Observers WhishList

Chris Fullmer · 11 Dec 2009, 21:51

I don't have a specific request off the topf of my head. I don't use observers very often.

But my request would be that the observers that are in the API should work. smoothly and consistently. And the API should document better how they work, what values they return, when they kick in, etc.

So for me it just be nice if everything worked.

Chris

Whaat · 12 Dec 2009, 01:31

Here's a few ideas:

make the materials observer for Mac and PC behave the same way. On Mac, the observer doesn't fire (onMaterialSetCurrent) unless you paint a face, whereas on PC, it fires when you click on the icon in the Materials Window
Make them more stable. Observers seem to be a leading cause of crashes especially if you are using entity and entities observers
An idea that's related to observers: I would like a method to determine the time a component definition was last modified. Modification would be any change for any entity in the component definition including any material assignments that have changed for any entity in the definition. I would think that the Entities Observer should accomplish this but, as I recall, this only would fire in a few specific circumstances.
It would be nice if the API docs would say all of the specific events that are supposed to cause an observer to fire. For example, OnEntityChanged, what constitutes a 'change'? It shouldn't be up to the developers to have to endure countless hours of testing to figure out what events actually trigger the observers.

Thanks for working to improve this stuff!

thomthom · 12 Dec 2009, 08:50

@whaat said:

It would be nice if the API docs would say all of the specific events that are supposed to cause an observer to fire. For example, OnEntityChanged, what constitutes a 'change'? It shouldn't be up to the developers to have to endure countless hours of testing to figure out what events actually trigger the observers.

Yea - I spent some time on this because I was setting attributes - which seems to be a change. Not sure if I want that to trigger. At least not most of the times. attributes are like meta data.
maybe onChange for geometric changes, and onAttribChange(dict, key, oldVal, newVal) for the attribute meta data?

fredo6 · 12 Dec 2009, 16:10

On my side, the priority should be:

an APP and Model Observer which behaves correctly in reliably for events: Create New Model, Open Existing model, whatever method is used (.i.e from whithin Sketchup, via revert, etc...). Today the Observer is bugged and does not trap events in case you open a model, but get through a dialog box to save the current one before. This is critical, because these observers are usually the place where you will set or reset all others.
a new observer event onSaveBefore, which is fires before the model is saved by the user. This gives an opportunity to save parameters (usually as attributes) into the model.

Fredo

Dan Rathbun · 13 Dec 2009, 07:39

@unknownuser said:

a new observer event onSaveBefore, which is fires before the model is saved by the user.

The tradition in naming events [using DHTML events as a reference,] seems to be:

Event names begin with "on".1. If a modifier is NOT needed, the event is usually always an after event, such as onclick. The event cannot fire until after the click occurs. (There are exceptions, such as onunload, which is a before event. Some browsers such as MSIE implement a non-standard alias called onbeforeunload. Anyhow.. unmodified names are or should be unambigious as to when they fire; ie: onunload would not be any use if it fired after the unload happens.) 1. If a modifier IS needed, there are two kinds:

Verb based: the eventnames end with a verb, with the modifier placed between "on" and the verb.:
onbeforeEventverb** onafterEventverb** Noun based: the noun follows the "on", with a modifier at the end.
onEventnounstart* onEventnounstop

So, referencing the URL below and the way most SU API events are named, the eventname Fredo wants should be called ModelObserver.onBeforeSaveModel.
The SU API for most Observer events follows close to the way that DHTML events are named. (There are events in both APIs that did not follow the naming 'convention', for whatever reason...?)

ref: DHTML events
http://msdn.microsoft.com/en-us/library/ms533051(VS.85).aspx

I don't know why the W3C named events this way, but they did. (IMHO, I think even some of the Noun-based event names would be more clear with a 'before' or 'after' between the 'on' and noun.)

The list at MSDN may give some readers ideas as to similar events or observers they would like in the SU API.

Dan Rathbun · 14 Dec 2009, 12:05

suggested Purgeevent methods,
fire ONCE regardless of entities removed:
(allows bulk handling in one event, instead of dealing with multiple events firing for each purged object.)

onBeforePurge(*collection, itemsObjectArray*)
onAfterPurge(*collection, itemsStringArray, purgeResult*)

(onAfterPurge must return a an array of string names because the objects are now gone; the purgeResult would be the boolean returned from .purge_unused)

for: DefinitionsObserver, LayersObserver, MaterialsObserver, StylesObserver
(and any other new observer added that gets a .purge_unused method)

thomthom · 14 Dec 2009, 12:37

@thomthom said:

An event that triggers when it's safe to perform model changes would be good. So we can stack up modification, based on information from the other events, and execute them during this safe event.

Do I understand it correctly that the TransactionCommit event is suppose to be doing this?
There is another issue here though. Say that TransactionCommit was working and you use that to perform actions which you have queued up during various model change events - then those changes will also trigger an TransactionCommit event. Would be useful to have some more into to what event was completed - or started. As oppose to just getting an event that says "some operation" started/ended.

Dan Rathbun · 14 Dec 2009, 12:43

@thomthom said:

Tools observer that let you determine what custom Ruby tool is being used.

YES I agree.
Currently all custom Tools are reported as name="RubyTool" id="50003" state=nil
(or similar.)
In my custom Tool, I've made attribute getters tool_name, tool_id, tool_state but the ToolsObserver didn't ask for them I suppose. (or maybe it's the Tools collection?)

Also the Tools collection does NOT work with custom RubyTools. Which may be the ToolObservers' problem as it gets it's info from the Tools collection.

%(#804000)[(Addition:) I'd like some VCB methods (or extra parameters passed by current methods,) to supply VCB Label and VCB value.
If new methods, something like:
onVCBlabelchange( *tools, tooldata, vcblabel* )
onVCBvalueentry( *tools, tooldata, vcbvalue* )
tooldata = Array[ tool_name, tool_id, tool_state ]

reason for labelchange event, is that we cannot rely on tool states as they vary too much, and worse, not all standard tools even use / change a tool state. So in order to know where a tool is in it's process we might be able to use what's displayed for the VCB prompt label.

It would also be nice to get the VCB value after user entry from outside of a standard tool. (I know we CAN do it within our own custom tools using the onUserText method.) Uses, well I can think of implementing a entry history for standard tools, for one.]

mtalbott · 16 Dec 2009, 19:52

I'm a newbie to Sketchup Ruby API so this might be me and not the observer...

When using a selection observer to control the visibility of all entities based on the attributes of the selected entity, it allows me(and I don't want to be allowed) to select hidden entities even when "show hidden geometry" is unchecked.

This is my first use of an observer so maybe I just have to play with it more but I was thinking this might be a limitation of the observer.

thomthom · 16 Dec 2009, 20:00

@mtalbott said:

When using a selection observer to control the visibility of all entities based on the attributes of the selected entity, it allows me(and I don't want to be allowed) to select hidden entities even when "show hidden geometry" is unchecked.

You will see that SU, when you triple click a mesh, will select hidden geometry, and display it as dotted edges.

Check the entity's .hidden? or .visible? property along with Sketchup.active_model.rendering_options['DrawHidden'] to determine if an entity is visible and if the user has Hidden Geometry on.

mtalbott · 16 Dec 2009, 20:45

@thomthom said:

Check the entity's .hidden? or .visible? property along with Sketchup.active_model.rendering_options['DrawHidden'] to determine if an entity is visible and if the user has Hidden Geometry on.

thomthom,
thanks for responding to my issue. However, with my limited knowledge, checking hidden or visible is only good after the selection has been made. I don't know how to effect what can be selected in the first place.

The issues can be simplified to this: all entities have a dictionary with an attribute value set to either "A" or "B". The script's job is to watch the selection that the user makes and based on what is selected, change the visibility. If an entity with the "A" tag is selected, only other "A" entity are visible and visa-versa. Furthermore, if nothing is selected both "A" and "B" entities should become visible again. The problem is that after I start the observer and select an "A" entity all of the "B" entities are hidden but if I go to select a different entity (which should be another "A" because that is all that is visible), the selection tool will pick the entity as if both "B" and "A" are visible. So if there is a "B" entity is in front of the "A" that I want to select it will select the invisible "B". This also goes for selecting in white space.

I figure that I need to tell sketchup that something has changed and it should get it's act together but that is where I've fallen short and thought maybe the selection observer was just not quite observant enough.

thomthom · 20 Dec 2009, 14:39

I'm a bit surprised there hasn't been more activity in this thread...

thomthom · 20 Dec 2009, 22:46

Before, Styles where called Rendering Options. So look at the RenderingOptions class. You have a RenderingOptionsObserver class: http://code.google.com/intl/nb/apis/sketchup/docs/ourdoc/renderingoptionsobserver.html

Same thing as Scenes in the SU UI is Pages in the API.

Dan Rathbun · 21 Dec 2009, 10:17

@thomthom said:

Before, Styles where called Rendering Options. So look at the RenderingOptions class. You have a RenderingOptionsObserver class: http://code.google.com/intl/nb/apis/sketchup/docs/ourdoc/renderingoptionsobserver.html

OK looked at that... seems like the RenderingOptionsObserver class needs an overhaul:

Need to alias (rename) as: StyleOptionsObserver* Need to deprecate the old name.
@unknownuser said:

The type is an internal number that indicates what was changed. You will need to watch the observer for numbers you are interested in.

That is clumsy! (or lazy programming.) Ruby is a high-level language, lets have the same identifier keys as used in the RenderingOptions collection class.* RenderingOptions also needs to be aliased or renamed StyleOptions.* Need to deprecate the old name.
This only covers the onStyleEdit method I proposed in the previous post, which is actually named onRenderingOptionsChanged, however, it fails to send any detail, it only tells (by ordinal,) what options WAS changed. I think it should pass oldValue and newValue as well.

So reproposing the method as:
onStyleOptionsEdit(styleOptionsCollection, optionKeyname, oldValue, newValue)
Could still have old method that sends ordinal number, as:
onStyleOptionsChange(styleOptionsCollection, optionNum)

Need to deprecate the old name.

STILL, need as in previous post, a 'higher' level StylesObserver for triggering events when Styles are manipulated as a whole set (saved to, or loaded from files.) We can call these a StyleSet.

It is possible that those methods could go in the AppObserver rather than create a new observer class.

TIG · 21 Dec 2009, 11:07

I think you are missing a practical point - you can change a "Rendering Option" without changing the current Scene's "Style" - e.g. switching Xray-mode on won't affect the "Style" BUT it will affect what is displayed on screen [i.e. "Rendered"]. Therefore shouldn't there be two sorts of Observer - one to watch for changes to the "Rendering Options Settings" and another to watch for "Style Changes"

thomthom · 21 Dec 2009, 11:26

@tig said:

I think you are missing a practical point - you can change a "Rendering Option" without changing the current Scene's "Style" - e.g. switching Xray-mode on won't affect the "Style" BUT it will affect what is displayed on screen [i.e. "Rendered"]. Therefore shouldn't there be two sorts of Observer - one to watch for changes to the "Rendering Options Settings" and another to watch for "Style Changes"

If you enable X-Ray mode you change the current Style.

thomthom · 21 Dec 2009, 12:01

Ah. Gotcha!

TIG · 21 Dec 2009, 12:03

@thomthom said:

@tig said:

I think you are missing a practical point - you can change a "Rendering Option" without changing the current Scene's "Style" - e.g. switching Xray-mode on won't affect the "Style" BUT it will affect what is displayed on screen [i.e. "Rendered"]. Therefore shouldn't there be two sorts of Observer - one to watch for changes to the "Rendering Options Settings" and another to watch for "Style Changes"

If you enable X-Ray mode you change the current Style.

That way you don't change an existing Scene's Style per se - the current Render Settings are temporary changes in that View not the Scene's defined Style - if you change back to that Scene tab the Style should be refreshed back as it was until you save its changes - I know that Xray mode is perhaps a bad example in that it transcends Style settings.

My point is that what is being Rendered in a View and a Scene's Style are different things. One is transient, one is fixed. Changing the Style changes the Render Setting it uses, but you can change the Render Settings without changing a 'named' Style - the confusion is that changing the Render Settings makes a 'temporary' Style that is used in that View from then on until you change settings again or go back to another saved Style - you can the save this temp-style as a new Style or use it to update another already defined Style...

My point was that Styles and Render Settings are linked but different as any Style's settings are not necessarily equal to the current Render Options ?

Dan Rathbun · 21 Dec 2009, 12:12

@tig said:

Therefore shouldn't there be two sorts of Observer - one to watch for changes to the "Rendering Options Settings" and another to watch for "Style Changes"

Yes possibly. The API is WAY out of date on this subject and not very clear.

Likely that Rendering Options came first (before Scene Styles were 'invented'?)

@unknownuser said:

(in regard to Rendering Options) The majority of the rendering information returned exists in the Model Info > Display section of SketchUp
But that is no longer true!
There is no 'Display' section in the Model Info dialog in 7.x!

thomthom · 21 Dec 2009, 12:15

Don't think I've ever seen a Display section. But I've only used SU since v6.x.