/CPy

Primary LanguageC

Method of Extension Library Between C and Python

Call a C Function From Python

By ctypes

from ctypes import cdll, c_int, c_float
clib = cdll.LoadLibrary("PATH/TO/YOUR/CLIB.so")
clib.func(c_int(5), c_float(0.3))

Call a Python Function From C

By Python/C API

#include <stdio.h>
#include <python3.6/Python.h>

PyObject *pow_func;
Py_Initialize();
PySys_SetPath(Py_DecodeLocale(".", NULL));

/* Get reference to the python function */
PyObject *u_name, *module;
u_name = PyUnicode_FromString("/YOUR/PYTHON/MODULE");
module = PyImport_Import("YOUR_FUNCTION_NAME");
Py_DECREF(u_name);
pow_func = PyObject_GetAttrString(module, symbol);

/* Build arguments */
args = Py_BuildValue("(dd)", x, y);
kwargs = NULL;

/* Call the function */
result = PyObject_Call(pow_func, args, kwargs);
Py_DECREF(args);
Py_XDECREF(kwargs);

/* Create the return value */
retval = PyFloat_AsDouble(result);
Py_DECREF(result);

return retval;

By Cython

Function

  • Cython just like C, if you declare a variable as specific type, you can't git it other type data, for example:
cdef np.ndarray[double, ndim=1] x
# Wrong Example 1
x = "1"

# Wrong Example 2
x = np.asaray(x, dtype=np.int)

# Correct method:
cdef np.ndarray[np.int_t, ndim=1] y = np.asaray(x, dtype=np.int)
  • Using cdef for c/c++, cpdef for c/c++ and ptrhon, such as:
cdef public np.ndarray[double, ndim=1] compute(char* path):
    print("path: ", path)
    // your algorithm go here ..
    return np.zeros(1)

Class

cdef public class Foo[object c_Foo, type c_Foo]:
    cdef public int attr 
    def __cinit__(self):
        self.attr = 0
    cdef int add(self, x):
        return x + self.attr
  • [object c_Foo, type c_Foo_t] ensures that c_Foo is available to C/C++ as struct c_Foo, and the Python TypeObject defining Foo is available as c_Foo_t.
  • If you don't need Foo class, you can replace cdef public class Foo[object c_Foo, type c_Foo]: with cdef class Foo:.

attributes

  • It's necessary to declare types of return object amd initial attributes.
from libcpp.string cimport string
from libcpp cimport bool

# common types in Cython
# cdef  int 
# cdef  float 
# cdef  string 
# cdef  object 
# cdef  list 
# cdef  bool 
# cdef YOUR_CUSTOMER_OBJECT
  • Note that if you using your own type, you're suppose to declare object with right type.

methods

  • if you want to call methods, you have to wrap method as function.
cdef public Foo get_method():
     return Foo()

cdef public int method_add(FOO foo, int x):
    cdef int ret = foo.add(x)
    return ret

cmake

  • Copy source files(.c) into ./cmake/src
  • Copy head files(.h) into ./cmake/include
  • Copy shared object files(.so) into ./cmake/ and rename to lib*.so
  • Create CMakeLists.txt.
  • Note that although you execute the commands in ./cmake/build, but you should define related path based on ./cmake
  • Then, run the following command in ./cmake/:
$ mkdir build
$ cd build
$ cmake ../ > /dev/null; make
  • After cmake, you can found executable file with name defined in ./cmake/CMakeLists.txt

reference:

[1] Extension between C and Python: https://medium.com/pyladies-taiwan/%E5%A6%82py%E4%BC%BCc-python-%E8%88%87-c-%E7%9A%84%E5%85%B1%E7%94%9F%E6%B3%95%E5%89%87-568add0ba5b8 [2] Cython for Python: https://www.cnblogs.com/freeweb/p/6548208.html [3] Python-C API: https://learning-python.com/class/Workbook/unit16.htm [4] C Extension: http://shouce.jb51.net/Python_jj/c_extensions/python_c_api.html [5] C Extension: https://segmentfault.com/a/1190000000479951#articleHeader1 [6] Numpy for C (Python side): https://stackoverflow.com/questions/28474931/return-numpy-array-in-cython-defined-c-function [7] Numpy type for C (Python side): https://cython-docs2.readthedocs.io/en/latest/src/tutorial/numpy.html [8] Numpy for C (C side): https://stackoverflow.com/questions/28483819/how-can-i-use-the-numpy-c-api-in-both-c-header-and-source-file [9] C function with Numpy object: https://github.com/macklin/C_arraytest [10] Numpy type in C(C side): https://docs.scipy.org/doc/numpy/reference/c-api.dtype.html [11] Numpy type in C(Python side): https://numpy.org/devdocs/user/basics.types.html [12] PyArrayObject: https://blog.csdn.net/zj360202/article/details/79214363 [13] PyClass for Cython: https://www.reddit.com/r/Cython/comments/8mqlca/convert_python_class_to_cython/ [14] Using Python Object in C++: https://www.tutorialspoint.com/How-to-use-Python-object-in-Cplusplus [15] Declare a public class: https://stackoverflow.com/questions/40572073/how-to-create-a-public-cython-function-that-can-receive-c-struct-instance-or-p