How to know whether Gt::Object_2 is a site or a line? (2D Apollonius diagram)

[stla]

Hello,

As written in the doc, the dual of a face of an Apollonius graph has type Gt::Object_2 and can be assignable to Site_2 or Gt::Line_2. Currently in my code I do Site2 site = ag.dual(f); because I know the dual of each face is a site for my example (because the weights are distinct).

But how to do in general? That is, I'm looking for something like

GT::Object_2 d = ag.dual(f);
if(typeof(d) == Site2) { 
  ......
} else {
  ......
}

I believe I have to use Assign_2 but I don't understand how it works.

[stla]

My attempts which do not work:

    Object2 obj = ag.dual(f);
    Site2 site;

    if(Traits::assign_2_object(site, obj)) {
      Point2 pt  = site.point();
      DualPoints(fid, 0) = pt.x();
      DualPoints(fid, 1) = pt.y();
    } else {
      // ...
    }

    Site2 site;

    if(Traits::assign_2_object(site, ag.dual(f))) {
      Point2 pt  = site.point();
      DualPoints(fid, 0) = pt.x();
      DualPoints(fid, 1) = pt.y();
    } else {
      // ...
    }

[stla]

This compiles so that probably works:

    Traits traits = ag.geom_traits();
    Traits::Construct_object_2 cst = traits.construct_object_2_object();
    Object2 obj = cst(ag.dual(f));
    Traits::Assign_2 assgn = traits.assign_2_object();
    Site2 site;

    if(assgn(site, obj)) {
      Point2 pt  = site.point();
      DualPoints(fid, 0) = pt.x();
      DualPoints(fid, 1) = pt.y();
    } else {
      // ...
    }

[lrineau]

This compiles so that probably works:

    Traits traits = ag.geom_traits();
    Traits::Construct_object_2 cst = traits.construct_object_2_object();
    Object2 obj = cst(ag.dual(f));

ag.dual(f) is already of type Traits::Object_2. The last line created a new Object_2 from an existing object of the same type: that creates a copy. Harmless, but useless. You can write instead:

Traits::Object_2 obj = ag.dual(f);

And then:

    Traits::Assign_2 assgn = traits.assign_2_object();
    Site2 site;

    if(assgn(site, obj)) {

that is the correct API.

[lrineau]

Actually, the source of incomprehension is because the doc and code do not agree.

Here are the real signatures for CGAL::Apollonius_graph_2::dual:

cgal/Apollonius_graph_2/include/CGAL/Apollonius_graph_2.h

Lines 525 to 529 in 6098856

	// ACCESS TO THE DUAL
	//-------------------
	typename Gt::Object_2 dual(const Face_handle& f) const;
	Site_2 dual(const Finite_faces_iterator& it) const

See that the version from Finite_faces_iterator does not return an Object_2 but directly a Site_2.

[stla]

Are you sure? Traits::Object_2 obj = ag.dual(f); is what I originally did, and this didn't compile (no conversion between Site_2 and Object_2. And the code confirms what I say. Should we convert the Finite_faces_iterator f to a Face_handle?

[stla]

Cross posted. Yes I understand now. I can open an issue for the problem in the doc.

[stla]

Something weird happens. I replaced ag.finite_faces with ag.all_faces and then assgn(site, obj) is never true, as well as assgn(line, obj). Will try to do a small reproducible example.

[stla]

And std::cout << ag.dual(f) always prints 1.

[stla]

I checked https://github.com/CGAL/cgal/blob/master/Apollonius_graph_2/include/CGAL/Apollonius_graph_2.h. There's no method dual defined for All_faces_iterator. The doc claims there is one.