Implicit conversion from float -43.31043956043956 to int loses precision