Implicit conversion from float -5.588235294117652 to int loses precision