boostorg
bug: geometry::buffer returns an empty multi-polygon in a very specific case
BOOST v1.84 / v1.85
Using the following code, lOuts is empty in the end:
boost::geometry::model::polygon<boost::geometry::model::d2::point_xy<double>> lPolygon;
lPolygon.outer().push_back({730.35, 750.00});
lPolygon.outer().push_back({730.35, 740.00});
lPolygon.outer().push_back({726.02, 740.00});
lPolygon.outer().push_back({726.02, 735.00});
lPolygon.outer().push_back({730.35, 735.00});
lPolygon.outer().push_back({730.35, 0.00});
lPolygon.outer().push_back({0.00, 0.00});
lPolygon.outer().push_back({0.00, 750.00});
lPolygon.outer().push_back({730.35, 750.00});
static const boost::geometry::strategy::buffer::side_straight side_strategy;
const boost::geometry::strategy::buffer::distance_symmetric<double> distance_strategy(5);
boost::geometry::model::multi_polygon<boost::geometry::model::polygon<boost::geometry::model::d2::point_xy<double>>> lOuts;
static const boost::geometry::strategy::buffer::join_miter join_strategy;
static const boost::geometry::strategy::buffer::end_flat end_strategy;
static const boost::geometry::strategy::buffer::point_circle point_strategy;
boost::geometry::buffer(lPolygon, lOuts, distance_strategy, side_strategy, join_strategy, end_strategy, point_strategy);
The shape looks like this:
If the small hole is moved a bit to the bottom, the boost function seems to work well. This is a very specific case that crashes in our code. Moreover, we need to keep "rectangular/cartesian" shapes (we cannot have arcs in the final shapes).
Do you guys have any clue of how to bypass this particular problem? Is this a wrong usage from on our side or something that Boost can fix directly?
If you need any detail, feel free to ask :)
Ugly workaround: add 1.0e-12 to the distance.
Smaller reproduction wkt: POLYGON((130.35 150,130.35 140,126.02 140,126.02 135,130.35 135,130.35 0,0 0,0 150,130.35 150))
It generates colocated turns. The green ones on the right side are all detected OK. But the top two are colocated (therefore you see only 2) and that causes the problem.
To be continued.
This was fixed in July 30 (but the issue was not closed automatically, somehow - and not manually). PR #1300 Closing.