Implicit conversion from float -5.089285714285708 to int loses precision