I\'m trying to implement std::is_enum
. Here is my code so far:
template
struct is_enum {
static bool value;
};
template
The best way to implement this is to use compiler magic, and I believe most implementations do this.
For example, here's libc++'s implementation for gcc >= 4.3 and any compiler that __has_feature(is_enum)
1
template <class _Tp> struct _LIBCPP_VISIBLE is_enum
: public integral_constant<bool, __is_enum(_Tp)> {};
For all other compilers libc++ does:
template <class _Tp> struct _LIBCPP_VISIBLE is_enum
: public integral_constant<bool, !is_void<_Tp>::value &&
!is_integral<_Tp>::value &&
!is_floating_point<_Tp>::value &&
!is_array<_Tp>::value &&
!is_pointer<_Tp>::value &&
!is_reference<_Tp>::value &&
!is_member_pointer<_Tp>::value &&
!is_union<_Tp>::value &&
!is_class<_Tp>::value &&
!is_function<_Tp>::value > {};
Some of those other type traits still require compiler magic.2 E.g. is_union
. However, that condition can be rewritten such that it doesn't need compiler magic. This can be done by replacing the seperate checks for unions and classes with a single check for both, as Johannes Schaub points out.
1. So far as I know only clang implements __has_feature
, unfortunately.
2. It's interesting that libc++ does have a version of is_union<T>
and is_class<T>
that do not use compiler intrinsics, but as a result they provide erroneous results for union types. But their erroneous results are complementary so libc++'s fallback implementation of is_enum<T>
provides accurate results.
You problem is that
template<enum E>
Is interpreted as unnamed parameter with type forward declared enum named E
.
Semantically same to
template<int>
Just substituting int
with enum E
.
This
template<enum E>
promises that the template argument is a value of type enum E
. The argument is NOT a type (Type template arguments are introduced by typename
, or for backward compatibility, class
. Even struct
isn't allowed). It's just like saying
template<int i>
except no name is given for the variable.
Things go wrong from there.