Implicit conversion from float -2.327586206896555 to int loses precision