Turns out it's a stdcall which has no mangling. I can debug the c extension in CBuilder. Very fast. Now I need to experiment with arguments etc. That will be the acid test for stdcall If anyone is interested I can post the files and the steps. Once I get a real good handle on this I'll try it out with XE7 which is a compiler that compiles Win32, Win64 and OSX. I will also test this on my lap top where Borland memory manager (a dll) doesn't exist. It will tell me if I need to include any of these files or not.

@jolran said: Bottom line is the crash does not occour when using GL_LINES. So for me all is fine.. We'd really like to close down any causes of crashes. So please send us examples if you come across it.

Yes, Thanks. I used that and noticed some increase in performance. But I have kind of abandoned the PolygonMesh for now. At least during the previewing state. I experimented building my own meshclass with indexing of points. Instead of using PolygonMesh class as a Point container. Keep arrays of indexes and a Hash containing Points. And the Transformation only deals with the uniq points(Hash) then. Quite a lot of speed improvements! The advantage here is that I can reuse the objects somewhat, while I can't purge/clear(?) a PolygonMesh. So the GC should work a little less. I'm sure you guys have/are doing something similar or more advanced...

entities.fill_from_mesh(mesh, true, 2 | 8) is working nicelly, but a small gotcha.. %(#FF0000)[mesh.add_polygon(1,2,3) mesh.add_polygon(2,3,4)] Does gives hard edges, as expected. %(#FF0000)[mesh.add_polygon(1,-2,3) mesh.add_polygon(-2,3,4)] gives hard soft edge between Points 2, as expected. but so does this: %(#FF0000)[mesh.add_polygon(1,2,3) mesh.add_polygon(-2,3,4)] or %(#FF0000)[mesh.add_polygon(1,-2,3) mesh.add_polygon(2,3,4)] So I'm starting to see some trouble with this method. To be able to stear edge visibility one need to use 2 Points in comparison. For ex a Collection of edge.start and end positions. For a triangulated quad this works but not if the face has innerloops, Or faces where triangulation is splitting the face without hitting innerloops. Was to fast here.. The docs actually explains this behavior, "value of [-1, 2, 3] indicates that the edge from 1 to 2 is hidden"

Draw method, in a tool. Sorry, my misstake. Calculations for nodes are done outside the tool class. Passing through those methods I collect information to be sent to a container that the Tools draw method fetches information for rendering. I have a bit of work left there to get things organized. The data changes constantly so it will probably cost more to iterate what is needed to be updated then to redraw everything in 1 go. Since mostly dealing with Points and vectors, the new data will probably be totally different then the old anyway. I just clear the old data and bring in new, sort of, without going into details.. There are different types of data though. Persistent data that references to a Sketchup entity for ex, can be cached and updated only when switched. Anyway got a new situation now(TOPIC), have to redesign a lot on the Ruby side. I've studied your render cache. It's very nice and good example. Need to study it more though. Thanks for the input

JSONPadding Will try that. Thanks a lot!

Thanks TIG. Your variant is probably more than I need for the moment, but still very clever. And I presume (since nobody mentioned it already) quite safe to use ?

Yep, solid tip. Thank you Dan.

For those still interested in this topic, I started a new thread here: http://sketchucation.com/forums/viewtopic.php?f=179%26amp;t=58195

@unknownuser said: NO not now, because the vertice positions are part of the definition, and not part of the individual instance(s). Thanks for answering Dan. It could be an useful addition to the API, IMO. What confuses me is that we are moving the vertices when scaling the instance(s) using Sketchup normal scaling behavior. So why would it be different if doing a tapering ? I know the calculations for tapering is much more complex, but... Maybe there is need for some symmetry calculation I havent thought about.

That would be better for me. I'll try to clean out the rubbish from the script first. BTW, thanks a lot. You are very helpful back soon..

@unknownuser said: The speed difference results in that it's doing less - and all the work is done in C. While the method I had used, and your offset method, was doing more - partially in Ruby. Yeah, thats probably true. Anyway, speed increase is always good. I'll switch to this method and report back if any oddities. Cheers.

@unknownuser said: What does "reuse the iteration" mean? Hm. maybe my explanations and english where bad. Hash.each{|k, v| v.do something } is called an "iteration", right? Iterate a Hash. Depending on something do the same iteration on other Hash with different values. Seams like waste retyping the same code again.. I know you are asking why I don't pass in a nested Hash, or array and just do everything in 1 go. In this case I just cant.

@unknownuser said: OK.. then why not ALSO have a premade collection of Materials in a "plugins/folder/materials" directory ? Might be better to grab the materials from components in this case. The idea is that the end user should be able to add new components to that folder and everything updates. Anyway, I "sort of" have it working already now with option 3 (wich was pointed out earlier). Thanks for the suggestion.

Thank you Thomthom. I think you answered all my questions. @unknownuser said: All is restricted to the scope of the HTML page loaded This was my main concern. I don't want to pollute other people's plugins if I create a global var. @unknownuser said: Namespacing is still good when you want to use other libraries, you then want to ensure you do not conflict. Ah, I think I understand now. Like if you create a Jquery plugin for public release, for ex? @unknownuser said: But as of TT_Lib, which is a library I have namespaced it because it's mean to be used in different environments. A, yeah. I forgot for a minute it was a library when I was pooking around. Your library is a very useful resource for me, since you are using Jquery .

@jolran said: Sounds really advanced. Responsive window. it's a prototype that seems to work on my mac, might find an excuse to finish it one day. john

Maybe not for this part. I already have a folder with components. But your code might be very useful for a different projekt I have. This code would have to run each time to see if the user has added new components, when launching the webdialog. And add thumbs to the list. I already had it working before with 128 px X 64 px, but it would be nice to be able to set the thumbs proportions. As of now I'm clipping the image in CSS with 2 divs. Never mind how I'm doing it BTW.. The main point is that I may not be able to support IE7 with several imagehacks. BTW a side question: Trying to stay "Sketchup like" with the dialog. Could there be any copyright issues, mimicing Sketchup GUI interface?

@unknownuser said: Isn't that just because the model in your screenshot are mainly thin edges? Yes, thats true. That the components that has faces and materials come out better. Increasing the contrast on those thumbs too much make things worse. So I have found an acceptable between value around + 45 contrast, that all thumbs can benifit from. How to apply it is a different story I can live with the bad aliasing... One can affect some color when saving the component, playing with styles. But havent found any settings that gives me what I want. At it only seems to affect face.materials. @unknownuser said: And is that thumbnail in 100% scale? Thumbs are 100% scale. They look the same in Photoshop. From PS: 80% increase in contrast look really good on edges. [image: eBKC_contrasts.jpg]

Aha, I see. So embedding HTML is more for convenience sake then. I remember even over half a year ago there where issues on Safari, thought they whould have been fixed by now. Kind of the answer I was hoping for though, if I understood it correctly. Thank you.