* CODE+DOC: Bbox_3 is a nested type in CGAL's Kernel; the code uses Bbox_3
* DOC: Add Compute_squared_distance_3 in AABBGeomTraits as it is needed by AABBTraits
* DOC: remove Splitting_direction from AABBTraits as it is not used
* DOC: Squared_distance in AABBTraits is only used between two points
(I guess the confusion from the fact that nearest_point use the other
version but without using the AABBTraits model).
* CODE+DOC add the function to construct Squared_distance functor in AABBTraits + AABB_traits
and use it instead of AABBTraits::Compute_squared_distance_3 which was not in the concept
* DOC: remove from AABB_traits types Ray_3, Line_3 and Segment_3 that are not required by AABBTraits
* CODE: The code was using AABBTraits::Point which is not in the concept. Use Point_3 instead
* CODE: reorder typedefs in AABB_traits to match concept order
AABB_tree documentation, Intersections_3 and Arrangement_on_surface_2.
The merge also introduced a regression for a bug fixed in R68387 which
fixed a bug in Triangle_3 Segment_3 intersections. The relevant
testcase is in triangle_other_intersection_test.cpp.
* simplify code of AABB_tree::First_intersection_traits
* change intersection(ray,triangle) and intersection(line,triangle):
returned object is now an empty object, a point or a segment
* remove do_intersect test from Intersection::operator() of
AABB_traits: intersection(ray,triangle) and
intersection(line,triangle) do it.
IMPORTANT THINGS
- We need to add Compute_squared_distance_3 (at least for Point_3, Point_3) to the traits requirements
- Since we use the (decorated) Traits as a template parameter for the Search_traits_3, we need to add
this as a requirement. Currently, the Search_traits_3 requests that the template parameter be a CGAL
kernel. This is much too strong!
the KD-tree uses decorated points to return the Id too.
There are just a few boost::optional to add in the traits, where we still
use the old bool intersection(..., Result&) approach, and some modernization
to do in the distance computation things.
Everything compiles.
I have removed the funtion closest_primitive from the code and documentation,
since it would not be consistent with the rest of the interface, and would not
be faster than closest_point_and_primitive.
I think we are almost done, not only for the documentation, but for the code too.
Sébastien Loriot
Philipp Möller
Laurent Rineau
Stéphane Tayeb
Pierre Alliez
Camille Wormser
Laurent Saboret