Implicit conversion from float -93.23863636363637 to int loses precision