Implicit conversion from float -7.023809523809533 to int loses precision