Yes, you read that right. Silverlight 1.0 - as in unmanaged JavaScript - and 3D.

Last week I was reminded of the FarSeer Physics engine for XNA and Silverlight 2.0. I have a strange fascination for 3D software - stemming back from the years when I was a beta tester and plugin developer for Caligari trueSpace 3D - so I decided to poke around in the source code from CodePlex to see what they were doing. Turns out it was just mostly linear algebra and some simple mechanics - things I remembered working with many years ago before the age of managed runtimes and rich visual APIs. Back then we had to do all the heavy lifting with highly optimized C++ routines and drawing directly to an HWND. But things are different now, and what seemed incredibly hard in years past really didn't seem all that hard anymore.

So I was inspired I guess.

I know I could have started with the FarSeer code, and built upon that using C# and targeting the managed runtime. But oh no - that was too easy! I figured that if it can be done using simple trig and a solid model design, I bet I could make it work in JavaScript. After all, JavaScript's Math intrinsic object supports Sin() and Cos(), extensible objects, and arrays - and at the end of the day, that's all one really needs in order to create a 3D runtime. Sure, it involves a fair amount of matrix operations and dot-products and cross-products, but all those things can be learned by reading up on Wikipedia these days (or mathreference.com).

So with that, I set out to create an unmanaged 3D visualization engine for Silverlight...

I am making the code freely available and have already published out to CodePlex here: http://www.codeplex.com/polygraph3d

My efforts so far are looking pretty good (I think). Over the course of about 8 hours of coding and testing, I have the basics working. I can create a 3-dimensional polyhedron (a mesh if you will), and I can display the object in a Silverlight Canvas. I can apply scaling, rotations, and positioning of the polyhedra, and I can also build complex objects using polyhedron groups, each of which has its own local scaling/rotation/position transforms. It also supports a movable camera and performs rudimentary perspective correction as well as small optimizations for backface-culling. It currently only supports simple brushes for the object faces - I would like to implement a real texture mapping system, but that will require some very creative use of image skewing to pull it off.

I have a test page up that demos a simple 3D cube that you can move around the virtual viewport space...

For sure, it is a very simple demo. Clicking the red, green, and blue axis bars will move the cube around. The "spin" button introduces a gradual rotation around the Y axis... but since my rotation transform code is currently broken, the effect isn't quite what I had in mind. Hopefully I will figure that particular bug out soon.

One major hurdle I encountered was the fact that Silverlight does not let you draw directly on a Canvas, nor does it let you create your own controls. This posed a problem since I wanted to display things that simply could not be expressed using 2D XAML. I "solved" this problem by performing all of the mathematics needed to project the 3D scene into a 2D viewport, and then I use the vertex data from the projected mesh to build a series of 2D Path elements. It was indeed a dirty trick - but hey, sometimes you just have to make do with the capabilities of the platform.