Indeed the falloff parameter seems to quite effectively address any sharp bend start/end points.
But I think I meant more "artistically" speaking, or to have the ability to more or less slowly or gradually go into the deform before the bending becomes (exponentially) more prominent, or to have very gradual 'ease-in' and 'ease-out' of the effect.
I understand and agree with curve profiles not being animatable, and concerning that :
how about using a gradient node (color gradient defaulting to linear from 0 to 1) which you can insert divots which have both animatable values AND position on the ramp, and using that to modulate the effect? (or have otherwise arbitrary profiles)
which would more or less be the equivalent of this:
Concerning search algorithms,
I thought about doing it right in ICE like I did in code. But search algorithms are not great in ICE due to poor looping capabilities.
It reminded me of this (by Pooby) if it would potentially be implementable?
(from Resource dump thread, and also on RRay) Post subject: Get Closest Filtered points
A compound that gets ONLY the closest points in the current pointcloud/geometry that have a »true« boolean. It's much much faster than rifling through arrays. Hopefully someone will find this useful. It works by temporarily moving the filtered points up and away on the Y axis, then the remaining points look in the area where the filtered ones got moved to. Then filtered ones are moved back to their original location. The temporary moving of them is nothing to be concerned about because you wont see it and it doesnt affect simulation.
In any event, even without any of that, I think the falloff parameter (or as shown in one of your shots, delving in the compound to set different profiles) would be more than sufficient for the large majority of cases.
BTW, everything in that video is nothing short of awesome! You have great talent my friend!