Implicit conversion from float -22.898936170212764 to int loses precision