Re: stk::rectangle

Dunno, my solution would be to make two rectangle classes, one for 
objects which provide a rectangle like interface (and this class should 
have an absolute minimum of virtual functions)
And a second class which is non-virtual (POD!) for usage in rendering 
etc, this is basicly where the vfuncs slow us extremely down atm!
Darren Vincent Hart wrote:
> I am beginning to feel the same way.  I had to wrap most of the rect
> functions to add the on_resize signal anyway.  Perhaps we should revert
> to the "has a" model and remove all the virtual crap from rectangle. 
> Let's discuss this thoroughly before we make any changes to the source
> though, this needs to be the last time we change how rects and widgets
> relate.
> 
> Darren
> 
> On Sun, 2003-10-19 at 05:29, Marc Strämke wrote:
> 
>>Hi team,
>>I have just looked at the changes darren commited today, and have found 
>>out that rectangle contains virtual functions, do you guys really think 
>>thats good? Imo we should create rectangle_interface class or something 
>>to derive from, and not use the regular rectangle. These changes make 
>>rectangle easyli a few 100 bytes(!!!) Big, this is really not the right 
>>thing to do for an object which the user expects to be lightweight is 
>>it? Imo we only got two options, either we stop deriving from rectangle, 
>>or we create two rectangle classes, one which serves as an interface and 
>>from which you can derive, so it behaves like a rectangle but isnt one, 
>>but we really shouldnt continue with using a struct full of virtual 
>>functions everywhere in the code, this is a major source of bloat!
>>
>>
>>
>>_______________________________________________
>>Libstk mailing list
>>Libstk@Óq«LD—"
>>http://www.dvhart.com/cgi-bin/mailman/listinfo/libstk
> 
>