|
|
|
|
| Re: widget objects [message #38840 is a reply to message #19152] |
Wed, 31 March 2004 12:48   |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Tue, 30 Mar 2004 15:56:25 -0500, Michael A. Miller wrote:
> I've got some questions regarding objects and widgets and am
> hoping that someone here can help point me toward a robust
> design. Here's the situation...
>
> I have a collection of IDL objects that include GUI components.
> I use these in two ways: by themselves in their own top level
> windows and swallowed in widget_bases of other GUIS, like a
> compound widget. This works pretty well for me, but has some
> limitations. For example, I have a color map object that allows
> me to select a color map and adjust the limits. It inherits from
> a publisher class so other objects that might use it can
> subscribe. When the color map is changed, it tells its
> subscribers. Typically a subscriber will redraw something using
> the color map when it is notified of a change.
>
> Now I've decided that in some cases, it's is ok to have a color
> map GUI on the screen (as part of a main window or in its own
> window) all the time. In other cases, I'd really like to put the
> color map in its own window which can be drawn and withdrawn at
> will, like with tk's withdrawn window state. I don't believe
> that this is possible with IDL, so instead I thought I'd separate
> the color map object from the gui so that an instance can create
> and activate the gui and destroy it at will (at the programmers
> will, that is) without destroying the entire object.
>
> This sounds fairly straight forward, but I'm confident that I can
> make it complex enough to be cumbersome. I wonder if any of yous
> idlers have given this sort of thing some thought and have ideas
> that you'd be willing to share or just discuss in general. The
> part in particular that I'm wondering about is the capability of
> putting a gui component in it's own top level base or in another
> window. I've implemented this with a widget object class which
> is appended below. Using it just seems a bit cumbersome to me
> and I can't quiet put my finger on why.
I do this without any difficulty. I usually have a separate method
call which actually contains the widget definition code and
(optionally) XManager call. I call it when necessary, and provide a
means to test if the GUI is running (XRegistered is convenient, as is
widget_info(/VALID_ID)). To have it function either standalone or as
a compound widget, pass an optional parent widget ID. If the parent
ID is passed and is valid, build all widgets beneath it, and don't
call XManager (since the parent application will do that). If it's
not passed, make your own TLB (perhaps MODAL) and proceed the same
way, calling XManager this time to manage your own events. In either
case, the events will pass through your own event handler (which,
since you can't rely on XManager to set for you, you should do using
WIDGET_CONTROL on the necessary top-level widget). The widget object
doesn't know whether it's running *inside* another widget or as a
stand-alone. Nor does it care whether it actually has any widgets
realized and showing. The only potential difficulty appears if you
want to pass events "up" from your widget-object-as-compound-widget to
be handled directly in the parent app's event handler. You can do
away with this difficulty is you simply stick to your
publish/subscribe based communication in all cases, and swallow all
events (keeping them internal to your object's widget hierarchy).
This methodology will pay off in the long run, since then you can
reuse the same widget object in a variety of ways, and its
communication pathways remain flexible.
To have, e.g., a colormap widget set appear and disappear within
another base, you can just wrap the widget building code in
widget_control, self.parent,UPDATE=0, widget_control,
self.parent,UPDATE=1 and destroy it when it's not popped up. Since
the object lives through this process, there is no problem maintaining
state: it simply appears that the widget disappears and reappears on
command. Except for the heaviest of widgets, this is acceptably fast.
JD
|
|
|
|
|
|
| Re: widget objects [message #38844 is a reply to message #19152] |
Wed, 31 March 2004 12:23 |
MKatz843
Messages: 98
Registered: March 2002
|
Member |
|
|
Regarding object widgets, a fun topic for discussion if there ever
were.
I was rummaging through IDL's libraries and came across this object
class. (Your path may be somewhat different)
/Applications/idl_6.0/lib/utilities/idlexwidget__define.pro
Here's the top of the file's header. This would appear to be an RSI
object cw_widget prototype. I just through it might be of interest to
the group to know that this is there.
-M. Katz
;$Id: idlexwidget__define.pro,v 1.7 2003/02/03 18:14:12 scottm Exp $
;
; Copyright (c) 1997-2003, Research Systems, Inc. All rights
reserved.
; Unauthorized reproduction prohibited.
;
; NAME:
; IDLexWidget
;
; PURPOSE:
; Provide a base class for defining objects that function
; as IDL compound widgets.
;
; CATEGORY:
; Widgets. Object Oriented Programming.
;
; CREATION:
; It is intended that subclasses of class IDLexWidget, rather than
; class "IDLexWidget" itself, get instantiated. It is possible to
; create a "IDLexWidget" object directly...
;
; oWidget = OBJ_NEW('IDLexWidget', wParent)
;
; ... but such objects would not have any buttons, sliders, etc.
; in them, and would have a propensity to throw errors.
;
|
|
|
|
| Re: widget objects [message #38910 is a reply to message #38844] |
Mon, 05 April 2004 14:18 |
Paul Sorenson
Messages: 48
Registered: May 2002
|
Member |
|
|
My visio diagram includes this class. The diagram is at:
http://www.paulsorenson.com/underthehood.html
-Paul
"M. Katz" <MKatz843@onebox.com> wrote in message
news:4a097d6a.0403311223.462cd994@posting.google.com...
> Regarding object widgets, a fun topic for discussion if there ever
> were.
> I was rummaging through IDL's libraries and came across this object
> class. (Your path may be somewhat different)
>
> /Applications/idl_6.0/lib/utilities/idlexwidget__define.pro
>
> Here's the top of the file's header. This would appear to be an RSI
> object cw_widget prototype. I just through it might be of interest to
> the group to know that this is there.
> -M. Katz
>
> ;$Id: idlexwidget__define.pro,v 1.7 2003/02/03 18:14:12 scottm Exp $
> ;
> ; Copyright (c) 1997-2003, Research Systems, Inc. All rights
> reserved.
> ; Unauthorized reproduction prohibited.
> ;
> ; NAME:
> ; IDLexWidget
> ;
> ; PURPOSE:
> ; Provide a base class for defining objects that function
> ; as IDL compound widgets.
> ;
> ; CATEGORY:
> ; Widgets. Object Oriented Programming.
> ;
> ; CREATION:
> ; It is intended that subclasses of class IDLexWidget, rather than
> ; class "IDLexWidget" itself, get instantiated. It is possible to
> ; create a "IDLexWidget" object directly...
> ;
> ; oWidget = OBJ_NEW('IDLexWidget', wParent)
> ;
> ; ... but such objects would not have any buttons, sliders, etc.
> ; in them, and would have a propensity to throw errors.
> ;
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
