Template metaprogramming

From Wikipedia, the free encyclopedia

Template metaprogramming is a metaprogramming technique in which templates are used by a compiler to generate temporary source code, which is merged by the compiler with the rest of the source code and then compiled. The output of these templates include compile-time constants, data structures, and complete functions. The use of templates can be thought of as compile-time execution. The technique is used by a number of languages, including C++, D, Eiffel, Haskell, ML and XL.

1 Components of template metaprogramming
2 Using template metaprogramming
3 Benefits and drawbacks of template metaprogramming
4 See also
5 References
6 External links

[edit] Components of template metaprogramming

The use of templates as a metaprogramming technique requires two distinct operations; a template must be defined, and a defined template must be instantiated. The template definition describes the generic form of the generated source code, and the instantiation causes a specific set of source code to be generated from the generic form in the template.

Template metaprogramming is generally Turing-complete, meaning that any computation expressible by a computer program can be computed, in some form, by a template metaprogram.

Templates are different from macros. A macro, which is also a compile-time language feature, generates code in-line using text manipulation and substitution. Macro systems often have limited compile-time process flow abilities and usually lack awareness of the semantics and type system of their companion language (an exception should be made with Lisp's macros, which are written in Lisp itself).

Template metaprograms have no mutable variables— that is, no variable can change value once it has been initialized, therefore template metaprogramming can be seen as a form of functional programming. In fact many template implementations only implement flow control through recursion, as seen in the example below.

[edit] Using template metaprogramming

Though the syntax of template metaprogramming is usually very different from the programming language it is used with it has practical uses. Some common reasons to use templates is to implement generic programming (avoiding sections of code which are similar except for some minor variations) or to perform automatic compile-time optimization such as doing something once at compile time rather than every time the program is run, for instance having the compiler unroll loops which would otherwise not be.

[edit] Compile-time class generation

What exactly "programming at compile-time" means can be illustrated with an example of a factorial function, which in non-templated C++ can be written using recursion as follows:

int factorial(int n) 
{
    if (n == 0)
       return 1;
    return n * factorial(n - 1);
}

void foo()
{
    int x = factorial(4); // == (4 * 3 * 2 * 1) == 24
    int y = factorial(0); // == 0! == 1
}

The code above will execute when the program is run to determine the factorial value of the literals 4 and 0.

Instead using template metaprogramming and template specialisation to provide the ending condition for the recursion, the factorials used in the program, ignoring any factorial not used, can be calculated at compile-time by

template <int N>
struct Factorial 
{
    enum { value = N * Factorial<N - 1>::value };
};

template <>
struct Factorial<0> 
{
    enum { value = 1 };
};

// Factorial<4>::value == 24
// Factorial<0>::value == 1
void foo()
{
    int x = Factorial<4>::value; // == 24
    int y = Factorial<0>::value; // == 1
}

The code above calculates the factorial value of the literals 4 and 0 at compile time and uses the result as if they were precalculated constants.

While the two versions are similar from the point of view of the program's functionality, the first example calculates the factorials at run time while the second calculates them at compile time. However to be able to use templates in this manner, the compiler must know the value of its parameters at compile time, which has the natural precondition that Factorial<X>::value can only be used if X is known at compile time. In other words, X must be a constant or the result of a call to a compiler function such as sizeof().

[edit] Compile-time code optimisation

The factorial example above is one example of compile-time code optimisation in that all factorials used by the program are pre-compiled and injected as numeric constants at compilation, saving both run-time overhead and memory footprint. It is, however, a relatively minor optimisation.

As another, more significant, example of compile-time loop-unrolling, templated metaprogramming can be used to create n-dimensional vector classes (where n is known at compile time) with the benefit over a more traditional n-dimensional vector is that the loops can be unrolled, resulting in very optimized code. As an example, consider the addition operator. An n-dimensional vector addition might be written as

template<int dimension>
Vector<dimension>& Vector<dimension>::operator+=(const Vector<dimension>& rhs) 
{
    for (int i = 0; i < dimension; ++i)
        value[i] += rhs.value[i];
    return *this;
}

When the compiler instantiates the templated function defined above, the following code will be produced:

template<>
Vector<2>& Vector<2>::operator+=(const Vector<2>& rhs) 
{
    value[0] += rhs.value[0];
    value[1] += rhs.value[1];
    return *this;
}

The compiler's optimizer is able to unroll the for loop because the template parameter dimension is a constant at compile time. For a real implementation of this technique, see this post on flipcode.com.

[edit] Static polymorphism

Polymorphism is a powerful common standard programming facility where derived objects can be used as instances of their base object but where the derived objects' methods will be invoked, as in this code

class Base
{
    public:
    virtual void method() { std::cout << "Base"; }
};

class Derived : public Base
{
    public:
    virtual void method() { std::cout << "Derived"; }
};

int main()
{
    Base *pBase = new Derived;
    pBase->method(); //outputs "Derived"
    return 0;
}

where all invocations of virtual methods will that of the most-derived class. This dynamically polymorphic behaviour is obtained by the creation of virtual look-up tables for classes with virtual methods, tables that are traversed at run-time to identify which method to be invoked. Thus run-time polymorphism necessary entails execution overhead.

However, in many cases the polymorphic behaviour needed is invariant and can be determined at compile-time. Then the Curiously Recurring Template Pattern can be used to achieve static polymorphism, which is an imitation of polymorphism in programming code but which is resolved at compile-time and thus does away with run-time virtual-table lookups. For example:

template <class Derived>
struct base
{
    void interface()
    {
         // ...
         static_cast<Derived*>(this)->implementation();
         // ...
    }
};

struct derived : base<derived>
{
     void implementation();
};

Here the base class template will take advantage of the fact that member function bodies are not instantiated until long after their declarations, and will use members of the derived class within its own member functions, via the use of a static_cast, thus at compilation generating an object composition with polymorphic characteristics. The CRTP is for instance pervasively used in the Boost iterator library[1].

Another similar use is the "Barton-Nackman trick", sometimes referred to as "restricted template expansion", where common functionality can be placed in a base class that is used not as a contract but as a necessary component to enforce conformant behaviour while minimising code redundancy.

[edit] Benefits and drawbacks of template metaprogramming

Compile-time versus execution-time tradeoff: Since all templated code is processed, evaluated and expanded at compile-time, compilation will take longer while the executable code may be more efficient. This overhead is generally small, but for large projects, or projects relying pervasively on templates, it may be significant.
Generic programming: Template metaprogramming allows the programmer to focus on architecture and delegate to the compiler the generation of any implementation required by client code. Thus, template metaprogramming can accomplish truly generic code, facilitating code minimization and better maintainability.
Readability: With respect to C++, the syntax and idioms of template metaprogramming are esoteric compared to conventional C++ programming, and advanced, or even most non-trivial, template metaprogramming can be very difficult to understand. Metaprograms can thus be difficult to maintain by programmers inexperienced in template metaprogramming (though this may vary with the language's implementation of template metaprogramming syntax).
Portability: With respect to C++, due to differences in compilers, code relying heavily on template metaprogramming (especially the newest forms of metaprogramming) might have portability issues.

[edit] See also

[edit] References

Ulrich W. Eisenecker: Generative Programming: Methods, Tools, and Applications, Addison-Wesley, ISBN 0-201-30977-7
Andrei Alexandrescu: Modern C++ Design: Generic Programming and Design Patterns Applied, Addison-Wesley, ISBN 3-8266-1347-3
David Abrahams, Aleksey Gurtovoy: C++ Template Metaprogramming: Concepts, Tools, and Techniques from Boost and Beyond, Addison-Wesley, ISBN 0-321-22725-5
David Vandervoorde, Nicolai M. Josuttis: C++ Templates: The Complete Guide, Addison-Wesley, ISBN 0-201-73484-2
Manuel Clavel: Reflection in Rewriting Logic: Metalogical Foundations and Metaprogramming Applications, ISBN 1-57586-238-7
What's Wrong with C++ Templates? by Jacob Matthews