Implicit conversion from float -108.2484076433121 to int loses precision