CGAL users discussion list

["" <>]

Thank you for you answer.

Sure, we end-up in a kd-tree-like strategy, but then we can use the same code path for query against any type of entity : 0D, 1D or 2D. (although less efficient for 0D, but that is not the core of our problems).

Best regards,

G.

---

De: "Sebastien Loriot (GeometryFactory)" <>
À:
Envoyé: Mardi 19 Décembre 2017 13:17:06
Objet: Re: [cgal-discuss] AABB Tree: closest_point

I might have miss something but it seems to me that what you achieve is
what a kd-tree would be doing, right? Provided a kd-tree is internally
use by the AABB-tree to speed up distance queries it seems to me
that using a kd-tree directly would be better.

Did I miss something?

Sebastien.

On 12/19/2017 12:27 PM, wrote:
> Dear all,
>
> first of all, thank you very much for the time you invest on CGAL, it is
> a very valuable library !
>
>
> In our library, we have a quite extensive use of AABB Tree for efficient
> distance computation. Lately, I tried to use CGAL (maybe a little bit
> out of his scope...) to compute point to point distances with AABB Trees.
>
> Indeed, only "point to 1D" and "point to 2D" queries are coded at the
> moment. I managed to "hack" CGAL/Cartesian/function_objects.h to make it
> work. Basically, it consists in adding operator()( const Point_3& s,
> const Point_3& p ) in class Construct_projected_point_3.
>
>
> My questions are the following:
>
> 1/ Is it a bad idea ?
>
> 2/ if not, would you be interested in an upstream merge request ?
>
>
> Best regards,
>
> G.
>
>
>