Implicit conversion from float -7.142857142857139 to int loses precision