State of Observers Redux — 18 August 2011

[thomthom]

I've begun on a revision of the State of Observers:
http://www.thomthom.net/software/sketchup/observers/

StateOfObservers.png

Could you please have a look and let me know if have information that would fit in this.

(p.s. The page is in HTML5 - so if you use silly old browsers it might not look good - I've not tested it yes, just in Firefox 3.6. Atm I'm focusing on the content.)

---

Added public bug tracker to the BitBucket repo:
https://bitbucket.org/thomthom/sketchup ... tatus=open

People can add reports of broken observer events there.

[thomthom]

Was there an observer hidden in the API docs somewhere that was listed in between the methods of another class instead of a class of its own?

[Dan Rathbun]

Was there an observer hidden in the API docs somewhere that was listed in between the methods of another class instead of a class of its own?

Yes.. it is the FrameChangeObserver, see the Pages class.

I am trying to animate objects based on Scene Changes and not with the interact tool.

topic: Dynamic Components that react to scene (aka page) changes

in this post: Re: Using an observer to delete a screen note I said the following:

That's interesting. I've never noticed that [FrameChangeObserver observer] ...

Yes that is because is one of the API missing class boo-boos. It IS mentioned within the API, in passing.
Pages.add_frame_change_observer
Pages.remove_frame_change_observer

It was featured (documented) in the Sketchup API BlogSpot:
Dynamic Components that react to scene (aka page) changes

('scenes.rb') Code properly formatted (by Todd Burch) at:
http://www.smustard.com/forum/viewtopic.php?f=9&t=25

It basically only has one event callback:
frameChange( fromPage, toPage, percent_done ),
the intialize method, plus any custom method(s) you need inside to keep your code readable and organized.

~

[Dan Rathbun]

And I think Jim had doubts whether it worked during an animation export.

PS, the table page looks fine under MSIE 7.
"Depreciated" has no "i" in it. And the methods that were removed can not be "Deprecated", they are "Abandoned", "Deceased".. whatever they are no longer even in the Observer protoclass.
.

[Krisidious]

they should have you on staff Thom...

[Dan Rathbun]

MaterialsObserver.onMaterialRefChange

"When purging materials, or right-clicking a material in the Material Browser this event triggers one time for every entity with a material. This cause a long series of events to trigger unnessesary."

should read:
"When purging materials, or right-clicking a material in the Material Browser, this event triggers one time for every entity with a material. This causes a long series of events to trigger unnecessarily."

~

[Jim]

Great work, and great resource. Thanks for doing it.

[thomthom]

Thanks Dan, knew it was hiding somewhere.

Thanks for the spelling control - I didn't run a spell-checker or proofread yet.

"Depreciated" has no "i" in it. And the methods that were removed can not be "Deprecated", they are "Abandoned", "Deceased".. whatever they are no longer even in the Observer protoclass.

It's what the API docs said. I've not tested if the events are actually removed, or just aliased to their new names...

[thomthom]

"Depreciated" has no "i" in it.

http://dictionary.reference.com/browse/ ... ref&ch=dic

[thomthom]

Oh, "Depreciated" refer to value...

*grumble something about English and too similar words*

[thomthom]

Another issue, I imagine I read some reports on MaterialObserver.onMaterialSetCurrent not working as it should on OSX in some versions of SU?

[Dan Rathbun]

It's what the API docs said. I've not tested if the events are actually removed, or just aliased to their new names...

I did test the MaterialObserver's Bulk change method. It has been removed from the superclass, so when you create an instance of a MaterialObserver subclass, it just is not there, and if you were to explicitly call the callback, a NoMethodError exception will be raised. Ie, there is no aliasing to the individual change callback(s).

So if you wanted to handle events in a bulk manner, you'd need to do it on the Ruby side, perhaps with a Ruby standard Observer class, or a bulkchange boolean flag variable and an Array of changed objects (each individual SU change callback would need to push it's object into the Array.)

[Dan Rathbun]

Another issue, I imagine I read some reports on MaterialObserver.onMaterialSetCurrent not working as it should on OSX in some versions of SU?

YES.. I posted the answer that Scott gave me. He says it's logged as an internal bug.

See GG post: Materials.current on OS X
.

[thomthom]

http://code.google.com/apis/sketchup/do ... e_observer
Sketchup::Pages.add_frame_change_observer(FrameChangeObserver.new)

This, surprisingly worked.

I'd thought you'd have to reference the model first...
Sketchup.active_model.pages.add_frame_change_observer(FrameChangeObserver.new)

[thomthom]

Another issue, I imagine I read some reports on MaterialObserver.onMaterialSetCurrent not working as it should on OSX in some versions of SU?

YES.. I posted the answer that Scott gave me. He says it's logged as an internal bug.

See GG post: Materials.current on OS X
.

So it's bugged on SU6, SU7.0, SU 7.1?

[thomthom]

It's what the API docs said. I've not tested if the events are actually removed, or just aliased to their new names...

I did test the MaterialObserver's Bulk change method. It has been removed from the superclass, so when you create an instance of a MaterialObserver subclass, it just is not there, and if you were to explicitly call the callback, a NoMethodError exception will be raised. Ie, there is no aliasing to the individual change callback(s).

So if you wanted to handle events in a bulk manner, you'd need to do it on the Ruby side, perhaps with a Ruby standard Observer class, or a bulkchange boolean flag variable and an Array of changed objects (each individual SU change callback would need to push it's object into the Array.)

bulk change event? Is that one that has been removed from the API?
That was a troublesome points when I made this chart - they recently removed some events from the API....

[Dan Rathbun]

So it's bugged on SU6, SU7.0, SU 7.1?

Careful.. the collection getter method Materials.current IS bugged, the question remains, is the MaterialsObserver bugged as well. You'd need to test versions on the Mac to find out.
.

[Dan Rathbun]

bulk change event? Is that one that has been removed from the API?
That was a troublesome points when I made this chart - they recently removed some events from the API....

Yes.. the MaterialsObserver.onMaterialRemoveAll callback was removed, as it never worked, and/or never got called. So there was no point in deprecating it.

[thomthom]

bulk change event? Is that one that has been removed from the API?
That was a troublesome points when I made this chart - they recently removed some events from the API....

Yes.. the MaterialsObserver.onMaterialRemoveAll callback was removed, as it never worked, and/or never got called. So there was no point in deprecating it.

I'd like to add it for historic purpose - anyone remembers it parameters?

[thomthom]

bulk change event? Is that one that has been removed from the API?
That was a troublesome points when I made this chart - they recently removed some events from the API....

Yes.. the MaterialsObserver.onMaterialRemoveAll callback was removed, as it never worked, and/or never got called. So there was no point in deprecating it.

I'd like to add it for historic purpose - anyone remembers it parameters?

Ah - already got it.
Thought there was an event named onBulkChange, like the SelectionObserver has.

[thomthom]

aiaiai... *sigh*

[thomthom]

Update to include SU 8.0 M1
http://www.thomthom.net/software/sketchup/observers/

[Dan Rathbun]

You have that the AppObserver::onQuit is working...

It's is bugged in v8 (at least. And I did report a bug last beta cycle.)

It does not wait until the callbacks return, before shutting down the application.
Trying to use the callback to save plugin settings, and so forth, often fails for me.

Also, trying to dettach the observer itself from within onQuit causes a GPF (if I remember right.)
In addition coders should NOT open any "owned" windows (like WebDialogs and messageboxes,) during the callback, until they fix the 'non-blocking' issue.

---

Can't remember if v7 had any of these problems.

[Dan Rathbun]

InstanceObserver

I'm wonding if you can do this

Code: Select all

#
class InstSpy < Sketchup::InstanceObserver

  if RUBY_PLATFORM =~ /(darwin)/
    def onQuit()
      Sketchup.active_model.definitions.instances.each{|i|
        i.remove_observer(self)
      }
      # what about Groups?
    end
  end #if

  def initialize(*args)
    Sketchup.add_observer(self) if RUBY_PLATFORM =~ /(darwin)/
  end

  # define your Instance callbacks:
 
  def onOpen(instance)
    puts("onOpen: " + instance.to_s)
  end

  def onClose(instance)
    puts("onClose: " + instance.to_s)
  end

end # class


(Note how I changed the example, from UI.messagebox to puts.

[thomthom]

You have that the AppObserver::onQuit is working...

It's is bugged in v8 (at least. And I did report a bug last beta cycle.)

Ok - I'll add that. Never used it. All the observers I've not tried and not heard any reports on I assume is working.

InstanceObserver

# what about Groups?

Groups and Images all are listed in model.definition.

(Note how I changed the example, from UI.messagebox to puts.

I wish the docs would do that. Messageboxes are double plus ungood - especially in observer events.

[thomthom]

Dan, how do you test AppObserver::onQuit ?

If I had a test case that illustrates the symptoms I could quickly run it on SU 6-8.

[Dan Rathbun]

Dan, how do you test AppObserver::onQuit

Tough one... as it DOES fire. (It just doesn't wait for the callback to finish.)
Well I first found it because I wanted to save the pos & size of several windows using Win32API calls.
But it wasn't working.

So I tried putting a debug popup messagebox inside onQuit(), thinking the shutdown process should stop, until I clicked "OK". But it DIDN'T stop, the messagebox came up, but everything else shutdown. And when I closed the messagebox, it's owner (the Sketchup App window,) no longer existed, so Windows poped up the VS Just in Time debugger query box (you know.. "DO want Debug now?")

So next I tried, to open a WebDialog.. and that made things even worse. (I got a post-shutdown BugSplat!) They were either GPF or "Unhandled Win32 Exception".)

I was trying to call a method that saved a whole bunch of data into the Registry, and that method may have called another method as well. Anyway, I even tried "lean" the process way down, putting all statements inside the onQuit(), and removing every extra reference assignment I could. But Sketchup just would not wait. It seemed to be processing all the onQuit callbacks, in the Ruby thread, but at the same time, cleaning up the main App thread, and disposing of all it's child windows.

Perhaps it's a unique thing because I want to save window object data, ...that most other plugins would not encounter.

I plan to try using Ruby define_finalizer, and see if that works for me.

As far as testing.. I'll have to think more on this... perhaps a trace function.

[Dan Rathbun]

InstanceObserver

# what about Groups?

Groups and Images all are listed in model.definition.

I don't have a Mac to test this InstanceObserver workaround.

(And I fixed the example. I don't know why, but every single time I go to write an initialize() method, I forget to put def in front of it!)

[thomthom]

In TT_Lib2 I got a function TT.debug that outputs a string to any system debugger, like DebugView. That can probably be used to track onQuit.

[thomthom]

Added public bug tracker to the BitBucket repo:
https://bitbucket.org/thomthom/sketchup ... tatus=open

People can add reports of broken observer events there.