Cython C ++ shell operator overload error ()
Linked to my previous question. Using Cython to port a C ++ class using OpenCV types as parameters
Now I am stuck with another error. My cython shell code like OpenCV Matx33d looks like this:
cdef extern from "opencv2/core/core.hpp" namespace "cv":
cdef cppclass Matx33d "cv::Matx<double, 3, 3>":
Matx33d()
Matx33d(double v0, double v1, double v2, double v3, double v4, double v5, double v6, double v7, double v8)
double& operator()(int i, int j)
Then I define a function to copy Matx33d into a numpy array.
cdef Matx33d2numpy(Matx33d &m):
cdef np.ndarray[np.double_t, ndim=2] np_m = np.empty((3,3), dtype=np.float64)
np_m[0,0]= m(0,0); np_m[0,1]= m(0,1); np_m[0,2]= m(0,2)
np_m[1,0]= m(1,0); np_m[1,1]= m(1,1); np_m[1,2]= m(1,2)
np_m[2,0]= m(2,0); np_m[2,1]= m(2,1); np_m[2,2]= m(2,2)
return np_m
When I compile the cython shell I get this error
geom_gateway.cpp(2528) error C3861: '()': identifier not found
This corresponds to the first use of Matx33d :: operator (), which is when accessing m (0,0) in the code above. If I look at the generated line 2528 geom_gateway.cpp I get:
*__Pyx_BufPtrStrided2d(__pyx_t_5numpy_double_t *, __pyx_pybuffernd_np_m.rcbuffer->pybuffer.buf, __pyx_t_6, __pyx_pybuffernd_np_m.diminfo[0].strides, __pyx_t_7, __pyx_pybuffernd_np_m.diminfo[1].strides) = operator()(0, 0);
I don't understand that this operator () (0, 0) is there alone at the end of the line without any object !! How is this possible? Is this a Cython bug? or is the syntax I'm using for the () operator wrong? Any help is appreciated!
source to share
Ok, I don't know why this error happened, to me it looks like the syntax
double& operator()(int i, int j)
should work, but it doesn't. This syntax works for other operators like +, -, /, *
An alternative syntax that works is the following:
double& get "operator()"(int i, int j)
then in cython code when we want to use operator () (i, j) we will call get (i, j) instead
source to share