Implicit conversion from float -18.51409978308027 to int loses precision