Implicit conversion from float -2.142857142857139 to int loses precision