RE: object-oriented programming in Octave

[Paul Billings]

I think this would be wonderful.  My reason for posting is to say that I
hope it is better than Matlab's implementation.  Their OO support is
rudimentary at best, given the lack of references.  The problem enters when
trying to share something between two other objects:
  r = common_obj(...);
  o1 = myobj(r, ...);
  o2 = myobj(r, ...);

Guess what?  O1 and O2 have separate r-objects.  If o1 modifies r (or asks
its r to change state), it is no longer the same as o2.r.  Any sort of
doc/view (subject/observer) pattern is impossible.

Secondly, you cannot access base class fields from subclasses, which
requires you to implement get/set hooks for access.  This either exposes
variables that shouldn't be exposed, or you have a "private" get/set
function (with a different name) and have to remember which variables are
private.

Paul


> -----Original Message-----
> From: Stefan van der Walt [mailto:address@hidden
> Sent: Saturday, January 07, 2006 12:24 PM
> To: address@hidden
> Subject: Re: object-oriented programming in Octave
>
>
> On Sun, Jan 01, 2006 at 09:49:57PM +0100, Arvid Rosén wrote:
> > Michael Schmid wrote:
> > >my next question:
> > >
> > >are there any future plans to introduce the "@directorys" like
> in Matlab?
> > >This directorys defines the object-oriented programming technology.
> > >If not, I will change this parts of the neural network toolbox,
> > >otherwise I will
> > >wait!
> > >Thanks
> > >
> > >Michael
> > >
> > That would be great!
> > Object-oriented programming is very useful for this kind of work. This
> > is also the main reason why we favor Scilab and not Octave at
> my office.
> > Scilab has a easy-to-use and very powerfull way of making
> > object-oriented programs. It would be a great addition to Octave.
>
> I saw no further responses on this topic.  Is this anywhere on the
> TODO-list for 3.0?
>
> Regards
> Stéfan