Implicit conversion from float -0.526315789473685 to int loses precision