Revenge of the strap rigging

You ever start something and think "oh boy this is going to be so simple!", and then it's not- it's really, really not?

Long story short, I've been bashing my head against this particular brick wall for the past couple of weeks. Originally I thought I could setup a system for handling the strap rigging using perfect circles and tangents between those circles (which are easy enough to calculate), however this didn't work for a variety of different reasons:

1) The breasts aren't perfectly spherical (they're kinda lopsided ellipsoids, mostly to simulate the swimsuit sorta "holding them back" a bit)

2) The shoulders don't form a perfect arc and never have

3) The back (more specifically, the shoulder blades) don't form any kind of easily expressible curve

4) Both the shoulders and back tend to deform quite wildly when the character moves around, which further complicates the curves formed between those two

All of this combined resulted in a system that was both ungodly complicated and still somehow managed to fail my #1 requirement- the straps must remain as straight as possible at all times. The shear number of individual components operating within the base character rig made it nearly impossible to keep track of everything inside the swimsuit rig, which meant that the straps would eventually get jogged out of alignment and start to bend in weird and mysterious ways.

So that didn't work.

Fortunately, I think I've come up with another system that will.

It occurred to me a while ago that the strap problem can actually be collapsed down onto a 2D plane as defined by three points- the starting point of the strap, the center point inside the shoulder, and the end point around the front of the breast. A well defined clipping line also exists between the beginning and end points, which means you only need to worry about everything on the side of that plane heading up towards the alignment vector.

If you were to generate a cross-section of the entire model along that 2D plane, you would then have a set of points in 2D space that could essentially be considered a convex hull. By using an algorithm such as the graham scan, it becomes possible to compute the ideal path of the straps in such a way that they will automatically stretch across the gap formed between the breasts and shoulders.

At that point, it also becomes possible to "add" or "subtract" point clusters from the same hull, which means you can use the same system to determine if and where the fingers intersect with the strap plane and position the strap so that it automatically loops over (or under) those points. Technically this means that the characters can't pull the straps to the left or right, but this was a limitation of the previous system as well and something I can easily fix by implementing an "override" system that lets me tweak the path of the straps after they've been computed by the hull algorithm.

In theory, this system should be capable of following the curves of the posed body nearly perfectly, since it will be operating off the actual deformed geometry rather than trying to predict what that final shape will be with approximated spheres and perfect circles. This is turning out to easily be the most complicated part of this entire project (as far as rigging is concerned), so I don't really foresee any other issues assuming I can get this working (since everything else is literally just a matter of hooking up a ton of controls to a bunch of systems I'm already well familiar with).