Implicit conversion from float -18.628158844765323 to int loses precision