Better handling of degenerate convex hull cases.

Contributes to CURA-1504

1 files changed, 42 insertions, 40 deletions

diff --git a/cura/ConvexHullDecorator.py b/cura/ConvexHullDecorator.py
index c42ab4f339..2a8f55d09e 100644
--- a/cura/ConvexHullDecorator.py
+++ b/cura/ConvexHullDecorator.py
@@ -130,46 +130,48 @@ class ConvexHullDecorator(SceneNodeDecorator):
             return rounded_hull
 
         else:
-            if not self._node.getMeshData():
-                return None
-            mesh = self._node.getMeshData()
-            world_transform = self._node.getWorldTransformation()
-
-            # Check the cache
-            if mesh is self._2d_convex_hull_mesh and world_transform == self._2d_convex_hull_mesh_world_transform:
-                # Logger.log('d', 'Cache hit in _compute2DConvexHull mesh path')
-                return self._2d_convex_hull_mesh_result
-
-            vertex_data = mesh.getConvexHullTransformedVertices(world_transform)
-            # Don't use data below 0.
-            # TODO; We need a better check for this as this gives poor results for meshes with long edges.
-            vertex_data = vertex_data[vertex_data[:,1] >= 0]
-
-            # Round the vertex data to 1/10th of a mm, then remove all duplicate vertices
-            # This is done to greatly speed up further convex hull calculations as the convex hull
-            # becomes much less complex when dealing with highly detailed models.
-            vertex_data = numpy.round(vertex_data, 1)
-
-            vertex_data = vertex_data[:, [0, 2]]  # Drop the Y components to project to 2D.
-
-            # Grab the set of unique points.
-            #
-            # This basically finds the unique rows in the array by treating them as opaque groups of bytes
-            # which are as long as the 2 float64s in each row, and giving this view to numpy.unique() to munch.
-            # See http://stackoverflow.com/questions/16970982/find-unique-rows-in-numpy-array
-            vertex_byte_view = numpy.ascontiguousarray(vertex_data).view(
-                numpy.dtype((numpy.void, vertex_data.dtype.itemsize * vertex_data.shape[1])))
-            _, idx = numpy.unique(vertex_byte_view, return_index=True)
-            vertex_data = vertex_data[idx]  # Select the unique rows by index.
-
-            hull = Polygon(vertex_data)
-
-            # First, calculate the normal convex hull around the points
-            convex_hull = hull.getConvexHull()
-
-            # Then, do a Minkowski hull with a simple 1x1 quad to outset and round the normal convex hull.
-            # This is done because of rounding errors.
-            rounded_hull = convex_hull.getMinkowskiHull(Polygon(numpy.array([[-0.5, -0.5], [-0.5, 0.5], [0.5, 0.5], [0.5, -0.5]], numpy.float32)))
+            rounded_hull = None
+            if self._node.getMeshData():
+                mesh = self._node.getMeshData()
+                world_transform = self._node.getWorldTransformation()
+
+                # Check the cache
+                if mesh is self._2d_convex_hull_mesh and world_transform == self._2d_convex_hull_mesh_world_transform:
+                    # Logger.log('d', 'Cache hit in _compute2DConvexHull mesh path')
+                    return self._2d_convex_hull_mesh_result
+
+                vertex_data = mesh.getConvexHullTransformedVertices(world_transform)
+                # Don't use data below 0.
+                # TODO; We need a better check for this as this gives poor results for meshes with long edges.
+                vertex_data = vertex_data[vertex_data[:,1] >= 0]
+
+                if len(vertex_data) >= 4:
+                    # Round the vertex data to 1/10th of a mm, then remove all duplicate vertices
+                    # This is done to greatly speed up further convex hull calculations as the convex hull
+                    # becomes much less complex when dealing with highly detailed models.
+                    vertex_data = numpy.round(vertex_data, 1)
+
+                    vertex_data = vertex_data[:, [0, 2]]  # Drop the Y components to project to 2D.
+
+                    # Grab the set of unique points.
+                    #
+                    # This basically finds the unique rows in the array by treating them as opaque groups of bytes
+                    # which are as long as the 2 float64s in each row, and giving this view to numpy.unique() to munch.
+                    # See http://stackoverflow.com/questions/16970982/find-unique-rows-in-numpy-array
+                    vertex_byte_view = numpy.ascontiguousarray(vertex_data).view(
+                        numpy.dtype((numpy.void, vertex_data.dtype.itemsize * vertex_data.shape[1])))
+                    _, idx = numpy.unique(vertex_byte_view, return_index=True)
+                    vertex_data = vertex_data[idx]  # Select the unique rows by index.
+
+                    hull = Polygon(vertex_data)
+
+                    if len(vertex_data) >= 4:
+                        # First, calculate the normal convex hull around the points
+                        convex_hull = hull.getConvexHull()
+
+                        # Then, do a Minkowski hull with a simple 1x1 quad to outset and round the normal convex hull.
+                        # This is done because of rounding errors.
+                        rounded_hull = convex_hull.getMinkowskiHull(Polygon(numpy.array([[-0.5, -0.5], [-0.5, 0.5], [0.5, 0.5], [0.5, -0.5]], numpy.float32)))
 
             # Store the result in the cache
             self._2d_convex_hull_mesh = mesh