Buck
Buck
I thought you needed a more generalized multi polyline combine operation - for your use case you may be able to just repeatedly call the combine function.
That's something I want to add, see issue https://github.com/jbuckmccready/CavalierContours/issues/11. I'm not sure when I'll get around to it.
Currently the union combine operation utilizes the the polylines being closed for orientation purposes (clockwise vs. counter clockwise which enables the slices to naturally stitch together end to start), and...
These issues have actually manifested in reported issue https://github.com/jbuckmccready/cavalier_contours/issues/23 - commits and improvements are ongoing.
Alright this issue should be fixed with commit: https://github.com/jbuckmccready/cavalier_contours/commit/9fb706efadc757e8274b672660422f8955d89c31 The issue was caused by numerical instability due to the implementation of `line_circle_intr`. I changed the implementation to something more stable...
I'll look into it more today. A few things come to mind: 1. I wonder if I need to explicitly check the line segment end points distance from arc center...
These are great bug cases, helping me correct the fuzzy float epsilon propagation and improve numeric stability! This latest case hit code paths where epsilon values for fuzzy comparing were...
Alright new commit with more improvements/fixes around epsilon values used for fuzzy comparing: https://github.com/jbuckmccready/cavalier_contours/commit/3bf2b379158c4139b5a91144568b33a3cd91e7a3 It fixes the second case you found and I added the second case as a test...
Fixed that case now as well, commit: https://github.com/jbuckmccready/cavalier_contours/commit/fa30fdd15f93417a0d31a004045638939188bf86 Issue was again in the `line_circle_intr` function, this time caused by squared values being fuzzy compared which are not to scale with...
I found using `1e-5` for position equal epsilon and `1e-4` for slice join epsilon to yield good results when using doubles (this is the default used in the library). Small...