Code walkthrough

Simple-3dviz uses moderngl to create OpenGL contexts, frame buffers etc. However, ModernGL is a low level abstraction over OpenGL which means that we get to decide how to manage the rendering pipeline ourselves. Moreover, ModernGL does not provide rendering primitives which means that we need to write our own shaders to render objects/scenes.

The following sections describe in high level the core parts of our rendering pipeline. For details, e.g. argument names and types, follow the links to the API docs or read the code.

Scene

A Scene is a representation of a scene to be rendered. It manages the ModernGL context and defines things like the background color or the size of the frame. The scene also provides information about the camera as well as any lighting to the shader programs that actually do the rendering.

All scenes have the ability to contain several objects that can be rendered, namely implement the Renderable interface, which can be added or removed with the add() or remove() methods of a scene. When the render() method of a scene is called the GPU actually renders all objects on the framebuffer.

The main scene implementation of simple-3dviz contains a single perspective camera and a single light source.

Renderables

The classes that implement the Renderable interface actually contain the code that renders objects in the frame. They get the uniform values (see the GLSL docs for uniforms) from the scene and they must be somewhat compatible. For instance, in order for a scene light to have any effect, the Renderable needs to have a GLSL shader that takes it into account.

Simple-3dviz starts out with the following Renderables:

  1. Mesh is the workhorse of most rendering pipelins and renders a number of triangles in the scene
  2. Spherecloud renders a collection of 3D spheres of specific color and size.
  3. Lines renders a set of 3D lines with specific width and color.

Window

Simple-3dviz, due to ModernGL providing the OpenGL context, does not require any graphical user interface to be able to render scenes. However, in most cases we want to explore a 3D model before settling towards an animation or presentation.

The BaseWindow interface implementations take care of the interaction between your code and the GUI. It is designed to be agnostic of the actual GUI library that implements it. All changes in a GUI scene are done using behaviours, which are explained below.

One implementation of BaseWindow is provided using wxpython.

Behaviours

The goal of the Behaviour interface is to decouple the changes of a scene with the scene management (creation, rendering, etc.). The interface defines simply a method behave(params) which is called every time that a frame is rendered (this is traditionally called a tick). The argument params is an instance of the class Behaviour.Params which provides to the behaviour all the necessary information:

  • A window if it exists
  • A scene object to be possibly altered
  • Access to the frame for saving
  • Objects to access mouse and keyboard information if available

The above design allows for many reusable behaviours that implement animations, mouse interaction and more. A non-exhaustive list follows:

  1. CameraTrajectory
  2. CameraTargetTrajectory
  3. LightTrajectory
  4. SaveFrames
  5. MouseRotate
  6. MouseZoom
  7. SnapshotOnKey