Skip to Content.
Sympa Menu

cgal-discuss - Re: [cgal-discuss] lower_hull_2() problem

Subject: CGAL users discussion list

List archive

Re: [cgal-discuss] lower_hull_2() problem


Chronological Thread 
  • From: Alessandro Attanasi <>
  • To: Philipp Moeller <>
  • Cc:
  • Subject: Re: [cgal-discuss] lower_hull_2() problem
  • Date: Fri, 17 Jan 2014 23:32:15 +0100

Thanks Philipp, I'm going to use the library as suggested by you :)


On 16 January 2014 23:23, Philipp Moeller <> wrote:
Alessandro Attanasi <> writes:

> Hi Philipp,
>
> I understood the behaviour with the STL, I was wondering about C way usage
> of CGAL as the following , that is working fine even if I used malloc() and
> free() while for the function lower_hull_points_2() a similar approach
> doesn't work at all, I was expecting the same behaviour but I missing
> something.

This has the *appearance* of working. The code causes undefined behavior
for reasons I pointed out before. I also don't understand what you mean
by "C way usage". CGAL is a C++ library. There is no straight-forward
way to use it from C.

> // definition of the input points
>
>   K::Point_2 p1(2,5);
>
>   K::Point_2 p2(0,4);
>
>   K::Point_2 p3(0,0);
>
>   K::Point_2 p4(4,0);
>
>   K::Point_2 p5(4,4);
>
>   K::Point_2 p6(1,3);
>
>   K::Point_2 p7(1,1);
>
>   K::Point_2 p8(3,1);
>
>   K::Point_2 p9(3,3);
>
>   K::Point_2 p10(2,2);
>
>   int totPoints = 10;
>
>  // array storing input points
>
>   K::Point_2 *points = NULL;
>
>   points = (K::Point_2*)malloc(totPoints * sizeof(K::Point_2));
>
>
>
>   points[0] = p1;
>
>   points[1] = p2;
>
>   points[2] = p3;
>
>   points[3] = p4;
>
>   points[4] = p5;
>
>   points[5] = p6;
>
>   points[6] = p7;
>
>   points[7] = p8;
>
>   points[8] = p9;
>
>   points[9] = p10;
>
>
>
>   // array storing results
>
>   K::Point_2 *result = NULL;
>
>   result = (K::Point_2*)malloc(totPoints * sizeof(K::Point_2));
>
>
> ptr = CGAL::ch_akl_toussaint( points, points+totPoints, result );
>
>
>  // see how many points are into the convex hull
>
>   int ch_points = (ptr - result);
>
>
>   // print out the number of points of the convex hull and the points
> themselves
>
>   std::cout << "1) Points on the convex hull = " << ch_points << std::endl;
>
>   for(int i =0;i<ch_points;i++)
>
>   {
>
>     std::cout << result[i] << std::endl;
>
>   }
>
>
>  free(points);
>
>   points = NULL;
>
>
>
>   free(result);
>
>   result = NULL;
>
>
>
>
> On 15 January 2014 23:58, Philipp Moeller <
> > wrote:
>
>> Alessandro Attanasi <> writes:
>>
>> > Ok, but why that procedure was working using all convex_hull
>> > functions?
>>
>> It was working because you were using std::vector with a back_inserter
>> which takes care of those things. At least this is what your code
>> did. There can be situations where using a malloc'd pointer as an
>> OutputIterator can work (with a POD-type assignment effectively becomes
>> a memcpy, IIRC this is still undefined behavior though).
>>
>> > I started from the cgal example
>> >
>> http://doc.cgal.org/latest/Convex_hull_2/Convex_hull_2_2array_convex_hull_2_8cpp-example.html
>> >
>> > I freed the malloc at the end :)
>>
>> Which is exactly the wrong thing to do because it won't call the
>> destructors of the elements.
>>
>> >
>> >
>> >
>> > On 15 January 2014 23:31, Philipp Moeller <
>> > > wrote:
>> >
>> >> Alessandro Attanasi <> writes:
>> >>
>> >> > Sorry, really stupid error :P. It works as expected!
>> >> >
>> >> > Just last question about this problem: if I want to use C pointers
>> >> instead
>> >> > of stl iterators, why this code is wrong? It gives to me zero points
>> on
>> >> the
>> >> > lower hull (I tried to use the same logic of the example for convex
>> hull
>> >> > computation)
>> >> >
>> >> > K::Point_2 *ptr2;
>> >> >
>> >> >   K::Point_2 *result2 = NULL;
>> >> >
>> >> >   result2 = (K::Point_2*)malloc(totPoints * sizeof(K::Point_2));
>> >> >
>> >> >   ptr2 = CGAL::lower_hull_points_2( points, points+totPoints, result2
>> );
>> >>
>> >> Why would you ever call malloc instead of new in such a situation? It is
>> >> exactly where your problem comes from. You cannot simply use raw
>> >> uninitialized memory as an OutputIterator. You should wrap the pointer
>> >> into a raw_storage_iterator provided by the standard header memory. [1]
>> >>
>> >> You also need to take special care when releasing the memory (call the
>> >> destructor explicitly).
>> >>
>> >> >
>> >> >
>> >> >   ch_points = (ptr2 - result2);
>> >> >
>> >> >   std::cout << "6) Points on the lower hull = " << ch_points <<
>> >> std::endl;
>> >> >
>> >> >   for(int i =0;i<ch_points;i++)
>> >> >
>> >> >   {
>> >> >
>> >> >     std::cout << result2[i] << std::endl;
>> >> >
>> >> >   }
>> >>
>> >> Footnotes:
>> >> [1]  http://en.cppreference.com/w/cpp/memory/raw_storage_iterator
>>

--
Philipp Moeller
GeometryFactory



--






Archive powered by MHonArc 2.6.18.

Top of Page