Implicit conversion from float -87.71084337349399 to int loses precision