CGAL users discussion list

[Pablo Miranda Carranza <>]

Hi, I should first thank everyone involved in CGAL for a fantastic library.
I am researching on a spatial analysis tool for architecture and urban 
planning which uses skeletons. I am at the moment employing  the "2D Segment 
Delaunay Graphs" package in CGAL for calculating the Medial Axis. It works 
very well, and after some fiddling it has been no problem to draw the voronoi 
using conics and straight lines, as well as extracting all the information 
for my analysis. I found that one of my test polygons produced a CGAL::error, 
while using Segment_Delaunay_graph_filtered_traits_without_intersections_2. I 
am not sure if this is a bug, or simply expected behavior of 
Segment_Delaunay_graph_filtered_traits_without_intersections_2 (most likely) 
.  Anyway, if it is not a bug, I guess that it may be useful to have a test 
example of a simple polygon that fails with the filtered traits (I have 
simplified the original input). I wonder also if anyone has any suggestions 
on how to handle failures (I am planning eventually to catch the exception 
and only use exact numbers when the filtered traits fails?) . Thanks!


OUTPUT:

processing 0 of 9
processing 1 of 9
processing 2 of 9
processing 3 of 9
processing 4 of 9
processing 5 of 9
processing 6 of 9
CGAL error: assertion violation!
Expression : v_type == PSS
File       : 
/Developer/SDKs/MacOSX10.7.sdk/usr/local/include/CGAL/Segment_Delaunay_graph_2/Voronoi_vertex_sqrt_field_new_C2.h
Line       : 1243
Explanation: 
Refer to the bug-reporting instructions at http://www.cgal.org/bug_report.html
terminate called throwing an exception




SETTINGS:

Hardware:
-MacBook Pro, 2.93 GHz Intel Core 2 Duo, 4Gb Memory.
OS, CGAL version and compiler:
-CGAL 3.9, GMP through Macports, version 5.0.2, OS X Lion (gcc version 4.2.1)



CODE:


#include <iostream>
#include <CGAL/Simple_cartesian.h>
#include <CGAL/Segment_Delaunay_graph_traits_2.h>
#include <CGAL/Segment_Delaunay_graph_2.h>
#include <CGAL/Segment_Delaunay_graph_hierarchy_2.h>


//THIS PRODUCES THE ERROR WITH THE DATA BELOW 
typedef CGAL::Gmpq   ENT;                                          //exact 
number type
typedef CGAL::Simple_cartesian<double>     CK; //construction kernel
typedef CGAL::Simple_cartesian<ENT>        EK; //exact kernel
typedef 
CGAL::Segment_Delaunay_graph_filtered_traits_without_intersections_2<CK,CGAL::Field_with_sqrt_tag,
 EK,  CGAL::Integral_domain_without_division_tag>  Gt;
typedef CGAL::Segment_Delaunay_graph_2<Gt>             SDG_2;


/*
//THIS WORKS (But it is obviously substantially slow)
typedef CGAL::Gmpq                                                      ENT;  
  //CGAL::Gmpzf and CORE::BigFloat work even faster than CGAL::Gmpq
typedef CGAL::Simple_cartesian<ENT >     Kernel;
typedef 
CGAL::Segment_Delaunay_graph_traits_without_intersections_2<Kernel,CGAL::Integral_domain_without_division_tag>
  Gt;
typedef CGAL::Segment_Delaunay_graph_hierarchy_2<Gt>  SDG_2; //this improves 
performance dramatically over Segment_Delaunay_graph_2, but is still slow
*/

typedef SDG_2::Point_2                                          Point2;

int main(int argc, char *argv[]){
        
        const int testSize=9;
        Point2 ptData[testSize]={
                Point2( 5000, 0 ), 
                Point2( 5000, 4000 ), 
                Point2( 0, 4000 ), 
                Point2( 2367, 3305 ), 
                Point2( 3854, 1366 ), 
                Point2( 3869, 1090 ), 
                Point2( 3779, 1360 ), 
                Point2( 2366, 3230 ), 
                Point2( 0, 3000 )
        };
        
        
        SDG_2   sdg;
        
        //create segment sites for the closed polygon with the points as 
vertices
        
        for (int i = 0; i < testSize; ++i) {
                std::cout<< "processing "<< i <<" of "<<testSize << std::endl;
                int next = (i + 1) % testSize;
                
                const SDG_2::Site_2 
&site=SDG_2::Site_2::construct_site_2(ptData[i], ptData[next]);
                sdg.insert(site);
        }
        
}
        
        






Thanks!