#2d physics

Following my induction, my first task was to begin development of the Canvas PhysicsDevice, adding ray-casts and convex-sweep support so that all the weapons in the Quake 4 tech-demo could be used outside of the plugin. After completion of this task, I began to work on the full implementation of the Canvas PhysicsDevice available in the SDK today!

Implementing the PhysicsDevice was a big project, helped greatly by having had previous experience with 2D physics, but there were a lot of new things to learn. A lot of the time was spent working out the most efficient ways of implementing the physics for JavaScript and tuning physics parameters for the best behaviour.

During the implementation, I worked on a side project in Python - which first required *learning* Python - for the optimisation of collision meshes in the Quake 4 tech-demo. This involved a fair amount of computational geometry that eventually found its way into the WebGL PhysicsDevice, optimising convex hull collisions through abuse of the triangulated hull topologies, and ray testing of convex hulls by using the triangulation.

After much perseverance, the PhysicsDevice was completed (Minus some constraints) together with the Physics Benchmark SDK sample.

My next task was the Draw2D library. This library was originally written for the Score Rush game, and required tidying and optimising before being released into the SDK. I made use of Chrome's profiling to analyse the bottlenecks of the API and was able to make some quite drastic speed improvements. We later had a meeting to discuss changes and extensions to the API which turned into the Draw2DSprite object, and drawRaw interface.

The Draw2D SDK sample was created to demonstrate the capabilities of the library.

Another aspect of the original Draw2D library, was a full-screen bloom and distortion effect. These were generalised and separated from the Draw2D library, and after adding the Gaussian Blur effect, Color Matrix effect and related utilities, the TextureEffects API was born.

Turbulenz now had fully functional 3D physics for both Canvas and Plugin, but with an expected influx of 2D games, including the Save The Day game, it was clear that the Turbulenz SDK should also provide 2D physics. I spent a short time on researching existing solutions including Box2DJS and Box2DHTML5, as well as my own library compiled for JavaScript, none of which proved to be performant enough; it was decided that I should implement a 2D physics library for Turbulenz which would be high performing, and fit in with the rest of the APIs instead. As a formal project, I wrote up a Project Charter and initial API specification that other members of the team could read over and give their input after which I took a well needed holiday.

Returning from my holiday, I had only 4 weeks left to implement the 2D physics; I took a pragmatic approach of breaking down the project into discrete, unit-testable tasks so that I could develop the library quickly, whilst being confident there were no major bugs. With the experience of the WebGL PhysicsDevice and Draw2D, I could see immediately how best to implement the API for speed in JavaScript from the get-go, and everything was completed in time together with the documentation and the three 2D Physics samples.