Implicit conversion from float -107.22222222222223 to int loses precision