Implicit conversion from float -37.072538860103634 to int loses precision