Writing Python bindings of existing C libraries – (3) – Building and Installing with distutils

Posted on Thu 13 August 2009 in HowTo, Igalia, Linux, Maemo (EN), Programmazione, Python • Tagged with binding, distutils, library, maemo, Python, setup

In the last post of this series, we saw how to write a simple binding and we finished to build and install it manually. This is of course not a good way to manage the building/installation procedure.

In Python we can use a library called distutils that let us to automatize the building and installing process. I'll use the foo source code to create the package, so it will be easier to understand.

Using distutils

All we have to do is to write a setup.py file similar to this one:

from distutils.core import setup, Extension

foomodule = Extension('foo', sources = ['foo.c'])

setup (
    name = 'Foo',
    version = '1.0',  
    description = 'This is a package for Foo',  
    ext_modules = [foomodule]
)

As you can see, we have to first import needed modules with: from distutils.core import setup, Extension
then we create an entry for each module we have (in this case just one, "foomodule"). We then call the setup() method passing it all the parameters and our setup.py is complete.

Building and installing

To test it we can try to build the package in this way:

python2.5 setup.py build

if we want to install the module in our system:

python2.5 setup.py install

References


Writing Python bindings of existing C libraries – (2) – A simple example of binding

Posted on Thu 06 August 2009 in HowTo, Igalia, Linux, Maemo (EN), Programmazione, Python • Tagged with binding, Igalia, maemo, Python

Introduction

As I promised in the preceding post, I'll provide a very easy example of a python binding. Let's suppose we don't want to use the methods included in Python to sum two integer values and we want to do it in C and then call the add method from a python script. I'll write the complete source code first and then I'll explain all the parts of it.

Source Code

#include <Python.h>

static PyObject *foo_add(PyObject *self, PyObject *args)  
{  
    int a;  
    int b;

    if (!PyArg_ParseTuple(args, "ii", &a, &b))  
    {  
        return NULL;  
    }

    return Py_BuildValue("i", a + b);  
}

static PyMethodDef foo_methods[] = {  
    { "add", (PyCFunction)foo_add, METH_VARARGS, NULL },  
    { NULL, NULL, 0, NULL }  
};

PyMODINIT_FUNC initfoo()
{  
    Py_InitModule3("foo", foo_methods, "My first extension module.");  
}

How it works

First of all we have to include Python.h in our C file. This allows us to write an extension for Python language. To be able to include this header, we must have the python development packages installed in our system. For example in Debian based distributions we can install them with this command:

sudo apt-get install python2.5-dev

Every module has at least three parts. In the first part we write methods we want to call from the final python module: in this case we have a method called foo_add where "foo" is the name of the module and "add" the name of the method. Every method is declared as static PyObject. The method does anything particular except calling PyArg_ParseTuple to validate the input (we'll discuss this later), adding the two passed numbers and returning the result.

In the second part we have something like a dictionary, defined as static PyMethodDef and called foo_methods (where "foo" again is the name of the module). For each method we want to expose in our python module, we have to add something like this:

{"add", (PyCFunction)foo_add, METH_VARARGS, NULL}

where "add" is the name of the method we want to be visible in our module, (PyCFunction)foo_add is a pointer to our foo_add method, implemented in the C module, METH_VARARGS means that we want to pass some parameters to the function and the last one would be the description of the method (we can leave it NULL if we want).

Third part allows us to register the defined method/s and the module:

Py_InitModule3("foo", foo_methods, "My first extension module.");

Parsing Parameters

The PyArg_ParseTuple function extracts arguments from the PyObject passed as parameter to the current method and follows almost the sscanf syntax to parse parameters (in this case we had "ii" for two integers). You can fin the complete reference here: http://docs.python.org/c-api/arg.html

Building and installing

To build the module, we have to be in the source directory and execute this command:

gcc -shared -I/usr/include/python2.5 foo.c -o foo.so

then we've to copy the generated module to the python's modules directory:

cp foo.so /usr/lib/python2.5/site-packages/

Testing our module

Testing the module is really easy. We've to start a python shell or create a python script with the following source code:

import foo
print foo.add(2, 3)

if all is working fine, the printed result should be 5

References


Writing Python bindings of existing C libraries - (1) - Introduction

Posted on Mon 03 August 2009 in HowTo, Igalia, Linux, Maemo (EN), Programmazione, Python • Tagged with bindings, C, Igalia, libraries, library, maemo, pymaemo, Python

This summer I'm having the pleasure of working in Igalia (a spanish free software company) for a couple of months and they assigned me to an interesting project: developing Python bindings for MAFW library (a Maemo multimedia library that will be used in Fremantle release).

Having the opportunity to work both with C (yes, Python bindings are almost C code) and Python (it's a good practice to write unittest of all implemented methods) it's a good way to improve my knowledges in both languages and since I wasn't able to find much documentation about these kind of things, I'm going to share my own experiences.

What is a Binding?

A binding is a Python module, written in C language, that allows Python developers to call functions from existing C libraries from their python applications. It's just like a "bridge" from C world to Python one.

Why writing bindings?

There are a couple of reasons to write python bindings instead of writing a library in python language from scratch.

First of all I don't think is good duplicating code, so if a library already exists and it's written in C, why writing it again in another language? There's no reason. A lot of code already exist in C world and all we have to do is to create a bridge with python world.

Another good reason, in particular when a C library doesn't exist yet, is the fact that python code is slower than C code for some tasks (for example multimedia codecs). In these cases is good to implement the core library in C language and then create a python binding for it.

Coming next

As the title of this post says, this is only an introduction to the subjects I'm going to write about. If you have any particular request about any argument you would like to read, please feel free to leave me a comment. Next posts will talk about these things:

  • A simple example of binding: I'll write a simple library in C language and I'll show how to create the relative python binding, providing complete source code and an example for python developers.
  • Building and installing python bindings with distutils: I'll explain how to use distutils to build and install the binding (using the well know method "python setup.py install").
  • Defining new types: this post will be about how to write new types in C language and being able to use them from python code.
  • Using codegen to write bindings: I'll explain how to use codegen utils to automate lot of tasks, to generate the most part of binding code and how to customize the generated code using overrides.