Implicit conversion from float -74.3686502177068 to int loses precision