jwilling / jnwspringanimation Goto Github PK

CGRectValue CGSizeValue CGPointValue are defined in UIKit, CGRect CGSize CGPoint ..are defined in CG, UIKit should be imported here.

I was doing some research on spring-based animations and found this library while in the process of building a new feature for @NimbusKit. Your implementation is a solid step towards providing spring-based animations, but initial velocity support is key to enabling natural gesture-based animations we're starting to see in more apps and it sadly hasn't been implemented here yet.

This paper by David Morin has a wonderful section on using initial conditions with harmonic oscillators (section 1.1.7). From there we can see that the addition of (velocity / freq) * sin(freq * t) to the harmonic oscillation allows us to incorporate initial velocity into the oscillation.

Sidenote: in practice I found that if the delta (A) is negative then you have to flip the sign of the velocity (v) in order for it not to behave erratically.

I made a grapher file so that I could play with the curves a bit easier. Screenshot follows (can't upload the grapher file to Github sadly). The shorter curve of the two is a damped harmonic oscillation without velocity taken into account (effectively v = 0). The taller curve is a damped harmonic oscillation with an initial velocity of 20 (intuitively: the sign affects which direction we overshoot in, the amplitude by how much).

Velocity pointing toward the delta.

Velocity pointing away from the delta.

Example of an iOS 7-style spring animation being thrown toward the destination.

Example of an initial-velocity spring animation being thrown away from the destination.

This is primarily an FYI right now, as I'm not going to use your library in the feature I'm building. The approach I plan to go with will likely not be a subclass of CAAnimation at all, opting instead for a more generalized approach to building key frames for a spring-with-initial-velocity-based tween. The lack of correct transform tweening in this library is problematic, but properly implementing transform tweening requires non-trivial quaternion work, and CAAnimation likely already does that underneath the hood (linked code is not official), so I plan to just let CAKeyFrameAnimation do its magic rather than try to rebuild that code.

Thanks for providing a test bed with which I could more quickly play with this! Your code definitely helped point me in the right direction as far as researching papers and playing with graphs.

Cheers!

• Jeff