Casting an array of C structs to a numpy array

Question

A function I'm calling from a shared library returns a structure called info similar to this:

typedef struct cmplx {
 double real;
 double imag;
} cmplx;
typedef struct info{
 char *name;
 int arr_len;
 double *real_data
 cmplx *cmplx_data;
} info;

One of the fields of the structure is an array of doubles while the other is an array of complex numbers. How do I convert the array of complex numbers to a numpy array? For doubles I have the following:

from ctypes import *
import numpy as np
class cmplx(Structure):
 _fields_ = [("real", c_double),
 ("imag", c_double)]
class info(Structure):
 _fields_ = [("name", c_char_p),
 ("arr_len", c_int),
 ("real_data", POINTER(c_double)),
 ("cmplx_data", POINTER(cmplx))]
c_func.restype = info
ret_val = c_func()
data = np.ctypeslib.as_array(ret_val.contents.real_data, shape=(info.contents.arr_len,))

Is there a numpy one liner for complex numbers? I can do this using loops.

Answer 1

Define your field as double and make a complex view with numpy:

class info(Structure):
 _fields_ = [("name", c_char_p),
 ("arr_len", c_int),
 ("real_data", POINTER(c_double)),
 ("cmplx_data", POINTER(c_double))]
c_func.restype = info
ret_val = c_func()
data = np.ctypeslib.as_array(ret_val.contents.real_data, shape=(info.contents.arr_len,))
complex_data = np.ctypeslib.as_array(ret_val.contents.cmplx_data, shape=(info.contents.arr_len,2)).view('complex128')

• Ashwith Rego: The shared library requires that the field be a pointer to the cmplx struct.
• Ashwith Rego: Ok, that did work. Could you please explain why? C requires it to be a pointer to the cmplx struct but we're using a pointer to a c_double object in python.
• Daniel: a complex number is nothing else, than two doubles. For the interface, the type is not important (you can even use bytes). In the end, the view of your numpy-array defines the type.
• Ashwith Rego: I see. Thank you so much for your help!