CGAL users discussion list

[Efi Fogel <>]

Hi David,

Thanks for pointing out all these issues and for the patches you keep 
sending.

I've quickly looked at the patch and it seems correct. I guess one can 
provide a more efficient implementation using the same idea of comparing 
the y-coordinate of a point in one curve and a segment in the other, but 
advancing in both curves in parallel in a controlled manner.

We saw your posting from last week regarding the IO problem in the 
Arrangement_2 package. We are looking into it, and we'll try to give you 
a detailed answer soon.

David Keller wrote:
> Hi,
>
> You find a patch to fix a bug in the equality-comparison of 
> polyline-arrangements as reported in my post on 03/14: The problem is 
> caused by the segment-wise comparison of polylines which does not take 
> into account that polylines with different representation may share 
> the same graph. Maybe someone is interested to check its validity or 
> even give a more efficient implementation.
>
> I tried to hunt down the bugs observed in IO of Arrangement_2 reported 
> in my last post. There are seemingly different inconsistencies in the 
> current implementation:
>
>   1. Unbounded faces are not read correctly. See my post from 03/28 for
>      a reproducer.
>   2. CGAL/Arr_accessor.h (1037) in function clear_all() : The call
>      p_arr->dcel.delete_all() removes DCEL-entries in the current
>      arrangement. This leads to dangling pointers in Arrangement_2,
>      i.e. the fictitious face and the vertices (un_face, v_br, v_bl,
>      v_tr, v_tl). The function is called from
>      CGAL/IO/Arrangement_2_reader.h (107). From my understanding these
>      pointer members in Arrangement_2 are not set when reading an
>      arrangement leading to an uncontrolled behavior of the
>      reconstructed arrangement.
>   3. An empty arrangement (result of default constructor of
>      Arrangement_2) contains one fictitious face (actually 2 DCEL
>      entries), 4 fictitious vertices and 4 fictitious edges. The
>      CGAL/IO/Arrangement_2_writer.h writes 0 fictitious vertices, 4
>      fictitious edges and 2 faces. It is not possible to read an empty
>      arrangement with the current implementation (comment Line 22 in
>      examples/Arrangement_2/io.cpp for a demonstration).
>
> Can somebody sketch a solution to these problems?
> Thanks and best regards
> David Keller

--
  ____  _        ____             _
 /_____/_) o    /__________  __  //
(____ (   (    (    (_/ (_/-(-'_(/
                        _/