Implicit conversion from float -7.093791281373839 to int loses precision