[GRASS5] detri license clarified (new hope for geom)

[Andreas Lange]

Hi David,

i highly agree that an API is needed instead of the modules hacks right
now. 

At least i need the s.geom and v.geom functionality right now. 
I did some superficial tests with s.delaunay and s.voronoi. s.delaunay
seems to work ok, but s.voronoi has problems with some point
distributions. With the bugsites points from the spearfish data set i
can not generate a correct voronoi diagram. This seems to me a problem
of the implementation of the sweepline algorithm.

Let me summarize what is needed (only as a first proposal from users
standpoint):

delaunay triangulation in 2d and 3d from sites and vectors (points)
voronoi diagram in 2d from sites and vector (points)
convex hull in 2d (and in 3d?) from sites and vector (points)
min-max angle triangulation (?)
min-max slope triangulation (?)
min-max height triangulation (?)
calculation of centroid points for polygons (for m.clump or something
similar).

All the functions should work on grass data structures for points and
vectors. This will require some more conceptual work as it can not
always be assumed that the data can be loaded in memory in one hunk. At
least the hacks used right now (writing the points in a specific format
to a temp file, processing this and re-importing again) can not be used
with library functions. 

Sorry that i can not present something more elaborate. But i think that
a library for computational geometry is urgently needed. 

Andreas

David D Gray wrote:
> For a later version of the vector library we want to have support for 3d
> vector maps directly in the vector API itself. The idea is to use
> triangulated surfaces as the geometric primitive. So a suitable library
> would have to be able to do 3d triangulations, also would have to be
> implemented as a well-exposed API - most of the triangulation hacks seem
> to be end-use programs. I suspect that whatever is used will have to be
> customised to some extent.
> 
> To create a suitable surface, triangulations should
> 
> * be constrained so that triangles are moderately well-behaved, but this
> should be flexible
> 
> * show some relationship between local cell size and relief, for example
> larger triangles in regions of low curvature. But then - how is point
> selection done?
> 
> There is another public domain (sensu stricto) implementation called
> dewall, which does 3d  (http://vcg.iei.pi.cnr.it/DVRunstruct.html). The
> documentation (dewall.pdf) gives some benchmarks for different
> algorithms.
> 
> In sum, I don't have a clear idea yet how to get the details of the
> vector library, but we need a triangulation API. If it's not Detri, it
> would be something else, perhaps the precise method is not critical.
> 

-- 
Andreas Lange, 65187 Wiesbaden, Germany, Tel. +49 611 807850
Andreas.Lange at Rhein-Main.de - A.C.Lange at GMX.net



---------------------------------------- 
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo at geog.uni-hannover.de with
subject 'unsubscribe grass5'