Implicit conversion from float -22.878787878787875 to int loses precision