RE: [gtkmm] Adding more C++ power to Gtkmm

[Murray Cumming Comneon com]

Please mention just one subject per email. Big multi-subject emails generate
big multi-subject threads that are hard to follow. Most of this should be in
bugzilla anyway.

> From: r ve [mailto:kaas_10 hotmail com] 
> Some time ago, I've been complaining that Gtkmm does things 
> too much C-like.

Personally I don't remember. I don't think that's true in general, unless
you are talking about the GTK+ implementation. gtkmm is generally thought of
as the most C++-like of C++ GUI APIs.

> Because things won't change when keeping my mouth shut, here's a few 
> suggestions that I would really like to see added in Gtkmm.
> 
> 
> 
> One example of something 'C-like' in Gtkmm is that some 
> methods contain far 
> too many arguments.
> 
> Here's a method with 16 (!!) arguments:
> 
> - Gdk::Pixbuf::composite_color(const 
> Glib::RefPtr<Gdk::Pixbuf>& dest, int 
> dest_x, int dest_y, int dest_width, int dest_height, double 
> offset_x, double 
> offset_y, double scale_x, double scale_y, InterpType interp_type, int 
> overall_alpha, int check_x, int check_y, int check_size, 
> guint32 color1, 
> guint32 color2)
> 
> These arguments could be reduced to 13 arguments by rewriting 
> it to this:
> 
> - Gdk::Pixbuf::composite_color(const Glib::RefPtr<Gdk::Pixbuf& dest, 
> Gdk::Rectangle& dest_rect, double offset_x, double offset_y, 
> double scale_x, 
> double scale_y, InterpType interp_type, int overall_alpha, 
> int check_x, int 
> check_y, int check_size, guint32 color1, guint32 color2)

Sure. Please submit a patch for gtkmm 2.4 - see the instructions for this on
the website.

> It could be reduced to 10 arguments by modifying the 
> Gdk::Point class to 
> become a template (would be useful for a few other classes as well):

That would be interesting. I would like to see that as a separate patch.
 
> - Gdk::Pixbuf::composite_color(const Glib::RefPtr<Gdk::Pixbuf& dest, 
> Gdk::Rectangle& dest_rect, Gdk::Point<double>& offset, 
> Gdk::Point<double>& 
> scale, InterpType interp_type, int overall_alpha, Gdk::Point<int>& 
> check_offset, int check_size, guint32 color1, guint32 color2)
> 
> Well, I think you get the point :)
> 10 arguments is still quite a lot imho, but it's much better 
> than 16 (!) 
> arguments.
> 
> 
> 
> Other things that could be done to improve Gtkmm is to add 
> more overloaded 
> operators.
> 
> The Gdk::Point class, for example, does only overload the == and != 
> operators, but this class could be much more useful if it had 
> operators for 
> adding, subtracting and multiplying Gdk::Points.
> On top of that, the Gdk::Point class doesn't have a 
> Point(const Point& rhs) 
> constructor, which can also prove to be useful at some times.

Again,  please provide a patch.

> The Gdk::Rectangle class doesn't overload _any_ operator, 
> that's bad because 
> it's a pain to write something like this:
> 
> if ((rect1->get_x() != rect2->get_x()) && (rect1->get_y() != 
> rect2->get_y()) 
> && (rect1->get_width() != rect2->get_width() && 
> (rect1->get_height() != 
> rect2->get_height()) bleh;
> 
> Instead of:
> 
> if (rect1 != rect2) bleh;
> 
> Which is not only much easier on the eyes, but faster to type as well.

Again,  please provide a patch.

> Also, it would be really nice if virtual methods (signals) 
> wouldn't be like 
> this (the C way):
> 
> virtual void on_set_focus(Widget* focus);   // set_focus signal for 
> Gtk::Window
> 
> But rather like this (the C++ way):
> 
> virtual void on_set_focus(Gtk::Widget* focus);

I don't see what's different there.

>  (or const 
> Gtk::Widget& 
> focus.. whatever)

It's possible that focus might be null sometimes (I don't know) so that
might not work. Feel free to investigate and provide a patch.
 
> And same goes for all other signals.

Well it's still not clear what you mean.

Murray Cumming
murrayc usa net
www.murrayc.com