[Boost-users] multi-type containers

Subject: [Boost-users] multi-type containers
From: László Marák (ujoimro_at_[hidden])
Date: 2009-04-16 07:11:13

Dear Boost Users,

I am getting started with boost.python. I want to port an image processing
library into python. My problem is, that I have several pixel types (like
int, float, etc.) which I'd love to handle in the same object by
downcasting. In c++ it's something like this:

class image {};

class float_image : image {
    float * pixels; // or boost::array_ptr<float>
}

class int_image : image {
    int * pixels; // or boost::array_ptr<int>
}

I can create the image object by:

image * A = new float_image()
image * B = new int_image()

Now I'd have a virtual functions

virtual float image::get_element()
virtual int   image::get_element()

obviously this won't pass in c++, but in python it should pose no
difficulty, as python uses dynamic casting. Is there a manner to do this
with c++/boost/python?

thank you for considering,

Laszlo