Implicit conversion from float -34.45820433436532 to int loses precision