Implicit conversion from float -18.22784810126582 to int loses precision