Time driven loop

[sfrdmn]

Currently main-loop diffs and renders views as needed, when the state is changed. This works great for state which is a function of user input, but seemingly not as well for state which is a function of the render loop itself, i.e. time. I'm thinking of a momentum scroller, for example, whose state is still changing after the user has removed all inputs.

I'm wondering whether you think it's in the scope of this module to handle the case where the application needs a more game-loop like model. Of course, this could be achieved with a second loop, but this doesn't seem quite right - at least for the fact that changing state in the one animation frame will schedule a rerender for the next frame, whereas you really want to update it immediately.

Maybe there could be an option to provide an 'ontick' hook, in which case there is a continual raf callback at work, which checks a 'dirty' flag if it needs to rerender. If there is no hook, it can function the same as it does now and schedule the raf callback as needed

Maybe the loop logic could be reused from another node module?

[yoshuawuyts]

Can you explain why using raf doesn't suffice?

[sfrdmn]

It suffices. My application currently has two raf loops, one for state updates, and the main-loop one for rendering. This issue is more wondering "does this functionality belong in this module"

Just identifying that there is this case where applications need more of a game loop, and that this module could also be implemented like a game loop to run a callback every animation frame, checking a dirty flag

[waylonflinn]

some work towards this in #28

the PR passes through the high resolution time from requestAnimationFrame to the render callback.

I added this to solve a problem similar to the one described (scrolling).