Hi,
I propose to remove the fastcsg experimental option. There are three
problems with fastcsg:
- Robustness: It is not as robust as manifold and must fallback to Nef
polyhedron for these degenerate cases. User may also turn on both
fastcsg and manifold and get weird result/performance.
- Performance: There are cases in which fastcsg is slower than Nef
polyhedron, and it is slower than manifold.
(https://gist.github.com/ochafik/70a6b15e982b7ccd5a79ff9afd99dbcf#results)
- Maintenance issue: Half of the files in |src/geometry/cgal| are
about fastcsg, retiring fastcsg may allow us to simplify a lot of
the logic.
Please let me know if you have any comments about this. Feel free to
reply to this thread or comment on
https://github.com/openscad/openscad/issues/4870
Cheers,
John
Hi,
I propose to remove the fastcsg experimental option. There are three
problems with fastcsg:
* Robustness: It is not as robust as manifold and must fallback to Nef
polyhedron for these degenerate cases. User may also turn on both
fastcsg and manifold and get weird result/performance.
* Performance: There are cases in which fastcsg is slower than Nef
polyhedron, and it is slower than manifold.
(https://gist.github.com/ochafik/70a6b15e982b7ccd5a79ff9afd99dbcf#results)
* Maintenance issue: Half of the files in |src/geometry/cgal| are
about fastcsg, retiring fastcsg may allow us to simplify a lot of
the logic.
Please let me know if you have any comments about this. Feel free to
reply to this thread or comment on
https://github.com/openscad/openscad/issues/4870
Cheers,
John