Where to Put Your Code: Update(), Flush() or Render()

The world script has a number of functions that are used to code logic of the application. Initialization function of the world script (init()) is used to create objects and initialize all other necessary resources on the world load. Shutdown function (shutdown()) is called when the world is unloaded and is used to delete resources that were created during script execution to avoid memory leaks.

But what about the frame-by-frame update? In the world script there are three blocks for that:

• in update() you can control what to render onto the screen and how to do it
• in flush() you can simulate physics, perform calculations and implement non-rendering game logic
• in render() you can correct behavior according to the updated node states in the same frame

In the world script update() you can specify all functions you want to be called every frame while your application executes. It serves for implementing rendering logic. In brief, everything graphics-related is to be controlled from within update(). Here you can:

• move nodes around the scene and rotate them
• change any parameters of nodes
• simulate particle systems
• control skinned animation
• set global rendering settings
• create and manipulate your GUI

World Callbacks

Using flush() function of the world script you can control physics in your application:

• Operate on all physical bodies and shapes.
• Apply forces, impulses and torques to rigid bodies.
• Create, attach or break joints, as well as modify their parameters.
• Create and adjust physicals.

But do not hurry to code everything else in the update(). flush() can be used not only for physics simulation, but also to implement logic and perform calculations you would not want to do each and every rendering frame. Let us look into the details why is it so.

Usually, the physics framerate is lower than a rendering one (see the details), which makes flush() more performance-friendly in case of heavy calculations. It goes in parallel with rendering, so CPU bottleneck is avoided. Besides that, we know exactly how long one physics tick takes, because physics has a fixed framerate. These factors combined make flush() ideal to perform calculations.

For example, we need to calculate a new position for the object. We can do it in the update(), but it would stall the rendering. Instead, we can start rendering and in parallel calculate a new position using flush(). The problem that rendering frames are more frequent is easily solved. Knowing the new position, we can linear interpolate positions in between and apply them to objects each rendering frame. As a result, performance load will be much better balanced, without load peaks.

Physics Callbacks

The render() function of the world script is an additional function used to correct behavior after the state of the node has been updated (for example, skinned animation has been played in the current frame or particle system has spawn its particles).

Imagine a situation when we need to attach an object (let's say, a sword) to the hand of a skinned mesh character. If we get transformation of the character hand bone and set it to the sword in the update() function, the attachment will be loose rather than precise. The sword will not be tightly hold in the hand, because the animation is actually played right after the world script update() has been executed. This means, returned bone transformations in the update() will be for the previous frame. World script render() is executed after animation is played, which means from within this function you can get updated bone transformations for the current frame and set them to the sword.