Implicit conversion from float -22.200000000000003 to int loses precision