Previously there was an algirithm that was fitting a curve through
the resulting seam points. This worked for smooth models, but was
failing for some very basic cases (such as a cylinder).
The new algorithm builds on the previously implemented visibility
algirithm but does not do the curve fitting anymore.
Now the code is more separated for the four seam placement
options (rear, random, aligned, nearest).
Nearest and random are handled as one would expect.
Aligned is handled in a more "brute force" manner
(trying multiple seams and picking the best one)
and rear smartly switches between two modes
(center straight line projection and max y).
The temporary variable used for reordering ExPolygons wasn't cleared, and it contained empty ExPolygons from previous reordering. This caused those empty ExPolygons replaced actual ExPolygons for infill.
Another issue was that m_fill_expolygons was reordered, but m_fill_expolygons_bboxes were left untouched.
Reworked (again!) connecting of islands into a Z-graph.
Implemented various heuristics to handle self-intersecting and
mutually intersecting ExPolygons on the same layer.
Follow-up to 52ea2edf842e81e0f81fcd8303b69d288d903ae3
1) There was a bug in accessing the "perimeter is external" property,
ExtrusionCollection returns "mixed", the embedded ExtrusionPath
has to be queried directly.
2) The search bounding box has to be extended by the maximum offset
introduced by fuzzy skin algorithm. For Arachne the fuzzy skin
algorithm observes fuzzy_skin_point_dist.
1) An extrusion sample is taken for sorting extrusions into island
so that a sample deep inside its island is taken with high probability.
2) With fuzzy skin active, the inexact search is done with bounding boxes
inflated with the fuzzying distance.
1) New region expansion code to propagate wave from a boundary
of a region inside of it.
2) get_extents() extended with a template attribute to work with
zero area data sets.
3) ClipperZUtils.hpp for handling Clipper operation with Z coordinate
(for source contour identification)
with int64s instead of doubles.
Polygon / ExPolygon: contains() reworked to use ClipperLib::PointInPolygon().
The Slic3r own implementation was not robust.
Fixed test_perimeters after recent refactoring (sorting of extrusions
into LayerIslands)
of Layer->LayerSlice->LayerIsland hierarchy. This should improve
tool path ordering of multiple parts within the same object #5511.
Some shells tests rewritten from Perl to C++.
FIXME: Gap fill extrusions are currently not handled by the initial
G-code preview!
into a graph with links to the layer above / below.
In addition:
Members of LayerRegion were made private, public interface const only.
this->m_xxx replaced with just m_xxx
SurfacesPtr was made a vector of const pointers.
For this particular model, ClipperLib numerical instability causes
one of the internal surfaces to turn into bridging surfaces
on reslicing. The issue was fixed by reverting to untyped slices if possible.
1) Areas inside modifier meshes were ironed multiple times.
2) Ironing areas were not properly merged.
Layer::lslices were not always properly merged with modifier meshes
applied, which lead to the ironed surface being split and not fully
ironed, as there were artificial gaps created between regions as if
they were covered by perimeters (we don't iron over perimeters).
1) Removed the already commented-out scaling / unscaling when doing
"safe offsetting"
2) Removed some of the "safe offsetting" at calls where it never was used.
3) Reworked Clipper & ClipperUtils to pass Polygons / ExPolygons / Surfaces
as input parameters without conversion to ClipperLib::Paths. This
should save a lot of memory allocation and copying.
4) Reworked conversions from ClipperLib::Paths & PolyTree to Polygons /
ExPolygons to use the move operator to avoid many unnecessary allocations.
5) Reworked some "union with safe ofsetting" to "offset_ex", which should
be cheaper.
