I am using a C Library and interfacing it with python, without going into too much detail, my library has a C function which we will call 'foo' that returns void const* (it either returns float const* or double const* depending on application). In python I declared this as:
from ctypes import *
mylib = cdll.LoadLibrary('mylib.so')
foo = mylib.foo
foo.restype = c_void_p
From now we will assume that we are using float not double and given the expected dimensions of the pointer as 'D' e.g. 12
D = 12
I am trying to do a deep copy of the return value into a python numpy array so that I don't need to worry about the memory allocated by the C Library being freed, I have tried implementing it as follows:
import numpy as np
x = np.frombuffer(foo(), np.float32, D * np.dtype(np.float32).itemsize).copy()
AttributeError: 'int' object has no attribute '__buffer__'
But it doesn't seem to work, I thought that the return type of 'foo' would be a ctypes c_void_p type python object but actually it is just an int. I am guessing the value of the int corresponds to the address in memory of the first float in the array but maybe I am wrong. There must be an easy way of doing this, has anybody got any suggestions?
If you would like more details (on the library, etc.) let me know I tried to keep the question simple.
Thanks.
x = np.frombuffer((c_float * D).from_address(foo()), np.float32).copy()