Implicit conversion from float -22.451108213820078 to int loses precision