specialization. For the details see this page. Note that the header dlib/enable_if.h brings these templates into the dlib namespace.

#include <dlib/enable_if.h>
[top]

get_stack_trace



This function allows you to query the current stack trace.

This macro is only enabled if DLIB_ENABLE_STACK_TRACE is defined.


More Details...
#include <dlib/assert.h>
[top]

is_array



This is a template where is_array<T>::value == true when T is an array object.
More Details...
#include <dlib/is_kind.h>
[top]

is_array2d



This is a template where is_array2d<T>::value == true when T is an array2d object.
More Details...
#include <dlib/is_kind.h>
[top]

is_built_in_scalar_type



This is a template where is_built_in_scalar_type<T>::value == true when T is a built in scalar type such as int, char, float, etc.
More Details...
#include <dlib/algs.h>
[top]

is_complex



This is a template that can be used to determine if a type is a specialization of std::complex.
More Details...
#include <dlib/matrix.h>
[top]

is_config_reader



This is a template where is_config_reader<T>::value == true when T is a config_reader or config_reader_thread_safe object.
More Details...
#include <dlib/is_kind.h>
[top]

is_const_type



This is a template where is_const_type<T>::value == true when T is a const type and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_convertible



This is a template that can be used to determine if one type is convertible into another type.
More Details...
#include <dlib/algs.h>
[top]

is_directed_graph



This is a template where is_directed_graph<T>::value == true when T is a directed_graph object.
More Details...
#include <dlib/is_kind.h>
[top]

is_float_type



This is a template where is_float_type<T>::value == true when T is a floating point type (i.e. float, double, or long double) and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_function



This is a template where is_function<T>::value == true when T is a function type.
More Details...
#include <dlib/algs.h>
[top]

is_graph



This is a template where is_graph<T>::value == true when T is a graph object.
More Details...
#include <dlib/is_kind.h>
[top]

is_matrix



This is a template where is_matrix<T>::value == true when T is a matrix object or some kind of matrix expression.
More Details...
#include <dlib/is_kind.h>
[top]

is_pair



This is a template where is_pair<T>::value == true when T is a std::pair object.
More Details...
#include <dlib/is_kind.h>
[top]

is_pointer_type



This is a template where is_pointer_type<T>::value == true when T is a pointer type and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_rand



This is a template where is_rand<T>::value == true when T is a rand object.
More Details...
#include <dlib/is_kind.h>
[top]

is_reference_type



This is a template where is_reference_type<T>::value == true when T is a reference type and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_same_object



This is a templated function which checks if both of its arguments are actually references to the same object. It returns true if they are and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_same_type



This is a template where is_same_type<T,U>::value == true when T and U are the same type and false otherwise.
More Details...
#include <dlib/algs.h>
[top]

is_signed_type



This is a template where is_signed_type<T>::value == true when T is a signed scalar type and false when it is an unsigned scalar type.
More Details...
#include <dlib/algs.h>
[top]

is_std_vector



This is a template where is_std_vector<T>::value == true when T is a std_vector_c or std::vector object.
More Details...
#include <dlib/is_kind.h>
[top]

is_unsigned_type



This is a template where is_unsigned_type<T>::value == true when T is an unsigned scalar type and false when it is a signed scalar type.
More Details...
#include <dlib/algs.h>
[top]

noncopyable



This is a simple class that makes it easy to declare a non-copyable object. To use it to make your own class non-copyable just inherit from it.
More Details...
#include <dlib/noncopyable.h>
[top]

portability_macros



This file #defines various macros depending on the platform being compiled under. See the file itself for the specifics.
More Details...
#include <dlib/platform.h>
[top]

promote



This is a template that takes one of the built in scalar types and gives you another scalar type that should be big enough to hold sums of values from the original scalar type. The new scalar type will also always be signed.

For example, promote<uint16>::type == int32


More Details...
#include <dlib/algs.h>
[top]

static_switch



To use this template you give it some number of boolean expressions and it tells you which one of them is true. If more than one of them is true then it causes a compile time error. It is useful for cases where you want to specialize a template and you want to specialize it not by the type of object it gets per say but instead according to the values of some type traits associated with the various template arguments. A simple example of this can be seen in the assign_pixel's implementation which can be found at the bottom of the dlib/pixel.h file.
More Details...
#include <dlib/algs.h>
[top]

tabs



This is a template to compute the absolute value a number at compile time.
More Details...
#include <dlib/algs.h>
[top]

tmax



This is a template to compute the max of two values at compile time.
More Details...
#include <dlib/algs.h>
[top]

tmin



This is a template to compute the min of two values at compile time.
More Details...
#include <dlib/algs.h>
[top]

unsigned_type



This is a template that allows you to obtain the unsigned version of any integral type. For example, unsigned_type<signed short>::type == unsigned short.
More Details...
#include <dlib/uintn.h>
[top]

wrap_function



This is a template that allows you to turn a global function into a function object. See the specs for more details.
More Details...
#include <dlib/algs.h>
[top]

_dT



This is a macro function for converting a string/character literal to either a char or wchar_t literal. Its form is _dT(target character type,string or character literal)
More Details...
#include <dlib/algs.h>