Good to hear.

I got a model last month with a tree apparently authored by a well known poster in these parts..
It has 18600 groups, 19000+ component instances and 260,000 triangles. For 1 tree. Ay Caramba!

Procedurally generated stuff does tend to go nuts very quickly. I was wondering whether to write a tool that reports how "dense" each Component Definition you have in your model. Polygons / cubic meter type thing as I keep seeing models with an innocent chair or table in the corner with absolutely insane detail in comparison with the rest of the model.

Adam

OK, here's an updated one with 3 color barcharts as suggested my remus. Also turns on scrolling.

Place in your plugins folder

aka 'Not too much, not too little'

One of the common problems I come across when speaking with people using texturing in any package, is how much resolution/detail is needed in textures. Is 256x256 pixels enough? 1024x1024? 4096x4096?

Have too much texture resolution, and you slow down your system, too little and you get blurry renderings.

The amount of resolution you need is "view dependent". If you're very close to a surface you need more detail in the textures than when you're far away.

Goldilocks analyses your materials and model with respect to the current viewpoint and produces a bar chart showing which textures are about the right resolution (green bars), and which are too high resolution resulting in wasted space/larger filesize, or too low resolution resulting in a blurry look - both shown as red bars.

Adam

Sounds completely reasonable to me.

Filesize increase is 50000000
Number of entities 350000

So each dictionary entry is 142 bytes which is completely believable.

rescue with no argument will default to StandardError.
What you want is:

begin require 'foods' rescue LoadError puts "not there" end

You can look at a photo of great building and appreciate this and that, but its a wholly different experience to walk through it. Being inside a building elicits a whole bunch of emotions that you just don't get from a photo.

In a similar vein, I can look at a schematic representation of building / artefact and understand the dimensions etc but a render brings out a whole new set of emotions/thoughts that I just don't get from looking at a schematic representation that I can hopefully take back to the schematic in an iterative way.

Adam

It never ceases to amaze me how stuff you did a very long tim ago pops in the weirdest of places...

Here's a video game I wrote 20+ years ago being used in a different manner...

Hmmm..

@djmx1 said:

Thank you Adam. Your plugin is based Ghostcomp really! is very useful but I still am disappointed, I just want an even lighter than wireframe in order to turn quickly in a complex scene.

What you're asking is Level-of-Detail switching which is going to be clunky in SU. You could do something that watched the camera position etc but its a real hack. So what you ask is something for Google not me.

@djmx1 said:

Adam if only you could make proxy boxes be rendered as original components (on export 2d grafics) it'd be the greatest break thru!

Ah, well if you're using that very fine LightUp software, this happens automagically when you switch to Tourtool mode.

Adam

Sounds like a plan. I guess something like:

[PLUGIN] Proxy View

[Burn after reading ]

Posted the LightUp version here a little later than planned.

Adam

Chucked your model into LightUp to see what happens. Any chance I can get the whole thing?

The walls and 'roof' don't match up so cast a self shadow..

Adam

[flash=600,338:2wkarqp4]http://www.youtube.com/v/brkkFUpemjI[/flash:2wkarqp4]

Right-click on a ComponentInstance and choose "Replace with Proxy Component" to flip it into a box.

Right-click again to restore the original geometry.

Available in the PluginStore

Adam

@malkomitch said:

try to track your observers by storing them in a global variable, for example. And if you find some duplicates, use Sketchup.active_model.entities.remove_observer before calling a new one

Very good advice. Also remove_observer will accept without error an observer that is currently not attached - I always use the cliche of "remove followed by add" to ensure you have just 1 observer.

Hence my point that extending a class has the potential of conflict. Deriving a new class does not.

@chris fullmer said:

It has to be put into a class and you have to instantiate the class as a tool. And you also have to attach an observer to the Sketchup object.

Check out the full example in the introduction paragraph on that page. It shows better how to use it in context.

Chris

Apart from onQuit is broken on Mac OSX which is annoying.

Also be aware that these "stuff has finished / been deleted" generally get invoked after the fact so don't expect to get useful info from the objects in question.

Adam

@unknownuser said:

It is not a good practice to subclass a built-in class, whether Ruby or Sketchup. This could generate clashes between scripts and conflicts in names. Just imagine I do the same, and my script is loaded after yours, then it would be my version prevailing, and your script would not work!

I'm sure what Fredo meant to say is that its not good practice to overload (or to some degree) extend a built-in class. Its very good practice to derive from / subclass an existing class to create a specialization you may want.

Adam

Personally I use Subversion for all the LightUp development as I develop on multiple platforms in multiple locations and it works very nicely. Its also got Tortoise SVN which is a specular frontend for PCs.

But slightly OT, I was just thinking today whether there is any interest in version control for Sketchup models. My guess is most people use a very adhoc "Save As.." approach to keeping previous versions of models as they make changes whereas you really want something like Subversion / Git offer for code - which is being able to roll back to a version at any point in time.

Adam

Its also the 'correct' thing to do.

Specifically, if you don't add real geometry at those points (known as "T-junctions"), you'll actually get gaps when rendering because of the way all renderers (realtime or offline) work.

Its an often misunderstood part of geometry conditioning that non-degenerate zero-area triangles are very useful to ensure a topologically sound surface.

Adam

You might be finding it hard because you're trying to hammer a square peg into a round hole... Like anything you've got to play how a system wants to play rather than try and beat it into submission.

If you want to have a bunch of components that you don't wish to be edited permanently, then you install your Components in Sketchup.find_support_file 'Components' so they appear in the Component browser but are 'copy by value' when used so any changes don't mess with the 'original'.

Or ship your components all wrapped as 'Locked' Groups.

Adam