Implicit conversion from float -0.33546325878594985 to int loses precision