Implicit conversion from float -3.559322033898326 to int loses precision