Simple C ++ 11 constexpr factorial with triple exceeds maximum template depth

Question

If I write a compilation time factorial function using specialization, the following code may be sufficient and will correctly provide 120 as a result of fact1<5>() :

 template <size_t N> constexpr size_t fact1() { return N*fact1<N-1>(); } template <> constexpr size_t fact1<0>() { return 1; } 

However, with one function body and ternary operator, as in the following code, g ++ 4.7 and Clang ++ 3.2 both exceed the maximum depth of the template increment. It seems like 1 never returns from fact2 . Why does this definition of fact2<5>() not return 120?

 template <size_t N> constexpr size_t fact2() { return N==0 ? 1 : N*fact2<N-1>(); } 

Answer 1

The problem is that no matter what, fact2<N-1> will always be created (even non-executable paths must be compiled, see Effective C ++, I think paragraph 47 or 48). You need to somehow do this only to create the next function, if you're not at the end. One way is to simply say “screw patterns” and go in the usual way constexpr , as @NicolBolas says in his comment.

Another will use one of the methods used in this similar question .