Implicit conversion from float -22.222222222222214 to int loses precision