三角函数的模板重载

Question

I am developing a larger library, where I would like to be able to change the floating point precision used. Trigonometric functions are used in a number of places, so I decided to include templated wrappers for eg sine and cosine.

I have made the following simple example, which gives a stack corruption and I cannot figure out why. Any hints

#include <cmath>
#include <iostream>

namespace sps {
template <typename T>
inline T sin(const T& v) { return sin(v); }

template <typename T>
inline T cos(const T& v) { return cos(v); }

template <>
inline float sin<float>(const float& v) { return sinf(v); }

template <>
inline float cos<float>(const float& v) { return cosf(v); }
}  // namespace sps

template float sps::sin(const float& v);
template float sps::cos(const float& v);
template double sps::sin(const double& v);
template double sps::cos(const double& v);

int main()
{
  double d = 2.0;
  std::cout << sps::sin(d) << std::endl; /* (*) */
  float f = 2.0f;
  std::cout << sps::sin(f) << std::endl;
  return 0;
}

If I explicitly state, which functions to use, eg sps::sin(f), I still get a stack corruption. The stack corruption is thrown in the line with the asterisk (*). If I omit the explicit instantiations, the stack corruption still occurs. I tried changing the inputs to by-value, but this has no effect either.

I get the same error using gcc 6.3.0 and MSCV 2017.

Thanks in advance Jens

Answer 1

You are getting an infinit recursion:

template <typename T>
inline T sin(const T& v) { return sin(v); }  // keep calling sin

what you need is to change the return value to:

template <typename T>
inline T sin(const T& v) { return std::sin(v); } 
                                  ^^^