Implicit conversion from float -19.45229681978799 to int loses precision