Small issue in Basic viewer

[pingpong74]

Wrong calculation in basic viewer

In the function initialize_vertices_and_edges_size(), the 2D case isn't handled correctly.
Source code:

void initialize_vertices_and_edges_size()
  {
    if(!m_scene.empty())
    {
      auto& bbox=m_scene.bounding_box();

      double d=is_two_dimensional()
              ?2.5
              : CGAL::sqrt(CGAL::squared_distance
                           (Local_point(bbox.xmin(), bbox.ymin(), bbox.zmin()),
                            Local_point(bbox.xmax(), bbox.ymax(), bbox.zmax())));

      m_size_vertices=1.5*d;
      m_size_edges=d;
      m_size_rays=m_size_edges;
      m_size_lines=m_size_edges;
      m_size_normals=d/3;
      m_height_factor_normals=0.02;
    }
  }

This can be fixed easily as the calculation is performed correctly in set_camera_mode() function.

Suggested change:

We can calculate d as

      double d;
      if (is_two_dimensional())
      {
        if (m_scene.has_zero_x())
          d = CGAL::sqrt(CGAL::squared_distance(
                Local_point(0, bbox.ymin(), bbox.zmin()),
                Local_point(0, bbox.ymax(), bbox.zmax())));
        else if (m_scene.has_zero_y())
          d = CGAL::sqrt(CGAL::squared_distance(
                Local_point(bbox.xmin(), 0, bbox.zmin()),
                Local_point(bbox.xmax(), 0, bbox.zmax())));
        else
          d = CGAL::sqrt(CGAL::squared_distance(
                Local_point(bbox.xmin(), bbox.ymin(), 0),
                Local_point(bbox.xmax(), bbox.ymax(), 0)));

      }
      else
      {
        d = CGAL::sqrt(CGAL::squared_distance(
              Local_point(bbox.xmin(), bbox.ymin(), bbox.zmin()),
              Local_point(bbox.xmax(), bbox.ymax(), bbox.zmax())));

        if (d < 1e-10) d = 1.0;
      }

[gdamiand]

Did you test your solution? Are you sure this solves the problem?

Please verify the new formula, and compare with the previous one...

[lrineau]

Your code, @pingpong74:

double d;
if (is_two_dimensional())
{
  if (m_scene.has_zero_x())
    d = CGAL::sqrt(CGAL::squared_distance(
          Local_point(0, bbox.ymin(), bbox.zmin()),
          Local_point(0, bbox.ymax(), bbox.zmax())));
  else if (m_scene.has_zero_y())
    d = CGAL::sqrt(CGAL::squared_distance(
          Local_point(bbox.xmin(), 0, bbox.zmin()),
          Local_point(bbox.xmax(), 0, bbox.zmax())));
  else
    d = CGAL::sqrt(CGAL::squared_distance(
          Local_point(bbox.xmin(), bbox.ymin(), 0),
          Local_point(bbox.xmax(), bbox.ymax(), 0)));

}
else
{
  d = CGAL::sqrt(CGAL::squared_distance(
        Local_point(bbox.xmin(), bbox.ymin(), bbox.zmin()),
        Local_point(bbox.xmax(), bbox.ymax(), bbox.zmax())));
}
if (d < 1e-10) d = 1.0;

that code is completely equivalent to the following one:

double d=CGAL::sqrt(CGAL::squared_distance
                           (Local_point(bbox.xmin(), bbox.ymin(), bbox.zmin()),
                            Local_point(bbox.xmax(), bbox.ymax(), bbox.zmax())));
if (!is_two_dimensional() && d < 1e-10) d = 1.0;

Let's call dx, dy, dx the the following:

auto dx = bbox.xmax() - bbox.xmin();
// and same for dy, dz, with the y and z coordinates

and then the squared distance expression is:

 dx*dx + dy*dy + dz*dz

If the scene is flat in one dimension (has_zero_x(), has_zero_y() and so on), then the corresponding dx/dy/dz is 0, and there is no need to compute a squared distance with 0 in the corresponding coordinates of the two points.

So, your contribution is only the last line:

if (!is_two_dimensional() && d < 1e-10) d = 1.0;

What is so special with 1e-10?

[pingpong74]

Yea, the code is the same as thay. The small value is added to ensure that we dont get 0 or veey small d. I am trying different functions of the bounding boxes now

[sainathmanda7]

Yes as @lrineau is Correct. I have gone through the code and i feel it is wrong.
Let's Consider a Scenario for a Single Point then xmin==xmax and ymin==ymax which results d=0 and geometry disappears completely.
To prevent from edges to scale to 0
we can fix this as:

if(d==0.0)
      d=1.0;
m_size_vertices= 1.5 *d;
m_size_edges =d;
m_size_rays= m_size_edges;
m_size_lines= m_size_edges;
m_size_normals =d/ 3.0;
m_height_normals =0.02;

[aminkhalsi]

I think the problem isn't here, but in the shaders. in GEOMETRY_SOURCE_LINE_WIDTH

void main(void)
{
  vec2 p0 = ToScreenSpace(gl_in[0].gl_Position);
  vec2 p1 = ToScreenSpace(gl_in[1].gl_Position);
  vec2 v0 = normalize(p1 - p0);
  vec2 n0 = vec2(-v0.y, v0.x) * u_PointSize * 0.5;

  // line start
  gl_Position = ToWorldSpace(vec4(p0 - n0 * gs_in[0].pointSize, gl_in[0].gl_Position.zw));
  fColor = gs_in[0].color;
  ls_fP = inverse(u_Mvp) * gl_Position;
  EmitVertex();
...

the normal n0 is being multiplied with u_PointSize which is not normalized (for very large scenes u_PointSize is very large) resulting in huge edges. This is incorrect since just in the next line n0 is multiplied by the gs_in[0].pointSize which is actually the correct size since in the previous shader it was divided by the distance. Normalizing the n0 normal results in correctly sized edges.

[aminkhalsi]

closed because duplicate of #9335