Implicit conversion from float -109.09090909090907 to int loses precision