Implicit conversion from float -7.870528109028953 to int loses precision