Yup. There is a weird little bug in numpy... sort of. You only need to watch out for it when subclassing numpy arrays (assuming that the subclasses you are using have avoided it.) Check out my stack overflow post about it. I'm not 100% sure, but it might just be a Python 2 thing.