A "real" type with sufficient precision for coordinates.

You may safely assume that double (or even float) provides enough precision for storing on-canvas points, and hence that double provides enough precision for dot products of differences of on-canvas points.

Definition at line 13 of file nr-coord.h.