Code:

void setup()
{
  // do setup stuff
}

void update()
{
  // compute stuff
}

void draw()
{
  update();
  // draw stuff
}
 

I don't think that having an update method that is called from draw is what aermartin was thinking of since it will not reduce the amount of computation in the draw function.

In the following code the setup() method is used to register the pre() method.

This effect of this is that pre() is called before each call to draw() i.e.
pre / draw / pre / draw/ .....

To stop the call to pre() then use deregisterPre()

There are other methods like pre() more info at
http://dev.processing.org/libraries/basics.html

Code:

void setup(){
   // blah blah blah

   registerPre(this);
}

void pre(){
   // do all the state calculations here
}

void draw(){
   // draw the objects here
} 

Hope this helps.  

I made a mistake in my original post when I said
Quote:

it should be unregisterPre()

PhiLo, I didn't mean my post to be a criticism of your solution I only meant to offer an alternative. For simple sketches I suspect, like you, that there is little difference between them in terms of efficiency.

I offered my solution because it offers a clear separation between updating the 'data' and drawing of that 'data', which is what I think aermartin is after.

Also aermartin said

Quote:


and this can be controled using registerPre()/unregisterPre()

My final thought:
Quote:

True but then surely they are convenient for regular sketches as well?

Even though they are documented in the developers section of the Processing website they are not so complicated to use especially for more experienced programmers like aermartin