Implicit conversion from float -17.784431137724553 to int loses precision