Implicit conversion from float -89.70049916805323 to int loses precision