C-Binding in python

A simple c-binding example in python with ctypes

This repository contains a simple example to test the c-binding capabilities of python. The module used to realize the binding is called `ctypes.

The ctypes binding is based on shared libraries. The process employed to bind c functions with python is the following:

• Compile the c code as a shared library
• Load the shared library in python
• Explicitely define the argument and return types in python
• Call the c function using the ctypes module

Compilation

In order to compile the shared library, just execute the makefile.

make clean
make

Optionally, a 32 bit compilation is provided for testing purposes. The loading of a 32 bit shared library is not supported by ctypes

make clean
make 32bit

Testing the binding

After the shared library has been compiled (in 64 bit format), the binding can be tested by executing the file c_wrapper.py

python3 c_wrapper.py

Numpy binding example

Numpy arrays can be passed as arguments to a c function wrapped with ctypes as pointers to contiguous memory areas. Let's consider for example a function definition in c that sums a vector y into a vector x of the same length len.

void sum_vectors(int* x, int* y, int len);

Calling lib the python object referencing the shared library that includes the sum_vectors function, the arguments can be specified as follows.

lib.sum_vectors.argtypes = (
	np.ctypeslib.ndpointer(dtype=int, ndim=1), # x
	np.ctypeslib.ndpointer(dtype=int, ndim=1), # y
	ctypes.c_int # len
)

The c function can then be called as follows.

x = np.ones(5)
y = np.ones(5)
lib.sum_vectors(x, y, n)
# x = [2, 2, 2, 2]