Based on a patch from Xyne, this should be a slightly less verbose way of safely calculating normals. Note that CGAL does have a unit_normal() function, but since the Gmpq kernel doesn't support sqrt(), this cannot be used.

1 files changed, 16 insertions, 24 deletions

diff --git a/src/export.cc b/src/export.cc
index 5ce2d15..5ac3ac0 100644
--- a/src/export.cc
+++ b/src/export.cc
@@ -83,32 +83,24 @@ void export_stl(CGAL_Nef_polyhedron *root_N, std::ostream &output, QProgressDial
 			stream.str("");
 			stream << x3 << " " << y3 << " " << z3;
 			std::string vs3 = stream.str();
+			CGAL_Polyhedron::Traits::Vector_3 normal(1,0,0);
 			if (vs1 != vs2 && vs1 != vs3 && vs2 != vs3) {
-				// The above condition ensures that vs1-vs2, vs1-vs3, and their cross
-				// product are non-zero. Floating point arithmetic may however truncate
-				// small values to 0. This can be avoided by first scaling the components
-				// of vs1-vs2 and vs1-vs3. This has no effect on the resulting unit
-				// normal vector.
-				double dn[6] = { x1-x2, y1-y2, z1-z2, x1-x3, y1-y3, z1-z3 };
-				double maxdn = 0;
-				int i;
-				for (i = 0; i < 6; ++i) {
-					double dx = dn[i];
-					if (dx < 0) dx = -dx;
-					if (dx > maxdn) maxdn = dx;
-				}
-				for (i = 0; i < 6; ++i) dn[i] /= maxdn;
-				double nx = dn[1]*dn[5] - dn[2]*dn[4];
-				double ny = dn[2]*dn[3] - dn[0]*dn[5];
-				double nz = dn[0]*dn[4] - dn[1]*dn[3];
-				double nlength = sqrt(nx*nx + ny*ny + nz*nz);
-				// Avoid generating normals for polygons with zero area
-				double eps = 0.000001;
-				if (nlength < eps) nlength = 1.0;
+				// The above condition ensures that there are 3 distinct vertices, but
+				// they may be collinear. If they are, the unit normal is meaningless
+				// so the default value of "1 0 0" can be used. If the vertices are not
+				// collinear then the unit normal must be calculated from the
+				// components.
+				if (!CGAL::collinear(v1.point(),v2.point(),v3.point())) {
+					// Pseudocode: CGAL kernel must be set up to enable unit_normal and
+					// Vector type must be declared as Vector_3<Kernel>.
+					// http://www.cgal.org/Manual/latest/doc_html/cgal_manual/Kernel_23_ref/Function_unit_normal.html
+					normal = CGAL::normal(v1.point(),v2.point(),v3.point());
+					normal = normal / sqrt(CGAL::to_double(normal.squared_length()));
+				}	
 				output << "  facet normal "
-							 << nx / nlength << " "
-							 << ny / nlength << " "
-							 << nz / nlength << "\n";
+							 << CGAL::to_double(normal.x()) << " "
+							 << CGAL::to_double(normal.y()) << " "
+							 << CGAL::to_double(normal.z()) << "\n";
 				output << "    outer loop\n";
 				output << "      vertex " << vs1 << "\n";
 				output << "      vertex " << vs2 << "\n";