Reimplement Poisson_implicit_surface_3 for Mesh_3 (for CGAL-6.1)

[lrineau]

Issue with Poisson_implicit_surface_3

For the moment, in master, we have an improved point oracle for Surface_mesher, than is only used one example file: Poisson_surface_reconstruction_3/examples/Poisson_surface_reconstruction_3/poisson_reconstruction.cpp.

History

Thanks to git log -p --reverse -S Poisson_implicit_surface_3 cgal/master.

• The class template CGAL::Poisson_implicit_surface_3 (and its companion Poisson_implicit_surface_oracle_3) was introduced by commit b0ce738 (Sep 27, 2011, first merged in CGAL by commit 485d1bc, Nov 20, 2012, for CGAL-4.2). In the same branch Surface_reconstruction_points_3-make_it_faster-afabri, there was also the following commits:

  • 197870a adding the use of that better oracle class to the poisson reconstruction demo (demo/Surface_reconstruction_points_3/PS_demo_poisson_plugin_cgal_code.cpp)
  • eb1dc07 adding the use of the improved oracle to the examples examples/Surface_reconstruction_points_3/poisson_reconstruction.cpp.

• then 028b6cc (in CGAL-4.5):

  remove the Point set demo as all functionalities are now in the Polyhedron demo

  but the polyhedron demo never had the speed improvements

• 2016, @sgiraudot added the improved oracle in examples/Poisson_surface_reconstruction_3/tutorial_example.cpp, commit f052a95, but then removed that usage in b9730aa in 2020.

Originally posted by @lrineau in #8248 (comment)

---

In a discussion with @afabri and @janetournois (link to a private Matrix discussion), we have decided the following:

For CGAL-6.0, we have decided to "drop" the potential optimization brought by Poisson_implicit_surface_3, and reimplement it for CGAL-6.1 (in master as soon as the release branch for 6.0.x is created).