I am having trouble slicing a Manhattan (rectilinear) polygon into rectangles using the get_rectangles(output_container_type& output, const T& polygon_set)
method in Boost.Polygon. It does not seem to work on self-intersecting polygons, such as the one below.
Here's my attempt:
#include <iostream>
#include <boost/polygon/polygon.hpp>
#include <vector>
namespace gtl = boost::polygon;
using namespace boost::polygon::operators;
int main() {
typedef gtl::polygon_90_with_holes_data<int> Polygon;
typedef gtl::polygon_traits<Polygon>::point_type Point;
Point pts[] = {gtl::construct<Point>(0, 0), // See the image
gtl::construct<Point>(0, 10),
gtl::construct<Point>(30, 10),
gtl::construct<Point>(30, 20),
gtl::construct<Point>(10, 20),
gtl::construct<Point>(10, 0)};
Polygon poly;
gtl::set_points(poly, pts, pts+6);
std::vector< gtl::rectangle_data<int> > rects;
get_rectangles(rects, poly);
std::cout << rects.size() << " rectangle: \n";
for(std::vector<gtl::rectangle_data<int> >::iterator it = rects.begin(); it !=
rects.end(); ++it) {
// Print out the corner coordinates
std::cout << "x1: "<< gtl::xl(*it) << ", x2: " << gtl::xh(*it)
<< ", y1: "<< gtl::yl(*it) << ", y2: " << gtl::yh(*it) << std::endl;
}
return 0;
}
And here's the output:
1 rectangle:
x1: 10, x2: 30, y1: 10, y2: 20
This method worked for a non-intersecting polygon, but here it only produces one rectangle, not two. This seems to indicate that what I am trying to do should be possible. What am I doing wrong, or is this possible?
Your code is "stress-testing" the polygon code. Strictly speaking, your list of points does not define a polygon. In theory, a planar region (a polygon shape in particular) is well defined only by a so-called Jordan (simple) curve that forms the perimeter. In the simplest terms, a Jordan (simple) curve must not cross itself.
When the same shape is redefined using a Jordan (simple) curve with the following list of 8 points,
your code produces the correct result:
2 rectangle: x1: 0, x2: 10, y1: 0, y2: 10 x1: 10, x2: 30, y1: 10, y2: 20