Implicit conversion from float -20.589812332439678 to int loses precision