“ Consider a sequence of one million values, each possibly one kilobyte when encoded -- roughly one gigabyte. It is often desirable to process such a dataset in an incremental manner without having to first read all of it before beginning to produce results. Traditionally, the way to do this with