C++ division operator overloading unwanted cast

[10se1ucgo]

When wrapping a class like so

# vec3.pxd
cdef extern from "math3d_c.h":
    cdef cppclass Vector3[T]:
        T x, y, z
        Vector3() except +
        Vector3(T x, T y, T z) except +

        Vector3[T] operator*(T &b)
        Vector3[T] operator/(T &b)

# math3d.pyx
from . cimport vec3

cdef class Vector3:
    cdef vec3.Vector3[double] c_vec

    def __init__(self, double x=0, double y=0, double z=0):
        self.c_vec.x = x
        self.c_vec.y = y
        self.c_vec.z = z

    def __mul__(Vector3 a, double scalar):
        return new_vector3_from(a.c_vec * scalar)

    def __truediv__(Vector3 a, double scalar):
        if scalar == 0:
            raise ZeroDivisionError("Vector3 division by zero")
        return new_vector3_from(a.c_vec / scalar)

# new_vector3_from just constructs a Python Vector3 object from a C++ Vector3 object, unrelated

The code generated by Cython for __mul__ is correct, however for __truediv__, Cython unnecessarily casts scalar to a Vector3.

// mul
 __pyx_t_1 = ((PyObject *)__pyx_f_6extlib_6math3d_new_vector3_from((__pyx_v_a->c_vec * __pyx_v_scalar))); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)

// truediv
__pyx_t_2 = ((PyObject *)__pyx_f_6extlib_6math3d_new_vector3_from((__pyx_v_a->c_vec / ((Vector3<double> )__pyx_v_scalar)))); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 43, __pyx_L1_error)

Which, of course, causes an error if the wrapped class doesn't provide an overload for that method.

[scoder]

That looks wrong. Might be a problem with the code that tries to find a common floatish type for true division.

Please also see
http://docs.cython.org/en/latest/src/userguide/special_methods.html#arithmetic-operators

[scoder]

That looks wrong. Might be a problem with the code that tries to find a common floatish type for true division.

Please also see
http://docs.cython.org/en/latest/src/userguide/special_methods.html#arithmetic-operators