Implicit conversion from float -137.0053475935829 to int loses precision