Implicit conversion from float -43.45323741007195 to int loses precision