Implicit conversion from float -5.478260869565219 to int loses precision