Implicit conversion from float -22.436974789915965 to int loses precision