Implicit conversion from float -20.078740157480297 to int loses precision