Enable non-linear easing

[atoktoto]

It seems to me that most of the work to get non-linear easing is already done.

In order to check what is left to be done I removed the linear-only enforcement, created a simple example with 3 key frames and compared the results with lottielab.com editor:

I have found that:

• only the easing curves from first two key frames seem to affect the animation as rendered in lottielab.
• Lottie spec mentions that: "All keyframes MUST have an i and o value, unless It is the last keyframe in the sequence OR ..."
• This comment mentions that the last one is unused: Eases on last keyframe: A question. Maybe not an issue. airbnb/lottie-web#2620 (comment)

The way I understand it is that while we are moving into position described by next key frame, we still should be using o and i parameters from the previous one.

After changing the Time::frames_and_weight to use the easing from current key frame, positions seem to match pretty well:

What do you think? Is there anything else missing to enable non-linear easing?

Thanks for creating an amazing library!

[simbleau]

Haven't tested yet but the code looks good!

I was the person who implemented the work for non-linear easing at first, but abandoned it after I failed to figure out why my example file (which included position keyframes) was failing. I'm really happy to see someone else do the rubber duck test and spot the issue I failed to see!

I'll give it a go later.

Bless

[DJMcNab]

Will this need to be ported into Interpoli as well?

[waywardmonkeys]

Yes, @atoktoto Could you also PR the same change into the https://github.com/linebender/interpoli/ repository, please?

[simbleau]

Local testing looks great. Super happy to land this; going to push a bug fix version for this as well.

[simbleau]

Yes, @atoktoto Could you also PR the same change into the https://github.com/linebender/interpoli/ repository, please?

linebender/interpoli#31 made a tracking issue for that