我正在寻找std::tuple
的投影功能。即,函数接收整数列表作为模板参数,并返回仅具有这些索引值的元组。
例如,假设我有一个元组tuple<int,String,char,float> t
,那么调用project<1,3>(t)
会产生tuple<String,float>
类型的元组。
当然,关键似乎是递归模板。但是,我无法让它工作。我尝试通过声明一个结构来计算投影操作的返回类型。即使这样也失败了这是我试过的:
template<typename T, size_t... indexes>
class Projection{
};
// Specialization for no index => Empty tuple
template<>
template<typename... T>
class Projection<std::tuple<T...>>{
typedef std::tuple<T...> InputTuple;
typedef std::tuple<> Tuple;
};
// Specialization for one or more parameters
template<size_t Index1, size_t... others>
template<typename... T>
class Projection<std::tuple<T...>,Index1,others...>{
typedef std::tuple<T...> InputTuple;
// The type of a projection with one column less
typedef Projection<std::tuple<T...>,others...> TupleMinusOne;
// The result type is the catenation of TupleMinusOne plus the column projected in this step
typedef decltype(std::tuple_cat(std::make_tuple(std::get<Index1>(InputTuple())),typename TupleMinusOne::Tuple())) Tuple;
};
这个编译。带有空元组的基本情况也有效,即:
Projection<std::tuple<int,std::string>>::Tuple t;
导致t
为空元组。但是,递归情况不会编译:
Projection<std::tuple<int,std::string>,1>::Tuple t;
我收到以下错误:
Test.cpp:79:1: error: ‘Tuple’ is not a member of ‘Projection<std::tuple<int, float>, 1ul>’
所以似乎无法识别递归情况,但为什么呢?
答案 0 :(得分:4)
很少需要递归模板; pack expansion通常更清晰,更简单。在这种情况下,只需使用tuple_element
:
template<typename T, size_t... indexes>
class Projection{
public:
using Tuple = std::tuple<typename std::tuple_element<indexes, T>::type...>;
};
同样为project
:
template<size_t... indexes, typename T>
auto project(const T &t) -> typename Projection<T, indexes...>::Tuple {
return typename Projection<T, indexes...>::Tuple(std::get<indexes>(t)...);
}
答案 1 :(得分:1)
您可以使用以下内容:
#define Return(ret) decltype(ret) { return ret; }
template<std::size_t... Is, typename T>
auto project(const T& t) -> Return(std::make_tuple(std::get<Is>(t)...))