I want to implement a range constructor for a certain object, but I want to restrict it to only accept two input iterators.
I've tried to compile this code with gcc 7.1.0.
File test.cpp
#include <vector>
#include <type_traits>
#include <typeinfo>
template <typename Iterator>
using traits = typename std::iterator_traits<Iterator>::iterator_category;
template <typename T>
class A{
private:
std::vector<T> v;
public:
template <typename InputIterator,
typename = std::enable_if_t<
typeid(traits<InputIterator>) ==
typeid(std::input_iterator_tag)>
>
A(InputIterator first, InputIterator last) : v(first, last) {}
};
int main(){
std::vector<double> v = {1, 2, 3, 4, 5};
A<double> a(v.begin(), v.end());
}
I get this compilation error with g++ test.cpp -o test
:
test.cpp: In function ‘int main()’:
test.cpp:27:34: error: no matching function for call to ‘A<double>::A(std::vector<double>::iterator, std::vector<double>::iterator)’
A<double> a(v.begin(), v.end());
^
test.cpp:22:7: note: candidate: template<class InputIterator, class> A<T>::A(InputIterator, InputIterator)
A(InputIterator first, InputIterator last) : v(first, last) {}
^
test.cpp:22:7: note: template argument deduction/substitution failed:
test.cpp: In substitution of ‘template<bool _Cond, class _Tp> using enable_if_t = typename std::enable_if::type [with bool _Cond = ((const std::type_info*)(& _ZTISt26random_access_iterator_tag))->std::type_info::operator==(_ZTISt18input_iterator_tag); _Tp = void]’:
test.cpp:18:16: required from here
test.cpp:19:49: error: call to non-constexpr function ‘bool std::type_info::operator==(const std::type_info&) const’
typeid(traits<InputIterator>) ==
test.cpp:18:16: note: in template argument for type ‘bool’
typename = std::enable_if_t<
^~~~~~~~
test.cpp:10:7: note: candidate: A<double>::A(const A<double>&)
class A{
^
test.cpp:10:7: note: candidate expects 1 argument, 2 provided
test.cpp:10:7: note: candidate: A<double>::A(A<double>&&)
test.cpp:10:7: note: candidate expects 1 argument, 2 provided
I decided to use default template parameter because is more suitable for constructors. The use of the operator typeid()
is because I find it really easy to read when mantaining the code, but I can't get it to work in any way.
Other solutions look very weird and are really obscure (like forcing the InputIterator parameter to have certain methods such as *it or ++it). If there's no way I can do this, I would appreciate a, more or less, easy-to-read solution.
In order to do SFINAE you need to assure that evaluation of involving expressions happen at compile time. For typeid
the following applies:
When applied to an expression of polymorphic type, evaluation of a typeid expression may involve runtime overhead (a virtual table lookup), otherwise typeid expression is resolved at compile time.
Thus, I wouldn't consider typeid
a good option for static (compile time) polymorphism.
One way you could solve your problem is by using tag dispatching in combination with a delegating costructor, as follows:
template <typename T>
class A{
std::vector<T> v;
template <typename InputIterator>
A(InputIterator first, InputIterator last, std::input_iterator_tag) : v(first, last) {}
public:
template<typename InputIterator> A(InputIterator first, InputIterator last)
: A(first, last, typename std::iterator_traits<InputIterator>::iterator_category()) {}
};
Live Demo
The use of the operator
typeid()
is because I find it really easy to read when mantaining the code, but I can't get it to work in any way.
typeid
is a facility mainly used to query type information at runtime. While you might find it more "readable", it's not the right tool for the job and it will confuse every other C++ developer.
Other solutions look very weird and are really obscure (like forcing the InputIterator parameter to have certain methods such as
*it
or++it
)
I strongly advise you to reconsider this. An InputIterator
is a concept that describers valid operations on a type. The whole idea behind a concept is checking for operation validity - there is nothing more than InputIterator
than its requirements.
The correct and idiomatic way of solving your issue is creating a constexpr bool is_input_iterator<T>
variable/function that returns true
if T
matches InputIterator
's requirements. You can easily implement that using the detection idiom... and it doesn't get more readable than std::enable_if_t<is_input_iterator<T>>
.
Here a simplified example (I'm not checking for all the requirements):
template <typename T>
using supports_increment = decltype(++std::declval<T>());
template <typename T>
using supports_dereference = decltype(*std::declval<T>());
template <typename T>
constexpr bool is_input_iterator =
std::experimental::is_detected_v<supports_increment, T>
&& std::experimental::is_detected_v<supports_dereference, T>;
template <typename InputIterator,
typename = std::enable_if_t<is_input_iterator<InputIterator>>>
A(InputIterator first, InputIterator last) : v(first, last) {}
live example on wandbox
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With