Ger*_*ago 9 c++ metaprogramming type-traits template-specialization c++14
我正在编写一些程序来通过代码生成自动调用一些API.
在某些情况下,我需要从一个类型转换Source为一个类型Target,但这些类型装饰有指针,const等.所以我需要做的是删除所有装饰,如指针,常量,数组等,得到普通类型将其映射到另一种类型,然后将装饰应用回新类型.
该实现有许多模板专业化.代码后的问题.我不能使用constexpr元编程,因为我需要使用VS2013.
template <class T>
struct TypeIs {
using type = T;
};
template <class T>
struct GetPlainType : TypeIs<typename std::decay<T>::type> {};
template <class T>
struct GetPlainType<T&> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T const &> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T &&> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T const &&> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T*> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T const *> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T const> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T[]> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
struct GetPlainType<T const[]> : TypeIs<typename GetPlainType<T>::type> {};
template <class T, std::size_t I>
struct GetPlainType<T[I]> : TypeIs<typename GetPlainType<T>::type> {};
template <class T, std::size_t I>
struct GetPlainType<T const [I]> : TypeIs<typename GetPlainType<T>::type> {};
template <class T>
using GetPlainType_t = typename GetPlainType<T>::type;
template <class Decorated, class Plain>
struct CopyDecorations : TypeIs<Plain> {};
template <class T, class Plain>
struct CopyDecorations<T const, Plain> :
TypeIs<typename CopyDecorations<T, Plain const>::type> {};
template <class T, class Plain>
struct CopyDecorations<T *, Plain> :
TypeIs<typename CopyDecorations<T, Plain *>::type> {};
template <class T, class Plain>
struct CopyDecorations<T const *, Plain> :
TypeIs<typename CopyDecorations<T, Plain const *>::type> {};
template <class T, class Plain>
struct CopyDecorations<T &, Plain> :
TypeIs<typename CopyDecorations<T, Plain &>::type> {};
template <class T, class Plain>
struct CopyDecorations<T const &, Plain> :
TypeIs<typename CopyDecorations<T, Plain const &>::type> {};
template <class T, class Plain>
struct CopyDecorations<T &&, Plain> :
TypeIs<typename CopyDecorations<T, Plain &&>::type> {};
template <class T, class Plain>
struct CopyDecorations<T const &&, Plain> :
TypeIs<typename CopyDecorations<T, Plain const &&>::type> {};
template <class T, class Plain>
struct CopyDecorations<T[], Plain> :
TypeIs<typename CopyDecorations<T, Plain[]>::type> {};
template <class T, class Plain>
struct CopyDecorations<T const [], Plain> :
TypeIs<typename CopyDecorations<T, Plain const []>::type> {};
template <class T, class Plain, std::size_t I>
struct CopyDecorations<T [I], Plain> :
TypeIs<typename CopyDecorations<T, Plain[I]>::type> {};
template <class T, class Plain, std::size_t I>
struct CopyDecorations<T const [I], Plain> :
TypeIs<typename CopyDecorations<T, Plain const [I]>::type> {};
template <class Decorated, class Plain>
using CopyDecorations_t = typename CopyDecorations<Decorated, Plain>::type;
int main()
{
static_assert(std::is_same<GetPlainType_t<int>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int const>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int *>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int **>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int * &>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int ** &>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int const * []>, int>{}, "");
static_assert(std::is_same<GetPlainType_t<int const **[][3][5]>, int>{}, "");
static_assert(std::is_same<CopyDecorations_t<int, double>, double>{}, "");
static_assert(std::is_same<CopyDecorations_t<int const, double>, double const>{}, "");
static_assert(std::is_same<CopyDecorations_t<int *, double>, double *>{}, "");
static_assert(std::is_same<CopyDecorations_t<int **, double>, double **>{}, "");
static_assert(std::is_same<CopyDecorations_t<int[], double>, double[]>{}, "");
static_assert(std::is_same<CopyDecorations_t<int[3], double>, double[3]>{}, "");
//******************THE TESTS BELOW DO NOT WORK
//static_assert(std::is_same<CopyDecorations_t<int[][3], double>, double[][3]>{}, "");
//static_assert(std::is_same<CopyDecorations_t<int * &, double>, double * &>{}, "");
// static_assert
// (
//std::is_same<CopyDecorations_t<int const * [], double>,
// double const * []>{}, "");
// static_assert
// (std::is_same<CopyDecorations_t<int const **[][3][5], double>,
// double const **[][3][5]>{}, "");
}
Run Code Online (Sandbox Code Playgroud)
问题:
main功能),我该如何解决?我发现这个问题是关于SO的C++元编程最有趣的问题之一.
我很乐意找到一个合适的解决方案.谢谢.:-)
它遵循一个最小的工作示例.
它并不完整,但它提供了一种可能的方法来实现它.
该函数f(好的,您可以在代码中选择更好的名称)接受两个模板参数:要清理的类型和要装饰的类型.
它返回一个模板类型(types),它引入了两个使用声明,basic并且decorated第一个模板参数被清理为basic,第二个模板参数被装饰为decorated.
它一次完成(清理和装饰).您仍然可以只使用第一个参数,在这种情况下decorated默认为装饰char类型.
这是完整的代码:
#include<type_traits>
#include<cstddef>
static constexpr std::size_t N = 42;
template<std::size_t N>
struct choice: choice<N-1> {};
template<>
struct choice<0> {};
template<typename T, typename U>
struct types {
using basic = T;
using decorated = U;
};
template<typename T, typename U>
constexpr auto
f(choice<0>) { return types<T, U>{}; }
template<typename T, typename U,
typename = std::enable_if_t<std::is_pointer<T>::value>>
constexpr auto f(choice<1>) {
auto t = f<std::remove_pointer_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_pointer_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_lvalue_reference<T>::value>>
constexpr auto f(choice<2>) {
auto t = f<std::remove_reference_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_lvalue_reference_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_rvalue_reference<T>::value>>
constexpr auto f(choice<3>) {
auto t = f<std::remove_reference_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_rvalue_reference_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_array<T>::value>>
constexpr auto f(choice<4>) {
auto t = f<std::remove_extent_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::conditional_t<(0==std::extent<T>::value), D[], D[std::extent<T>::value]>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_const<T>::value>>
constexpr auto f(choice<5>) {
auto t = f<std::remove_const_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_const_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_volatile<T>::value>>
constexpr auto f(choice<6>) {
auto t = f<std::remove_volatile_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_volatile_t<D>>{};
}
template<typename T, typename U = char>
constexpr auto f() {
return f<T, U>(choice<N>{});
}
int main() {
// something complex to show that it seems to work
static_assert(std::is_same<
decltype(f<const int ** const &&, char>()),
types<int, const char ** const &&>
>::value, "!");
// some of the OP's examples (the most interesting)
static_assert(std::is_same<decltype(f<int, int>()), types<int, int>>::value, "!");
static_assert(std::is_same<decltype(f<int const, int>()), types<int, int const>>::value, "!");
static_assert(std::is_same<decltype(f<int *, int>()), types<int, int *>>::value, "!");
static_assert(std::is_same<decltype(f<int **, double>()), types<int, double **>>::value, "!");
static_assert(std::is_same<decltype(f<int *&, int>()), types<int, int *&>>::value, "!");
static_assert(std::is_same<decltype(f<int **&, float>()), types<int, float **&>>::value, "!");
static_assert(std::is_same<decltype(f<int [3], char>()), types<int, char [3]>>::value, "!");
static_assert(std::is_same<decltype(f<int [], int>()), types<int, int []>>::value, "!");
static_assert(std::is_same<decltype(f<int [][3], double>()), types<int, double [][3]>>::value, "!");
static_assert(std::is_same<decltype(f<int const **[][3][5], int>()), types<int, int const **[][3][5]>>::value, "!");
// of course, you don't need to provide the second type if you don't need it
// in this case, types::decorated is defaulted to a decorated char type
f<int const **[][3][5]>();
}
Run Code Online (Sandbox Code Playgroud)
constexpr由于s的原因,它不能在没有s的情况下编译static_assert,你可以自由地删除它们并在运行时使用该函数.
实际上,它可能转向一个无定义的解决方案,为声明提供正确的返回类型并使用一堆decltypes,但我怀疑它远不是可读的.
编辑
如OP所述,他不想(或至少,他不能使用)constexprs.
它遵循略有不同的解决方案,仍然基于前一个解决方案.
基本思想是f用作未评估的操作数decltype.
这是完整的代码:
#include<type_traits>
#include<cstddef>
static const std::size_t N = 42;
template<std::size_t N>
struct choice: choice<N-1> {};
template<>
struct choice<0> {};
template<typename T, typename U>
struct types {
using basic = T;
using decorated = U;
};
template<typename T, typename U>
auto f(choice<0>) { return types<T, U>{}; }
template<typename T, typename U,
typename = std::enable_if_t<std::is_pointer<T>::value>>
auto f(choice<1>) {
auto t = f<std::remove_pointer_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_pointer_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_lvalue_reference<T>::value>>
auto f(choice<2>) {
auto t = f<std::remove_reference_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_lvalue_reference_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_rvalue_reference<T>::value>>
auto f(choice<3>) {
auto t = f<std::remove_reference_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_rvalue_reference_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_array<T>::value>>
auto f(choice<4>) {
auto t = f<std::remove_extent_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::conditional_t<(0==std::extent<T>::value), D[], D[std::extent<T>::value]>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_const<T>::value>>
auto f(choice<5>) {
auto t = f<std::remove_const_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_const_t<D>>{};
}
template<typename T, typename U,
typename = std::enable_if_t<std::is_volatile<T>::value>>
auto f(choice<6>) {
auto t = f<std::remove_volatile_t<T>, U>(choice<N>{});
using B = typename decltype(t)::basic;
using D = typename decltype(t)::decorated;
return types<B, std::add_volatile_t<D>>{};
}
template<typename T, typename U>
auto f() {
return f<T, U>(choice<N>{});
}
template<typename T, typename U = char>
using my_type = decltype(f<T, U>());
template<typename T, typename U = char>
using my_type_basic_t = typename decltype(f<T, U>())::basic;
template<typename T, typename U = char>
using my_type_decorated_t = typename decltype(f<T, U>())::decorated;
int main() {
int i = 42;
my_type_decorated_t<char *, int> ptr = &i;
// of course, it can still be used in a constant expression if needed
// constexpr my_type_decorated_t<char *, int> ptr = nullptr;
}
Run Code Online (Sandbox Code Playgroud)
因此,您可以使用模式匹配并执行转录的一步的函数来完成此操作,如下所示:
template<class In, class Out>
struct types {
using type=types;
using in=In;
using out=Out;
};
// transcribe cv:
template<class In, class Out>
types<In, const Out> transcribe( types<const In, Out> ) { return {}; }
template<class In, class Out>
types<In, volatile Out> transcribe( types<volatile In, Out> ) { return {}; }
template<class In, class Out>
types<In, const volatile Out> transcribe( types<const volatile In, Out> ) { return {}; }
// references and pointers:
template<class In, class Out>
types<In, Out*> transcribe( types<In*, Out> ) { return {}; }
template<class In, class Out>
types<In, Out&> transcribe( types<In&, Out> ) { return {}; }
template<class In, class Out>
types<In, Out&&> transcribe( types<In&&, Out> ) { return {}; }
// arrays
template<class In, class Out>
types<In, Out[]> transcribe( types<In[], Out> ) { return {}; }
template<class In, class Out, std::size_t N>
types<In, Out[N]> transcribe( types<In[N], Out> ) { return {}; }
// return type of a function
template<class In, class...In_Args, class Out>
types<In, Out(In_Args...)> transcribe( types<In(In_Args...), Out> ) { return {}; }
// return type of a function
template<class In, class...In_Args, class Out>
types<In, Out(*)(In_Args...)> transcribe( types<In(*)(In_Args...), Out> ) { return {}; }
// default case
template<class X>
X transcribe( X ) { return {}; }
Run Code Online (Sandbox Code Playgroud)
超载规则做正确的事。唯一令人讨厌的就是这种情况const volatile。
一旦我们有了上述内容,我们就可以将其变成一个特质:
template<class In, class Out>
struct transcribe_one:
decltype(transcribe( types<In,Out>{} ))
{};
Run Code Online (Sandbox Code Playgroud)
容易地。请注意,transcribe永远不需要调用。
别名使其更易于使用:
template<class T>
using strip_one=typename transcribe_one<T, int>::in;
template<class T>
using can_strip=std::integral_constant<bool, !std::is_same<T, strip_one<T>>{}>;
template<class T, class U>
using typescribe_one=typename transcribe_one<T, U>::out;
Run Code Online (Sandbox Code Playgroud)
并表示“有什么要脱的吗?”。
这只会从左到右移动一种类型的装饰。要移动它们,我们只需这样做:
template<class In, class Out, class=void>
struct transcribe_all:types<In, Out> {};
template<class T>
using strip=typename transcribe_all<T, int>::in;
template<class T, class U>
using typescribe=typename transcribe_all<T, U>::out;
template<class In, class Out>
struct transcribe_all<In, Out, std::enable_if_t<
can_strip<In>{}
>> :
types<
strip< strip_one< In > >, // must strip on strip_one, trust me
typescribe_one<
In,
typescribe< strip_one<In>, Out >
>
>
{};
Run Code Online (Sandbox Code Playgroud)
当你无法剥离时,它什么也不做。
当您可以剥离时,它会剥离 In 类型中的一个,将剩余部分转录到 Out 上,然后对其结果进行一步转录。
这给你两个别名:
template<class T>
using strip=// ...
template<class T, class U>
using typescribe=// ...
Run Code Online (Sandbox Code Playgroud)
第一个采用类型T并将其剥离为底部的“原始”类型。
第二个采用类型T并将其所有装饰移动到U。
template<template<class...>class M, class U>
using under_map = typescribe< U, M<strip<U>> >;
Run Code Online (Sandbox Code Playgroud)
它会剥离类型装饰,应用M,然后重新应用它们。
请注意,最“用户友好”的工具是strip<T>和typescribe<In, Out>。第一个删除所有装饰器并获取“底层”类型。第二个将装饰器从一种类型复制到另一种类型。使用transcribe_allortranscribe_one或strip_one等可能会导致混乱,它们是实现细节。
只需将类型映射写为 atemplate并将其传递给under_map。
使用的唯一 C++14 功能是std::enable_if_t<?>,可以在 C++11 上替换为typename std::enable_if<?>::type。
MSVC 可能对这些decltype东西有问题,因为它在 SFINAE 中对 decltype 的支持很糟糕。
| 归档时间: |
|
| 查看次数: |
499 次 |
| 最近记录: |