Icon Unrolling Rotations


Icon Animation Blend Spaces without Triangulation


Icon Quaternion Weighted Average


Icon BVHView


Icon Dead Blending Node in Unreal Engine


Icon Propagating Velocities through Animation Systems


Icon Cubic Interpolation of Quaternions


Icon Dead Blending


Icon Perfect Tracking with Springs


Icon Creating Looping Animations from Motion Capture


Icon My Favourite Things


Icon Inertialization Transition Cost


Icon Scalar Velocity


Icon Tags, Ranges and Masks


Icon Fitting Code Driven Displacement


Icon atoi and Trillions of Whales


Icon SuperTrack: Motion Tracking for Physically Simulated Characters using Supervised Learning


Icon Joint Limits


Icon Code vs Data Driven Displacement


Icon Exponential Map, Angle Axis, and Angular Velocity


Icon Encoding Events for Neural Networks


Icon Visualizing Rotation Spaces


Icon Spring-It-On: The Game Developer's Spring-Roll-Call


Icon Interviewing Advice from the Other Side of the Table


Icon Saguaro


Icon Learned Motion Matching


Icon Why Can't I Reproduce Their Results?


Icon Latinendian vs Arabendian


Icon Machine Learning, Kolmogorov Complexity, and Squishy Bunnies


Icon Subspace Neural Physics: Fast Data-Driven Interactive Simulation


Icon Software for Rent


Icon Naraleian Caterpillars


Icon The Scientific Method is a Virus


Icon Local Minima, Saddle Points, and Plateaus


Icon Robust Solving of Optical Motion Capture Data by Denoising


Icon Simple Concurrency in Python


Icon The Software Thief


Icon ASCII : A Love Letter


Icon My Neural Network isn't working! What should I do?


Icon Phase-Functioned Neural Networks for Character Control


Icon 17 Line Markov Chain


Icon 14 Character Random Number Generator


Icon Simple Two Joint IK


Icon Generating Icons with Pixel Sorting


Icon Neural Network Ambient Occlusion


Icon Three Short Stories about the East Coast Main Line


Icon The New Alphabet


Icon "The Color Munifni Exists"


Icon A Deep Learning Framework For Character Motion Synthesis and Editing


Icon The Halting Problem and The Moral Arbitrator


Icon The Witness


Icon Four Seasons Crisp Omelette


Icon At the Bottom of the Elevator


Icon Tracing Functions in Python


Icon Still Things and Moving Things


Icon water.cpp


Icon Making Poetry in Piet


Icon Learning Motion Manifolds with Convolutional Autoencoders


Icon Learning an Inverse Rig Mapping for Character Animation


Icon Infinity Doesn't Exist


Icon Polyconf


Icon Raleigh


Icon The Skagerrak


Icon Printing a Stack Trace with MinGW


Icon The Border Pines


Icon You could have invented Parser Combinators


Icon Ready for the Fight


Icon Earthbound


Icon Turing Drawings


Icon Lost Child Announcement


Icon Shelter


Icon Data Science, how hard can it be?


Icon Denki Furo


Icon In Defence of the Unitype


Icon Maya Velocity Node


Icon Sandy Denny


Icon What type of Machine is the C Preprocessor?


Icon Which AI is more human?


Icon Gone Home


Icon Thoughts on Japan


Icon Can Computers Think?


Icon Counting Sheep & Infinity


Icon How Nature Builds Computers


Icon Painkillers


Icon Correct Box Sphere Intersection


Icon Avoiding Shader Conditionals


Icon Writing Portable OpenGL


Icon The Only Cable Car in Ireland


Icon Is the C Preprocessor Turing Complete?


Icon The aesthetics of code


Icon Issues with SDL on iOS and Android


Icon How I learned to stop worrying and love statistics


Icon PyMark


Icon AutoC Tools


Icon Scripting xNormal with Python


Icon Six Myths About Ray Tracing


Icon The Web Giants Will Fall


Icon PyAutoC


Icon The Pirate Song


Icon Dear Esther


Icon Unsharp Anti Aliasing


Icon The First Boy


Icon Parallel programming isn't hard, optimisation is.


Icon Skyrim


Icon Recognizing a language is solving a problem


Icon Could an animal learn to program?




Icon Pure Depth SSAO


Icon Synchronized in Python


Icon 3d Printing


Icon Real Time Graphics is Virtual Reality


Icon Painting Style Renderer


Icon A very hard problem


Icon Indie Development vs Modding


Icon Corange


Icon 3ds Max PLY Exporter


Icon A Case for the Technical Artist


Icon Enums


Icon Scorpions have won evolution


Icon Dirt and Ashes


Icon Lazy Python


Icon Subdivision Modelling


Icon The Owl


Icon Mouse Traps


Icon Updated Art Reel


Icon Tech Reel


Icon Graphics Aren't the Enemy


Icon On Being A Games Artist


Icon The Bluebird


Icon Everything2


Icon Duck Engine


Icon Boarding Preview


Icon Sailing Preview


Icon Exodus Village Flyover


Icon Art Reel




Icon One Cat Just Leads To Another


Created on April 13, 2012, 2:01 p.m.

Developing some Python bindings for my pet game engine I got kind of sidetracked and over the last few days created what has become my new baby. That is PyAutoC - a python library that helps automatically wrap C functions and structs at run-time to be accessible to the Python/C API. The power in this approach over something like SWIG is that it allows for developers to register new functionality for your module/API at run-time.

This kind of behaviour is essential for what I want for my Corange Python bindings because I want users to be able to register new asset, entity and ui_elem types in C - which all then just work out of the box with the existing C API. For a bit of an introduction here are some details copied over from the GitHub page:



PyAutoC automatically wraps C functions and structs at runtime so that they can be called from the Python/C API.

  • Don't fancy the idea of hand wrapping every function and struct in your codebase?
  • Don't like the look of the monster that is SWIG?
  • Want a way for developers to register extra functionality at runtime?

PyAutoC is here to help.


Basic Function Usage

#include "Python.h"
#include "PyAutoC.h"

static float add_numbers(int first, float second) {
    return first + second;

int main(int argc, char **argv) {


    PyAutoFunction_RegisterArgs2(add_numbers, float, int, float);

    PyObject* args = Py_BuildValue("(if)", 5, 6.13);
    PyObject* result = PyAutoFunction_Call(add_numbers, args);
    PyObject_Print(result, stdout, 0);
    Py_DECREF(result); Py_DECREF(args);


    return 0;

PyAutoC will call add_numbers with values converted from their Python counterparts. It will then convert the return value back into a Python Object. No editing of the original function required.


Basic Struct Usage

#include "Python.h"
#include "PyAutoC.h"

typedef struct {
    float x, y, z;
} vector3;

int main(int argc, char **argv) {


    PyAutoStruct_RegisterMember(vector3, x, float);
    PyAutoStruct_RegisterMember(vector3, y, float);
    PyAutoStruct_RegisterMember(vector3, z, float);

    vector3 position = {1.0f, 2.11f, 3.16f};
    PyObject* y = PyAutoStruct_Get(vector3, &position, y);
    PyObject_Print(y, stdout, 0);


    return 0;

Structs work similarly to their functional counterparts. They can be accessed at runtime and do automatic conversion of types.


New Argument Types

To call functions or access struct members which have non-primitive types it is possible to register your own conversion functions.

typedef struct {
    int x, y;
} pair;

static PyObject* convert_from_pair(void* data) {
    pair p = *(pair*)data;
    return Py_BuildValue("(ii)", p.x, p.y);

static void convert_to_pair(PyObject* pyobj, void* out) {
    pair* p = (pair*)out;
    p->x = PyInt_AsLong(PyTuple_GetItem(pyobj, 0));
    p->y = PyInt_AsLong(PyTuple_GetItem(pyobj, 1));

PyAutoConvert_Register(pair, convert_from_pair, convert_to_pair);

Now it is possible to call any functions with pair as an argument or return type and PyAutoC will handle any conversions automatically. You can also use the registered functions directly in your code by using PyAutoConvert.

pair p = {1, 2};
PyObject* pypair = PyAutoConvert_From(pair, &p);

Extended Usage 1

You can use PyAutoC to very quickly and easily create Python C modules for a bunch of functions such as might be done via SWIG or similar.

#include "Python.h"
#include "PyAutoC.h"

static float add_numbers(int first, float second) {
    return first + second;

static void hello_world(char* person) {
    printf("Hello %s!", person);

static PyObject* call(PyObject* unused, PyObject* args) {
    PyObject* func = PyTuple_GetItem(args, 0);
    PyObject* fargs = PyTuple_GetSlice(args, 1, PyTuple_Size(args));
    PyObject* ret = PyAutoFunction_CallByName(PyString_AsString(func), fargs);
    return ret;

static PyMethodDef method_table[] = {
    {"call", call, METH_VARARGS, ""},
    {NULL, NULL, 0, NULL},

PyMODINIT_FUNC initpyautoc_demo(void) {

    PyAutoFunction_RegisterArgs2(add_numbers, float, int, float);   
    PyAutoFunction_RegisterArgs1Void(hello_world, void, char*); 

    Py_InitModule("pyautoc_demo", method_table);

Then in Python...

import pyautoc_demo
pyautoc_demo.call("add_numbers", 5, 6.13);
pyautoc_demo.call("hello_world", "Daniel");

Once you have this basic interface it is easy to intergrate more complicated and transparent APIs with some more complicated Python. For Example...


Extended Usage 2

Because PyAutoC lets you register and access new Python functions at run time it is perfect for automatically wrapping existing C Structs as Python classes. By overriding __getattr__ and __setattr__ of a class we can easily make a Python object that behaves as if it were a C struct.

import birdie

class python_birdie(object):
    def __getattr__(self, attr): return birdie.get_attr(self, attr)
    def __setattr__(self, attr, val): return birdie.set_attr(self, attr, val)

bird = python_birdie()
print bird.name
print bird.num_wings

Then in the C API you can write something that looks like this.

typedef struct {
  char* name;
  int num_wings;
} birdie;

static PyObject* birdie_get_attr(PyObject* unused, PyObject* args) {
    PyObject* self = PyTuple_GetItem(args, 0);
    PyObject* attr = PyTuple_GetItem(args, 1);

    birdie* bird = get_instance_ptr(self);
    return PyAutoStruct_GetMember(birdie, bird, PyString_AsString(attr));

static PyObject* birdie_set_attr(PyObject* unused, PyObject* args) {
    PyObject* self = PyTuple_GetItem(args, 0);
    PyObject* attr = PyTuple_GetItem(args, 1);
    PyObject* val = PyTuple_GetItem(args, 2);

    birdie* bird = get_instance_ptr(self);
    return PyAutoStruct_SetMember(birdie, bird, PyString_AsString(attr), val);

static PyMethodDef method_table[] = {
    {"get_attr", birdie_get_attr, METH_VARARGS, ""},
    {"set_attr", birdie_set_attr, METH_VARARGS, ""},
    {NULL, NULL, 0, NULL},

PyAutoStruct_RegisterMember(birdie, name, char*);
PyAutoStruct_RegisterMember(birdie, num_wings, int);

Py_InitModule("birdie", method_table);

A lot less work than writing a bunch of getters and setters!

The get_instance_ptr function is left for the user to implement and there are lots of options. The idea is that somehow the python instance should tell you how to get a pointer to the actual struct instance in C which it represents. One option is to store C pointers in the python instance using something like PyCObject_FromVoidPtr. An alternative I like is to just store a string in the python instance which uniquely identifies it. Once you have this, in C it is possible to just look this string up in a dictionary or similar to find the actual pointer.

For fun why not try also overriding __init__ and __del__ to call some C functions which allocate and decallocate the structure you are emulating, storing some data to let you identify the instance later. It is also easy to extend the above technique so that, as well as members, the class is able to look up and execute methods!

The true power of PyAutoC comes if you look a level deeper. If you use PyAutoStruct_GetMember_TypeId or PyAutoStruct_SetMember_TypeId you can even extend the above code to work for arbritary structs/classes which developers can add to.

For this to work you need to somehow get a PyAutoType value. This can be found by feeding a string into PyAutoType_Register. The PyAutoType_Register function is a simple function which gives a unique identifier to each new string it encounters. Run time type-ids are generated for types in this way - the macro preprocessor just turns the token into a string and feeds it into this function. This means that if you give it a string of a previously registered data type it will return a matching Id. One trick I like it to use is to feed into it the .__class__.__name__ property of a python instance. This means that I can create a new python class with overwritten __getattr__ and __setattr__ it will automatically act like the corrisponding C struct with the same name - providing it has been registered by some developer.


Managing Behaviour

Often in C, the same types can have different meanings. For example an int* could either mean that a function wants an array of integers or that it outputs some integer. We can change the behaviour of PyAutoC without changing the function signature by using typedefs and new conversion functions. Then on function registration you just use the newly defined type rather than the old one. Providing the types are truely the same there wont be any problems with converting types or breaking the artificial stack.

static void print_int_list(int* list, int num_ints) {
    for(int i = 0; i < num_ints; i++) {
        printf("Int %i: %i\n", i, list[i]);

typedef int* int_list;

static int list_space[512];
static void convert_to_int_list(PyObject* pyobj, void* out) {
    for(int i = 0; i < PyList_Size(pyobj); i++) {
        list_space[i] = PyInt_AsLong(PyList_GetItem(pyobj, i));
    *(int_list*)out = list_space;

PyAutoConvert_RegisterTo(int_list, convert_to_int_list);

PyAutoFunction_RegisterArgs2Void(print_int_list, void, int_list, int);

As you can probably see, automatic wrapping and type conversion becomes hard when memory management and pointers are involved. I'm looking at ways to improve this, perhaps with the ability to register 'before' and 'after' methods for certain functions or conversions.



  • The handling of errors is done via Python Exceptions. Checking for them in C is important or you'll probably end up with a segfault or garbage values.

  • I've not yet managed to get PyAutoC to compile using MSVS and due to the heavy macro use and lack of C99 I don't know if I can. Until I look into it more, compile any extensions under windows with MinGW and it will probably be okay.

  • The function registration macros are a little verbose and limited though this is unavoidable. Always remember that the argument count must be specified in the name and also if the function returns void.

PyAutoFunction_RegisterArgs2(add_numbers, float, int, float);
PyAutoFunction_RegisterArgs3Void(add_numbers_message, void, char*, int, float);
  • Using PyAutoC for functions creates a small memory and performance overhead. This is because it duplicates much of the process involved in managing the stack such as copying stack data. Because most of the logic happens at run-time it also uses a lot of function pointers. These cannot be optimised and inlined very easily so processes such as converting lots of Python data to C data can be slower than if the process is declared statically. Still, the overhead is fairly minimal and if you are wrapping with a scripting language like python then perhaps it is less of a concern.


github twitter rss