Implicit conversion from float -105.1219512195122 to int loses precision