Re: [Boost-users] [Geometry] 3D Oriented Bounding Box

Subject: Re: [Boost-users] [Geometry] 3D Oriented Bounding Box
From: Barend Gehrels (barend_at_[hidden])
Date: 2011-12-15 17:48:38

On 15-12-2011 22:07, Simonson, Lucanus J wrote:
> Jeremy Coulon wrote:
> Le 14/12/2011 19:37, Barend Gehrels a écrit :
>>> Yes, I can tell you that alas this is not possible with the current
>>> version. OBB's are not yet supported. Besides that, 3D intersections
>>> (besides trivial ones) are far from implemented...
>> Another little question. When you say 3D intersections are not
>> implemented, does it mean that
>> a) it would not even compile
>> b) it compiles but gives wrong results
>> c) it gives correct results but is not optimized for 3D ?
> The answer is (a) for the non-trivial intersections, but there seems to be a misunderstanding about the nature of 3D intersections.  3D solid modeling/meshing/graphics algorithms are not achieved by putting a template parameter for dimensionality on the 2D algorithm, passing 3 and getting the 3D algorithm.  The algorithms for 3D intersections on 3D solids represented by a surface mesh, for example, are completely different than the algorithms for 2D intersections of polygons.  They are less efficient (quadratic versus near-linear wrt. vertex count) and much harder to make numerically robust because of the higher order arithmetic involved.  A 3D solid modeling library is an order of magnitude larger in scope than a 2D geometry library.  What is meant by "non-trival" in this context is multiple man years of work, up to a decade or more to be competitive with the closed source offerings in terms of features, reliability and performance.
>

Thanks for this detailed explanation Luke, I agree with this.

Regards, Barend