Implicit conversion from float -4.264339152119703 to int loses precision