Observers WhishList

thomthom

So what's the latest gossip from the Google camp on this topic?

Dan Rathbun

need AppObserver.onCloseModel

It is often used to attach other observers to each model as it is opened or started. This ensures that your observers are watching all open models.
http://code.google.com/apis/sketchup/docs/ourdoc/appobserver.html

In an MDI application (as the Mac is now, and hopefully the PC will be in upcoming versions,) there needs to be a way to detect that an MDI child window has been closed, so scripts can do cleanup, such as unattaching observers that may have been attached using onNewModel or onOpenModel.

Of course on the PC, at the current time, Sketchup is not yet an MDI application; and to close the active model, a user would either open a new model (firing the onNewModel event,) open another model from a file (firing the onOpenModel event,) OR shut down Sketchup (firing the onQuit event.)
_

Dan Rathbun

_
I've noticed there does not seem to be Observers for classes:

Importer* Exporter* TextureWriter
Anyone think there's any need here?
_

thomthom

Can't see any reason.
It's the changes in the model that we need to be aware of. Importer, Exporter, and TextureWriter doesn't affect the model.
..well, the importer imports geometry, but the EntitiesObserver and DefinitionObserver and InstanceObserver should take care of that.

Dan Rathbun

_
Sketchup::SelectionObserver

This observer needs fixing! I would put this in the high priority group.

Per ThomThom's Observer Review "State of Observers"
http://forums.sketchucation.com/viewtopic.php?f=180&t=20676&start=0#p173630

Sketchup::SelectionObserver

.onSelectionAddednever triggers
.onSelectionRemovednever triggers.

Instead,
.onSelectionBulkChangeand .onSelectionClearedalways triggers.
_

thomthom

Why doesn't this observer have its own class documented as the other observers? http://code.google.com/intl/nb/apis/sketchup/docs/ourdoc/pages.html#add_frame_change_observer

Dan Rathbun

FrameChangeObserver

@thomthom said:

Why doesn't this observer have its own class documented as the other observers? http://code.google.com/intl/nb/apis/sketchup/docs/ourdoc/pages.html#add_frame_change_observer

I would guess that this was a pre-version 6 observer implementation, added before they 'set a standard' in version 6 for true observer prototypes defined under the Sketchup module. And that when they did add those formally defined Observer classes in version 6, this one fell through the cracks.

I checked the Object list... and there is no observer prototype named 'FrameChangeObserver' defined.
I checked the Sketchup::Pages class for it's methods:

pages.public_methods(false).sort
["[]", "add", "add_matchphoto_page", "add_observer", "count", "each", "erase", "parent", "remove_observer", "selected_page", "selected_page=", "show_frame_at", "size", "slideshow_time"]

There are NO methods:

add_frame_change_observer* remove_frame_change_observer
This information is OBSOLETE in the API, and should be removed.

A coder (currently) would have to use the Sketchup::ViewObserver, onViewChanged method to detect when the change of view finished.

IF there is a need (for the functionality of a FrameChangeObserver,) you might propose a Sketchup::AnimationObserver class that has this callback function and perhaps others?

Jim might want to weigh in on this, as he has been working on Animation lately.

@Jim : Questions:

Would an attached AnimationObserver be expected to slow the framerates?* Would an AnimationObserver even be necessary? (As really the Animation interface, is somewhat like an observer itself. The pause, resume and stop methods act like observer callbacks.)
_

ptitjes

@dan rathbun said:

_
Sketchup::SelectionObserver

This observer needs fixing! I would put this in the high priority group.

I highly second that request! SelectionObserver is broken.

@dan rathbun said:

Per ThomThom's Observer Review "State of Observers"
http://forums.sketchucation.com/viewtopic.php?f=180&t=20676&start=0#p173630

Sketchup::SelectionObserver

.onSelectionAddednever triggers
.onSelectionRemovednever triggers.

Instead,
.onSelectionBulkChangeand .onSelectionClearedalways triggers.
_

Here even onSelectionCleared is not even triggered. (SU 7.1 (7.1.5803) on MacOS SnowLeaopard) This makes working on selections quite impossible and you can't do that in a right way.

Also, plugins should be loaded before any new model is created in SU. Because AppObservers are not triggered for the initial new model open when lauching sketchup.

Both those issues could be easily fixed. I don't know whether sketchup dev team is reading this forum, but please fix that!

Didier.

thomthom

@ptitjes said:

Both those issues could be easily fixed. I don't know whether sketchup dev team is reading this forum, but please fix that!

This post was started by a Google Sketchup employee.

Dan Rathbun

Sketchup::AppObserver

Anyone interested in any of the following event methods?

Extensions
Sketchup.register_extension**%(#BF0000)[onRegisterExtension( extObj, loadFlag, regResult )
onLoadExtension( extObj, loadResult )]**

onUnloadExtension needs fixing. Currently passes the extension name, BUT text names are not necessarily unique, for example there are several extensions, by several different authors called "SelectionMemory".
In addition what would we do with a textname anyway? There is no method to find an ExtensionObject by searching using a textname, as well as even having access to the ExtensionsCollection/Manager.
We definately need the object handle passed, then we can access any of the extensions attributes, including the name, example:
onUnloadExtension( extObj )

Internet Connection
IF Sketchup.is_online returns false, an Observer could be used to watch for an when an Internet Connection is established. This may be needed for some extensions that download help information, or textures or other files from the internet (especially when the user has a dial-up connection.)
%(#BF0000)[onInternetConnect
onInternetDisconnect]

Send Action
fired by Sketchup.send_action%(#BF0000)[onBeforeSendAction( action )
onAfterSendAction( action, result )]

Thumbnails Bulk events
Would likely be used by Gallery or webpage generator Plugin extensions.
%(#BF0000)[onAfterSaveThumb( obj, img, result )
onBeforeSaveThumb( obj, img )]

When fired by:
%(#BF0000)[ComponentDefinition.save_thumbnail
Model.save_thumbnail
View.write_image] (if img is thumbSize)

obj is the source object handle* img is the output imgStringPathname
When fired by:
Sketchup.save_thumbnail
obj is the input skpStringPathname* img is the output imgStringPathname
_

thomthom

Can't say I've ever wanted/needed those events. It's the events that relates to model changes that I'd wish where working reliably.

dkendig

We have always had a hard time ensuring our plugin's data is properly serialized in to the scene prior to a save. Now we serialize after every - single - change... and it slows things down a little. The proposed solution for onSaveModel would be wonderful... oh please magical Google genies... please fix these observer issues. Here's a post we put in the discussion groups a while back when we were pulling our hair out:

http://groups.google.com/group/Ruby-API/browse_thread/thread/bf156b08bdb9ba71/5b9ac6cc4a3bb83c#5b9ac6cc4a3bb83c

tomot

It really bugs me! when I'm doing a load "Filename.rb" into the Ruby Console, that the revised "Filename.rb" may run as expected, only to find out that "Filename.rb" really didn't work. because I had not actually quit the currently running SU.

I'm not sure if this is an Observer issue: It would be really nice if there was a closer OBSERVATION on the part of the API that scripts work or don't work. I don't ever recall having to Quit AutoCad, and restarting it each time I was testing Lisp routines.

In fact GLD is Archicad's API, is even more elegant, the code is written totally within the ArchiCad application in its own GLD window, errors are flagged without ever having to quit the main ArchiCad application.

Dan Rathbun

@tomot said:

It really bugs me! when I'm doing a load "Filename.rb" into the Ruby Console, that the revised "Filename.rb" may run as expected, only to find out that "Filename.rb" really didn't work. because I had not actually quit the currently running SU.

Is ALL of the code written in the "Filename.rb" wrapped within a second level module block?
It should be.
When it is, from the console, you can type:
<1stLevelModule>.module_eval('remove_const :<2ndLevelModule>')
and the entire 2nd Level module should be undefined. Then type
GC.start to run garbage collection.
Then you can reload your script, and redefine the second level module.

What I'm saying is that all of your plugins should be inside modules, which are ALL inside a master module that is YOUR namespace, such as "Tomot".
example:

module Tomot
  module My_Plugin_One
    # CONSTANTS
    # @@module vars
    # class definitions
    # method definitions
    begin
      # run once code
    end
  end # My_Plugin_One
end # Tomot

Then when you wish to reload (at the console):
Tomot.module_eval('remove_const :My_Plugin_One')
GC.start
load 'tomot/plugin_one.rb'
(You might even put that in a temporary Tomot module method, to automate the cleanup and reload of any Tomot submodule or subclass, etc. Make a couple of constants Tomot::MODNAME and Tomot::MODPATH that you can change at will, and have the method use those constant's String values in the 1st and 3rd lines, respectively.) Sorta like this:


Tomot.const_set(;MODNAME,'My_plugin_One')
Tomot.const_set(;MODPATH,'tomot/plugin_one.rb')
def Tomot.reload
  module_eval( "remove_const ;#{MODNAME}" )
  GC.start
  load("#{MODPATH}")
end

You'd repeat the process, for other plugins, always defining them WITHIN your master namespace (module Tomot,) and each WITHIN it's own submodule. Yes if you create classes that are common to more than one of your plugins, they can be defined at the 2nd level, directly under Tomot, and they ALSO can be removed and reloaded because the name of a Class, is also a constant (just as a module identifier is.)

Top Level modules and classes (which should only be the Ruby base classes and modules,) cannot be removed (by design.)

Now this will not solve ALL of the challenges with SU API developement, because many objects are really C++ objects and cannot be garbage collected by Ruby. But we can reduce many of the problems, by at least coding on the Ruby side, in the proper Ruby way.
There have been many discussions in the forum on unattaching and 'cleaning up' observers.

As far as reloading SU. Sometimes it just must be done, especially if your working with menus. We cannot at present remove menu items.
To make things faster, in development, I have a bare-bones Sketchup copy with minimal development plugins only (and the DC extension,) that loads fast. On Windows, we can run Sketchup differently, by using extra user accounts, and using RunAs.. to access these differnt setups from the main user account.
See my posts Saving multiple workspaces (and run batch file from ruby as admistrator perhaps if you wanted to use a batchfile and icons.)

Now.. I agree with you and others, it would be great to have a good bridge between Ruby GC and C++ GC (if that's what it's called in C++.) So that things get cleaned on both sides when no longer in use.

tomot

Dan: Thank you for your detailed explanation. Early on when I first looked at Ruby scripting, I knew nothing about Ruby and I still know very little. Many of the examples I used were from donated scripts, from which I borrowed and referenced code to develop my own self taught scripts, there were no module examples to speak or copy from. Its easy to see why not more people get engaged with Ruby.
To a novice trying to make sense of the current SU Ruby API descriptions and examples, is like shining a light on a subject without a battery in the flashlight.

Dan Rathbun

@tomot said:

Many of the examples I used were from donated scripts, from which I borrowed and referenced code to develop my own self taught scripts, there were no module examples to speak or copy from.

See this thread at GoogleGroups, begining with post 4, in which Harald asks the question "... so, are all my .rb scripts really just one huge script?"
He was having trouble with the Ruby environment getting corrupted with classes, objects, methods and so forth.. that he didn't want loaded.
Read my detailed responces, which explain how things get loaded (both in Sketchup Ruby, and Ruby in general,) and an example of file spanning.

thomthom

Unfortunately, the Google examples does not encapsulate their code. But you'll find most plugins here at SCF are. Usually it's the older ones that doesn't.
The norm is encapsulate the script in a module which is prefixed with your initials - that goes a long way to prevent namespace collisions. I use TT_, Jim use JF_, Chris use Clf_, etc.

thomthom

@tomot said:

To a novice trying to make sense of the current SU Ruby API descriptions and examples, is like shining a light on a subject without a battery in the flashlight.

Yea - - that description sounds just about right.

PurpLev

+1 for EntityObserver currently it's pretty much unusable as it crashes SU when it intercepts anything and tries to perform more than a single printout to the console.

kwalkerman

I definitely add my vote for an onBeforeSave for the Model Observer. Right now, I am doing some stuff and then re-saving...

A few things I came across:

re-saving triggers the onSave method in your application, and in any other plugins that have a Model Observer. This can lead to an annoying infinite loop. I fixed this problem by requiring that the times between saves be greater than a certain set time.
Once you re-save your model, the model observer immediately calls the onSave method. This means that any code below Sketchup.active_model.save "my model.skp" will not be executed until after the onSave is called and executed again... so any infinite loop prevention should be above the save command.

--
Karen