passing a C struct to Python and back

[Kevin]

Thanks Alex... I guess I gave up too easily.

I got it working "as-is" with a better cast:

theProcs = *((KPDCProcs *) (PyCObject_AsVoidPtr(pyTheProcs)));

Kevin.


"Alex Martelli" <aleax at aleax.it> wrote in message
news:KazHa.100629$pR3.2188842 at news1.tin.it...
> <posted & mailed>
>
> Kevin wrote:
>
> > In a Python extension module, I have a struct that I need to pass to
> > Python,
> > then back into the extension module later for use.  Python doesn't need
to
> > do anything with the struct in the meantime though.
> >
> > I'm trying to do it using a CObject through the following, but
> > PyCObject_AsVoidPtr() returns a void*, not a struct...
>
> Right, a void*, i.e. "a pointer to anything".  In particular, therefore,
> it can be a pointer to the struct you want to pass back and forth (it
> can live on the heap or in statical-lifetime memory -- not on the C stack,
> obviously).
>
> > Is there an easy way to do this?  Creating a new Python type seems like
a
> > big pain in the butt for something as "simple" as this.
>
> There are many easy ways.  One is via a void*, as above.  Another is to
> pass back and forth a string of bytes instead --
PyString_FromStringAndSize
> to build the string object, PyString_AsString to get back the pointer to
> the buffer area (as a char* to be cast) -- you shouldn't use this one if
> you want your C code to modify the struct after it's been passed to Python
> (strings should be immutable!).  Returning a tuple of the struct's fields
> is another possibility (tuples are immutable, too), requiring more work to
> pack and unpack (wasted work if Python will never need to access the
> individual fields).
>
> You could also use a "buffer object", as documented at URL
> http://www.python.org/doc/current/api/bufferObjects.html -- not a very
> fashionable solution nowadays but perhaps the simplest q&d fix to your
> problem.  PyBuffer_FromReadWriteMemory to build (in order to use the
> struct's own memory for the buffer -- this needs a struct that won't just
> go away, e.g. by living on the stack;-) or PyBuffer_New if the buffer
> must maintain its own memory -- and you can use the PyBufferProcs (see
> http://www.python.org/doc/current/api/buffer-structs.html) to get the
> pointer to the buffer's start (e.g. to memcpy the struct at start, &tc).
>
> As I said -- plenty of choices...
>
>
> Alex
>