Implicit conversion from float -22.23776223776224 to int loses precision