Implicit conversion from float -87.94376098418277 to int loses precision