Trying to pass a custom C++ matrix to a numpy array

Question

I am trying to do some Python wrapping to use custom C++ stuff. The main type we use is a 2D gray image type with data allocated in a 1D buffer. I try to wrap it this way (following an example in an ubuntu forum):

PyObject* to_python_object(const custom2DImage& img) {
  int type_num = (int)NPY_UBYTE;
  long int dims[2] = {img.nr(), img.nc()};

  uchar** tmp_img = new uchar*[img.nr()];
  tmp_img[0] = new uchar[img.nr() * img.nc()];
  for (int i = 1; i < img.nr(); ++i)
    tmp_img[i] = tmp_img[0] + img.nc();
  for (int i = 0; i < img.nc(); ++i)
    memcpy(tmp_mat[i], &img(i, 0), img.nc() * sizeof(uchar));
  PyObject* py_img = PyArray_SimpleNewFromData(2, dims, type_num, tmp_img[0]);

  Py_INCREF(py_img);
  delete[] tmp_img[0];
  delete[] tmp_img;

  PyObject *repr = PyObject_Repr(py_img);
  const char *s = PyString_AsString(repr);
  cout << s << endl;
  Py_XDECREF(repr);

  return py_img;
}

My Python object representation from the C++ function is ok but as soon as I try to print it in my python main, it segfault (even though the shape is good). The Python code is as follow:

img = cst.read_image(filename);
if img is None:
    print("Can not load image from", filename)
    sys.exit(-1)
print(img)

Do you have an idea why do I have this issue ?

baptiste · Accepted Answer · 2016-11-14 10:43:26Z

1

Ok so I thought the PyArray_SimpleNewFromData function was copying data but it seems it doesn't. My mistake was to free the tmp_img 2D array.

answered Nov 14, 2016 at 10:43

baptiste

1,1693 gold badges15 silver badges33 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Trying to pass a custom C++ matrix to a numpy array

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related