Implicit conversion from float -87.68595041322314 to int loses precision