weymouth / parametricbodies.jl Goto Github PK

The example provided contains basic b-spline routines that should work on a GPU but cause problems when used inside ParametricBodies. Having this functionality would be a major boost as it would allow creating complex shapes and kinematics.

Hi Gabe, thanks for the great code. My question is, would it be possible to make a ParametricBody compatible with the WaterLily.∮nds method in Waterlily/Metric.jl? So far I think they are not compatible due to ParametricBodies missing the sdf field. Thank you!

We need to define a more general normal vector for parametric bodies. The rest carries nicely from 2D to 3D but not those lines.

https://github.com/weymouth/ParametricBodies.jl/blob/48e49e990d7a47f72289f90f74c2f15f64f00b3b/src/ParametricBodies.jl#L99C1-L102C4

https://github.com/weymouth/ParametricBodies.jl/blob/9e78f08be2c28aabf74a105753b71ac8e7cc8fac/src/ParametricBodies.jl#L104C1-L107C4

This is annoying as it results in a division by zero for the normal vector computation and pollutes the distance field. Luckily it doesn't occur very often and the fix seems simple enough lims = clamp.(lims,lims[1]+eps(T),lims[2]-eps(T)) and clamping any uv to these limits ensure that the normal is always a vector with non-zero magnitude.

We need a way to detect if a NurbsCurve is open or closed, as the function ParametricBodies.notC¹(l::NurbsLocator,uv) is only valid for open NurbsCurve for now. I will do that as soon as possible.