The class Snap_rounding_traits_2<Kernel> is a model of the SnapRoundingTraits_2 concept, and is the only traits class supplied with the package. This class should be instantiated with an exact geometric kernel that conforms to the CGAL kernel-concept, such as the Cartesian<gmpq> kernel.
This geometric kernel must provide an (arbitrary-precision) rational number type (FT), Point_2, Segment_2 and Iso_rectangle_2. It should be possible to cast numbers of the number type FT to double-precision representation. That is, the function CGAL::to_double(FT) must be supported.
The CGAL::to_double() function is used to implement the operation that rounds the coordinates of a point to a center of a pixel. This operation is one of the traits-concept requirement. The coordinates are converted to double, rounded down to the nearest grid point, and finally adjusted to lie on a center of a pixel. Notice that if CGAL::to_double() returns the closet value, then when it rounds up a given coordinate, the resulting ISR, may be imprecise, and the distance between some vertex and a non-incident edge can be slightly less than the guaranteed half-the width-of-a-pixel.