Implicit conversion from float -4.682320441988935 to int loses precision