extending with C

[Jesper Olsen]

Thanks for all the replies - they were useful.

I am not sure if the "dirty" tupel example would work directly;
the PyTuple_GET_ITEM() function
> 		floatlist[loop] = PyTuple_GET_ITEM(tuple, loop);
returns a PyObject, which I think is not a float, but rather a pointer to 
a structure containing a reference count and in this case a pointer to
a float?
So it would be necessary to explicitly extract the float from the float object:

   float f=PyFloat_AsFloat(PyObject *pyfloat) 

Anyway, in the end I chose the array solution, which I think is quite
elegant:


static PyObject * get_array_of_float
(
    PyObject *self,
    PyObject *args
)
{
    float* farray=get_float_array();

    return Py_BuildValue("s#", (char*) farray, SIZE*sizeof(float));
} /* get_array_of_float */


static PyObject * pass_float_array
(
    PyObject *self,
    PyObject *args
)
{
    char* fbuf;
    float* farray;
    int i;
    
    if(!PyArg_ParseTuple(args,"si", &fbuf, &frame_counter)) return NULL;
    farray=(float *) fbuf;
    for (i=0; i<SIZE; i++)
        printf("%f ", farray[i]);
    Py_INCREF(Py_None);
    return Py_None;
} /* pass_float_array */

>From python these functions can be called as:

str=get_array_of_float()
a=array("f")
a.fromstring(str)

and

pass_float_array(a.tostring())


I am not super enthusiastic about the python array type
- I have not cheked the implementation, but given the operations
on it, it can not map well to c-arrays.


Cheers
Jesper





Pete Shinners <pete at shinners.org> wrote in message news:<3C3C6E95.9020103 at shinners.org>...
> Jesper Olsen wrote:
> 
> > I need to write a python extension in C, and the method interface should
> > be able to handle "arrays of float".
> > 
> > In the python code, the array could be represented as a tupel, a list or 
> > an array it is not really important.
> > 
> > The extension needs to be able to receive such arrays, and to return them
> > to python.
> 
> 
> here's a quick dirty sample, reading a set of floats from inside a tuple...
> 
> 
> PyObject* do_floats(PyObject* self, PyObject* args)
> {
> 	PyObject* tuple;
> 	float* floatlist;
> 	int size, loop;
> 
> 	if(!PyArg_ParseTuple("O", &tuple))
> 		return NULL;
> 	if(!PyTuple_Check(tuple))
> 	{
> 		PyErr_SetString(PyExc_TypeError, "need tuple");
> 		return NULL;
> 	}
> 
> 	size = PyTuple_Size(tuple);
> 	floatlist = PyMem_New(float, size);
> 	if(!floatlist)
> 		return NULL;
> 	for(loop = 0; loop<size; ++loop)
> 		floatlist[loop] = PyTuple_GET_ITEM(tuple, loop);
> 
> 	/* DO SOMETHING WITH THE FLOATLIST HERE */
> 
> 	PyMem_Del(floatlist);
> 	Py_INCREF(Py_None);
> 	return Py_None;
> }
> 
> 
> of course, my email composer probably parses this better than a real c 
> compiler, so you might find some typos in there. :]