How can I create and pass as argument a multidimensional ctypes array?

Question

I have a shared library with a function that takes a int ** like this:

void printarray(int **array, int n, int m)
{
        int i, j;
        for (i = 0; i < m; i++)
                for (j = 0; j < n; j++)
                        printf("%d\n", array[i][j]);
}

I would like to call the function from Python withctypes. I have written the following to do this:

from ctypes import *

cdll.LoadLibrary("libssa.so")
libssa = CDLL("libssa.so")

a = [[1, 2, 3], [4, 5, 6]]
 
T = ((c_int * 3) * 2)
array = T()
for i in range(2):
    for j in range(3):
        array[i][j] = a[i][j]

libssa.printarray(array, c_int(3), c_int(2))

I expected that the type T would be an array of pointers to arrays of ints, and that array would be a pointer to an object of that type.

However, a segmentation fault occurs whenever an access to array is made in the C code. In particular, for this example, valgrind points to this line: printf("%d\n", array[i][j]); as the source of the segfault.

The question is, what is the proper way to construct a ctypes object that can be used as an argument of type int **?

"I expected that the type T would be an array of pointers to arrays of ints, and that array would be a pointer to an object of that type." It is, but just because you have pointers to arrays of ints, doesn't mean those pointers to arrays of ints actually have arrays of ints that they're pointing at. But regardless, a multidimensional array works differently from that in C. — Karl Knechtel
– Karl Knechtel, Commented Jul 16, 2020 at 20:58

Karl Knechtel · Accepted Answer · 2020-07-16 21:01:35Z

2

Be aware that int** is not equal to int[][] (see this post). This is due to the internal structure of an multidimensional array: It could be represented as

                      +-------------+
array ->  array[0] -> | array[0][0] |
                      +-------------+
                      | array[0][1] |
                      +-------------+
                      |     ...     |
                      +-------------+
          array[1] -> | array[1][0] |
                      +-------------+
                            ...

where each square shows how the corresponding value can be accessed. But int** would be a structure like

         +------+      +-----+
int** -> | int* | ---> | int |
         +------+      +-----+
         | int* | -\   | int |
         +------+  |   +-----+
           ...     |   | ... |
                   |   +-----+
                   |
                   |   +-----+
                   \-> | int |
                       +-----+
                         ...

Each square should represent a memory location, with the type in it. In contrast to Python, the size of the array is not saved in c, so access with array[i][j] fails when the pointer was passed as an parameter. Better use a one-dimensional array and calculate the second dimension yourself, so

void printarray(int *array, int n, int m)
{
        int i, j;
        for (i = 0; i < m; i++)
                for (j = 0; j < n; j++)
                        printf("%d\n", array[i + j*m]);
}

or array[j + i*n], depending on your implementation.

edited Jul 16, 2020 at 21:01

Karl Knechtel

61.4k14 gold badges133 silver badges193 bronze badges

answered Jul 16, 2020 at 19:39

fcdt

2,5215 gold badges17 silver badges27 bronze badges

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

1 Comment

Karl Knechtel Over a year ago

The problem isn't just that sizes aren't tracked intrinsically; the problem in OP's code is that the int*s in the main array don't have allocations of their own to point to.

Mark Tolonen · Accepted Answer · 2020-07-16 20:55:40Z

An array of pointers is different than a two-dimensional array. Here's what would call your function.

DLL code (modified for Windows and m/n in for loops were backward):

#include <stdio.h>

__declspec(dllexport)
void printarray(int **array, int n, int m)
{
        int i, j;
        for (i = 0; i < n; i++)
                for (j = 0; j < m; j++)
                        printf("array[%d][%d] = %d\n", i, j, array[i][j]);
}

test.py:

from ctypes import *

dll = CDLL('./test')
dll.printarray.argtypes = POINTER(POINTER(c_int)),c_int,c_int
dll.printarray.restype = None

def wrap(a):
    x = len(a)
    y = len(a[0])
    arr = (POINTER(c_int) * x)()  # allocate array of pointers in X dimension
    for i in range(x):
        arr[i] = (c_int * y)(*a[i]) # allocate array in Y dimension and populate it
    return arr,x,y

a = [[1,2,3],[4,5,6]]

dll.printarray(*wrap(a))

Output:

array[0][0] = 1
array[0][1] = 2
array[0][2] = 3
array[1][0] = 4
array[1][1] = 5
array[1][2] = 6

Collectives™ on Stack Overflow

How can I create and pass as argument a multidimensional ctypes array?

2 Answers 2

1 Comment

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related