Implicit conversion from float -0.20134228187919234 to int loses precision