Question

I am trying to use the GMP number library together with Eigen matrix library. I try to instantiate the template:

Matrix<typename Scalar, int RowsAtCompileTime, int ColsAtCompileTime>

with

Matrix<mpz_class, 3, 3> matrix;

where mpz_class is a number class from the GMP library.

I get a compiler error:

 /usr/include/eigen3/Eigen/src/Core/MathFunctions.h:409: error: 
 invalid static_cast from     
 type ‘const __gmp_expr<__mpz_struct [1], __mpz_struct [1]>’ 
 to type ‘int’

When I examine the source code of the Eigen library, I find out, that the problem is that
mpz_class cannot be static_cast -ed to int in this template:

template<typename OldType, typename NewType>
struct cast_impl
{
  static inline NewType run(const OldType& x)
  {
    return static_cast<NewType>(x);
  }
};

How can I bypass this problem? I know how to convert mpz_class to int in runtime, but it must be done by the compiler, since static_cast is compile time.

Answer 1

Assuming mpz_class is safe to subclass, you can just use a subclass and write a conversion operator:

class your_mpz_class : public mpz_class
{
public:

  // Write public constructors as needed

  operator int()
  {
    return /* Whatever must be returned */;
  }
};