[C++-sig] Re: Iterators + shared_ptr + Abstract base classes + Pass by reference = help :)

[David Sankel]

> David Sankel <camio at yahoo.com> writes:
> 
> > Hello,
> >   I have several questions and I think they are
> all
> > most easily answered by example:
> >
> > class Abstract
> > {
> > public:
> >   shared_ptr<Abstract> clone()=0;
> > };
> >
> > class Filter
> > {
> > public:
> >   vector<shared_ptr<Abstract> > data;
> >   void doFilter( vector<shared_ptr<Abstract> >
> &result )=0;
> > };
> >
> > How would I use boost.python to allow python code
> to
> > use Filter as a base class?
> >
> > I figure there is going to be:
> >
> > class PythonFilter : public Filter
> > {
> > public:
> >     void doFilter( vector<shared_ptr<Abstract> >&
> result)
> >     {
> >         //What goes here?
> 
> Here you need to invoke call_method<void>(...) with
> some argument(s). 
> You could write:
> 
>     call_method<void>( ref(result) );
> 
> But that will only work if you've exposed 
> 
>     vector<shared_ptr<Abstract> >
> 
> to Python.  My guess is that you haven't.
> 
> What type would you like your Python derived class
> to see when you
> override doFilter in Python?

I would like the Python derived class to see a list of
Abstracts.  I guess this is where I don't understand
how to expose vector<shared_ptr<Abstract> > with the
iterators.

> 
> >     }
> > }
> >
> > and:
> >
> > BOOST_PYTHON_MODULE(embedded_Filter)
> > {
> >     class_< Abstract > ("Abstract", no_init );
> >     class_< Filter,
> >             PythonFilter >("Filter")
> >     //What goes here?
> 
> So far, nothing.  You don't need to (and shouldn't)
> expose a pure
> virtual function to Python.

Wouldn't I need something so a python derived class
could access the Filter::data member?

Also, how would I allow python to see the
Abstract::clone() function?

> >     ;
> >
> > }
> >
> > The python code to look something like this:
> >
> > import ...
> > class SpecificFilter( Filter ):
> >   def doFilter( self, result )
> >      #make result some modification of self.data
> 
> I don't understand the comment above.
> 

For example:
class specificFiler( Filter ):
  def doFilter( self, result ):
    result = [x.clone() for x in self.data]
    result.reverse()

Does this make more sense?

Thanks,

David J. Sankel