Implicit conversion from float -107.44058500914079 to int loses precision