Implicit conversion from float -22.55474452554745 to int loses precision