The problem is for a subpatch that isn't vis'd:

[inlet]
|
[checkbutton]
|
[outlet]

Will give an error on "bang" because checkbutton_bang has:
sys_vgui("%s invoke\n", x->widget_id->s_name);
That could be solved by using a -variable option, but then that limits you because the nonzero value cannot be changed.*

So yeah, the best solution is stored state in the struct so you can access it in the c functions without caring about the vis state of the obj.  Is that the idea behind the options_binbuf?

* also note that checkbutton's default behavior is the opposite from tgl:
tgl: zero = off, nonzero = on
checkbutton: onvalue = on, everything else = off

From: Hans-Christoph Steiner <hans@at.or.at>
To: Jonathan Wilkes <jancsika@yahoo.com>
Cc: pd-dev List <pd-dev@iem.at>
Sent: Wednesday, August 17, 2011 7:24 PM
Subject: Re: tkwidgets


On Aug 17, 2011, at 6:23 PM, Jonathan Wilkes wrote:

How does your private git branch differ from what's currently in svn?

I pushed my git to github, my latest work is in the 'newentry' branch.  I basically focused on the [entry] widget to see if I could get it going with this new approach.


One thing I'd like to point out is that your tkwidgets suffer from the same problem tot/widget did -- by handling all the widget state in tcl you make it impossible to use your objects inside a subpatch/abstraction that doesn't have a visible canvas (because the widget no longer exists).  I was considering create a "master" widget as a child of the main window and sync it to the widget drawn on the canvas, but that seems like a lot of trouble.  Is there some other workaround?

That is true, but tkwidgets are all about using Tk widgets in Pd as easily as possible.  If the Tk widget is not drawn to the screen, then the Tk widget is not involved.  What is the problem in particular?  You want to change its state while its not visible?  I think the idea there is to have the state stored in the standard struct as a dump from the Tk widget.  So when its not visible, store the state-changes, when it becomes visible, dump the state to the Tk widget.

Another thing: to get "Pd-style" interaction, bind $canvas <<EditMode>> to a proc that sets -state to disabled for all tkwidgets in that $canvas when editmode == 1.  That way you don't end up triggering the widget when you want to edit it.

Yeah, tkwidgets was mostly written about the same time as I starting the pd-gui-rewrite.  As I rewrote a lot of pd-gui, I realized I should wait on tkwidgets since I could fix a lot of things in pd-gui first.  The <<EditMode>> message is one example.  tkwdigets is not complete yet, so this is not fully in there.  I think the [entry] in github does this kind of stuff.

.hc

Hans-Christoph Steiner <hans@at.or.at>
To: Jonathan Wilkes <jancsika@yahoo.com>
Cc: pd-dev List <pd-dev@iem.at>
Sent: Wednesday, August 17, 2011 2:54 PM
Subject: Re: tkwidgets


Hey Jonathan,

I'm cc'ing pd-dev since this is a topic that could interest others and others could contribute to.  I've started a private git branch of tkwidgets that I intent to push once I get somewhere with it.  The idea is to try out a new idea for how GUI objects can work.  Basically, I think I can make it so that Tcl handles more of the interaction with the user, minimizing on pd-gui <--> pd communications, and making it easier to write GUI objects.  Its not trivial to do, but should be doable.

.hc


On Aug 17, 2011, at 12:26 PM, Jonathan Wilkes wrote:

Never mind, I see it now inside canvas_vis... too bad canvas' "window" subcommand doesn't have something like pack's "-in" option...

But I guess I could make a toplevel checkbutton widget and just manually clone it.

-Jonathan


From: Jonathan Wilkes <jancsika@yahoo.com>
To: Hans- Christoph Steiner <hans@at.or.at>
Sent: Wednesday, August 17, 2011 3:14 AM
Subject: tkwidgets

Hi Hans,
     Do I have it right that your tkwidgets get destroyed when the containing patch is vis'd 0?  If so, any hints on how this happens?

Specifically, I'm playing around with [checkbutton], and even if I comment out everything in eraseme and checkbutton_free, and every single "destroy" subcommand, I still get a tcl error when sending a bang or float to a [checkbutton] that's in a subpatch with no window mapped:

(Tcl) INVALID COMMAND NAME: invalid command name ".x252a690.c.widget25272b0"
    while executing
".x252a690.c.widget25272b0 cget -onvalue"
    ("uplevel" body line 2)
    invoked from within
"uplevel #0 $cmd_from_pd"







----------------------------------------------------------------------------

Man has survived hitherto because he was too ignorant to know how to realize his wishes.  Now that he can realize them, he must either change them, or perish.    -William Carlos Williams







----------------------------------------------------------------------------

"We have nothing to fear from love and commitment." - New York Senator Diane Savino, trying to convince the NY Senate to pass a gay marriage bill