Re: Subclassing Gtk2::CellRenderer (reprise)

[muppet <scott asofyet org>]

On Wednesday, October 15, 2003, at 09:50 AM, muppet wrote:

> i will post more info about this, and some code, later today.

everything you need is at http://asofyet.org/muppet/software/gtk2-perl/
  Gtk2-CellRenderer-Base-0.03.tar.gz

this requires that you patch Glib and Gtk2:
  Glib-1.00-type_utils.patch
  Gtk2-1.00-get_child_requisition.patch
  Gtk2-1.00-style_ornull.patch

for those working against current cvs, there's also
  cellrendererstuff.patch


the archive includes both of Tortsten's way cool renderers, my original 
hack at the popup renderer (now new and improved with y'all's patches), 
and a new hack, fancy.pl.

fancy.pl is very ugly, but illustrates deriving directly from 
Gtk2::CellRendererText via the new tricks that 0.03 brings in.  what 
are those tricks?

  Gtk2::CellRenderer::register_subclass (parent_package, new_package, 
...)

does something very similar to Glib::Type::register, but requires that 
parent_package descends from GtkCellRenderer.  the varargs stuff is the 
same as what you'd pass to Glib::Type::register or the use args of 
Glib::Object::Subclass.

therefore:

   # derive a new object, named ReallyCool, from Gtk2::CellRendererText.
   # the resulting object has all of the properties of a text renderer,
   # plus the new one 'nlines'.  for each event, a perl sub will be 
called.
   Gtk2::CellRendererText->register_subclass ('ReallyCool',
           properties => [
                 Glib::ParamSpec->int ('nlines', 'number of lines',
                                      'the number of lines you want in 
the cell',
                                      1, 65535, 1, [qw/readable 
writable/]),
           ]);

   # implement the get_size() virtual method:
   sub on_get_size {
          # we're just extending the parent.  let's call it first,
          # and see what it wants.
          my @results = shift->parent_on_get_size (@_);
          ...
   }

all of this is subject to change, i have designs already on how the 
chaining may be better done; as it is currently implemented, it is very 
likely not to work on multiply-derived renderers.

in case you hadn't noticed, this method allows us to avoid the use of 
Gtk2::CellRenderer::Base altogether (well, it's used under the hood, 
but), which i think is somewhat cleaner.

comments?  thoughts?  suggestions?


--
muppet <scott at asofyet dot org>