Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!rutgers!mcnc!ece-csc!ncrcae!ncr-sd!hp-sdd!hplabs!ucbvax!CS.UCLA.EDU!bilbo.geoff
From: bilbo.geoff@CS.UCLA.EDU (Geoff Kuenning)
Newsgroups: comp.windows.x
Subject: Toplevel widget's geometry manager
Message-ID: <8711030020.AA11296@lanai.cs.ucla.edu>
Date: Mon, 2-Nov-87 19:20:23 EST
Article-I.D.: lanai.8711030020.AA11296
Posted: Mon Nov  2 19:20:23 1987
Date-Received: Sun, 8-Nov-87 11:21:31 EST
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 31


    Okay, so I played around with the new toolkit and figured out how
to write a new widget.  The new toolkit's pretty neat, though somewhat
limited by the realities of the C language.  But now I want to put a bunch
of widgets (e.g., menus) around the borders of a window, the center of
which is graphics.

    The proper solution to this, of course, is to write a handy composite
widget with geometry manager that magically handles the layout of the
various widgets exactly the way I want.  This is nontrivial, of course.
A simpler solution is just to hardwire widget positions into some composite
widget that has a dumber geometry manager.

    It though of using the Toplevel widget as the composite widget.
However, looking at the Toplevel widget's code shows that it can really
only handle a single subwidget, and it restricts that widget's geometry
fairly strongly.

    I think that Toplevel should have a very cooperative geometry manager.
It should always endeavor to do exactly what a child asks with a minimum
of changes and a minimum of features.  x/y/width/height requests should
be obeyed verbatim;  the only special processing is that the Toplevel
window should dynamically expand and shrink to reflect the largest extents
of any of its children.  This will make it easy to decorate a test
window with widgets during early development.

    I have written a widget, called "Obedient," which implements such
a scheme.  It's pretty easy.  What do other people think about this
approach?

	Geoff Kuenning	geoff@lcc.ucla.edu	geoff@ITcorp.com