Creating an instance of a C++ class and calling methods on it in Python

Question

I am trying to use a C++ library in my python app. I can load the dll in python but could not find any solution on how to create an instance of a class that is inside that c++ dll and invoke methods on that onject.

Following is what I did and want

C++ code inside My.dll

class MyClass
{
  public:
    MyMethod(int param);
}

Python code

from ctypes import *
myDll = windll.LoadLibrary("My.dll")

I want to do the following

myClassInstance = myDll.InstantiateMyClass()
myClassInstance.MyMethod(5)

Answer 1

While it might be possible with ctypes, it certainly won't be as straightforward as that. It'll be much easier to use e.g. Boost.Python or Cython to create a proper CPython extension that exposes that class as a Python type.

Answer 2

Loading a C++ dll with Ctypes is dangerous and has some strong limitations. The exported function name is not the same as you declared, unless you declared the function in C++ as 'extern "C"'. This is only possible for pure functions, not for member functions. The C++ compiler does something called "name mangling", see http://en.wikipedia.org/wiki/Name_mangling#Name_mangling_in_C.2B.2B.

I suggest two solutions:

1. You write some C++ code with pure functions, declared as 'extern "C"', which expose the functionality you need.
2. I really recommend using Cython http://cython.org/, especially http://docs.cython.org/src/userguide/wrapping_CPlusPlus.html should help you. I used it a lot, and for me it is the best way to wrap C++ code to Python.

ADDITIONAL NOTES:

I tried boost python some times but found it hard to use. It has its own build system which you have to learn, the compiling process is very very slow, and due to the template syntax the code gets hard to read. It thing the context behind boost python is very cool, but in my opinion it is hard to use.

I also tried SIP and SWIG which I did not feel very comfortable with.

I really recommend to use Cython.