It would be nice to report on how many IRIs, bnodes, literals, repeated terms, and so on are in the stream. Possibly also grouped by term (subject, predicate, object, graph).

The size tracking option (#74) should also work with this.