would you accept a patch to make the point-type flexible? So instead of forcing you to use a float, you could use an int (or something else).
Probably the most straight forward implementation would be to make the current lib a functor -- and providing a float and an int applied one.