Implicit conversion from float -19.335038363171364 to int loses precision