I have some static const floating point member variables, on which I want to do some compile time static asserts. The following code compiles in gcc but fails in both clang and Visual Studio:
#include <boost/static_assert.hpp>
#include <iostream>
template<typename Scalar>
class ProbModel {
public:
static const Scalar probA;
static const Scalar probB;
private:
BOOST_STATIC_ASSERT_MSG(probA < 1, "Weird Parameter");
BOOST_STATIC_ASSERT_MSG(probB < 1, "Weird Parameter");
BOOST_STATIC_ASSERT_MSG(probA < probB, "Weird Parameter");
};
// Initializations
template<typename Scalar> const Scalar
ProbModel<Scalar>::probA = 0.3;
template<typename Scalar> const Scalar
ProbModel<Scalar>::probB = 0.6;
int main(int argc, char* argv[]) {
typedef ProbModel<float> Modelf;
std::cout << "ProbA = " << Modelf::probA << std::endl;
}
I get errors in Visual Studio 2013 and Clang for the static assert statements complaing about constant expression:
2>..\static_assert_experiments.cpp(11): error C2057: expected constant expression
2>..\static_assert_experiments.cpp(12): error C2057: expected constant expression
2>..\static_assert_experiments.cpp(13): error C2057: expected constant expression
A non c++11 solution for doing this will be highly appreciated. Also is there other ways to do these kind of checks on static parameters?
As dyp stated in comment, floating-point type variables cannot be used in constant expression. A possible workaround would be to use rational number.
std::ratio
is C++11 but can easily be ported to C++03
template<typename Scalar>
class ProbModel {
public:
static Scalar getProbA() { return Scalar(probA::num) / probA::den; }
static Scalar getProbB() { return Scalar(probB::num) / probB::den; }
private:
typedef std::ratio<3, 10> probA; // 0.3
typedef std::ratio<6, 10> probB; // 0.6
BOOST_STATIC_ASSERT_MSG(probA::num < probA::den, "Weird Parameter");
BOOST_STATIC_ASSERT_MSG(probB::num < probB::den, "Weird Parameter");
BOOST_STATIC_ASSERT_MSG(probA::num * probB::den < probB::num * probA::den, "Weird Parameter");
};