How to return array from C++ function to Python using ctypes

Question

I am using ctypes to implement a C++ function in Python. The C++ function should return a pointer to an array. Unfortunately I haven't figured out, how to access the array in Python. I tried numpy.frombuffer, but that was not successful. It just returned an array of arbitrary numbers. Obviously I didn't used it correctly. Here is a simple example with an array of size 10:

Content of function.cpp:

extern "C" int* function(){ int* information = new int[10]; for(int k=0;k<10;k++){     information[k] = k; } return information; } 

Content of wrapper.py:

import ctypes import numpy as np  output = ctypes.CDLL('./library.so').function()  ArrayType = ctypes.c_double*10 array_pointer = ctypes.cast(output, ctypes.POINTER(ArrayType)) print np.frombuffer(array_pointer.contents) 

To compile the C++ file i am using:

g++ -c -fPIC function.cpp -o function.o g++ -shared -Wl,-soname,library.so -o library.so function.o 

Do you have any suggestions what I have to do to access the array values in Python?

Answer 1

Your python code will work after some minor modifications:

import ctypes  f = ctypes.CDLL('./library.so').function f.restype = ctypes.POINTER(ctypes.c_int * 10) print [i for i in f().contents] # output: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] 

Basically there are two changes:

1. remove numpy-related code and ctypes.cast call since we don't need them.

2. specify the return type to ctypes.POINTER(ctypes.c_int * 10).

   By default foreign functions are assumed to return the C int type, hence we need change it to the desired pointer type.

BTW, returning a newed array from C code to Python code seems inappropriate. Who and when will free the memory? It's better to create arrays in Python code and pass them to C code. This way it's clear that the Python code owns the arrays and takes the responsibility of creating and reclaiming their spaces.

Answer 2

function.cpp returns an int array, while wrapper.py tries to interpret them as doubles. Change ArrayType to ctypes.c_int * 10 and it should work.

---

It's probably easier to just use np.ctypeslib instead of frombuffer yourself. This should look something like

import ctypes from numpy.ctypeslib import ndpointer  lib = ctypes.CDLL('./library.so') lib.function.restype = ndpointer(dtype=ctypes.c_int, shape=(10,))  res = lib.function()