Implicit conversion from float -1.3138686131386947 to int loses precision