Implicit conversion from float -31.839080459770116 to int loses precision