<   one column or expression indexes, perhaps using per=index statistics
>   one column or expression indexes, perhaps using per-index statistics