1)template <class T = int, class U = double> //compiles
2)template <class T, class U =double> //compiles
3)template <class T = int, class U> //fails
Why does 1 and 2 compile whereas 3 does not?
Add a comment
|
3 Answers
For the same reason why:
void f(int = 0, int);
fails.
There is no way to use 3rd version default parameter:
template<class T = int, class U> class B { ... };
B<, short> var; // ??? no such syntax
Comments
(3) is ill-formed because
C++03 [Section 14.1/11] says
If a template-parameter has a default template-argument, all subsequent template-parameters shall have a default template-argument supplied.
6 Comments
JoshD
Great answer +1. Where can I get a copy of the standard?
Prasoon Saurav
@JoshD : You can find it here (ISO/IEC 14882 2003 is C++03)
Yakov Galka
@JoshD: AFAIK you can't get it for free. However you can look at the drafts of the next standard.
sellibitze
Although, I would have liked (3) to be valid in case of function templates. We still have TAD (template argument deduction) to determine the trailing, non-defaulted template parameters.
Yakov Galka
@sellibitze in that case you can add a default
class = void and it'll solve your problem. |
If you put that into some context, the third way may actually be legal, provided that the second default has been given earlier.
template <class T, class U = double>
struct X;
template <class T = int, class U> //here
struct X {};
int main()
{
X<> x;
X<float> y;
X<char, char> z;
}
Comments
