Cursor

An opaque position in a result sequence: the hits strictly past (docId, nodePre) in the order of the search that produced it — after it for SearchOrder.ASCENDING, before it for SearchOrder.DESCENDING. toString gives a compact URL-safe text, and parse reads it back. A cursor resumes only what produced it: a search of the same SearchOrder and kind, or JsonStore.objectsOfTypePage for the cursor of a TypePage; anything else throws InvalidCursorException.

A cursor used in a new search sees the store's newer state. A document whose insert started before the cursor's page was produced but was acknowledged later can land in the stretch the earlier pages already covered and is then missed; mayMissLateCommits tells whether that's possible. A descending search resumed from a cursor never returns documents above the cursor's, so documents inserted since the first page aren't returned either.

The cursor of a DocumentKeyQuery is a position in key order: it resumes strictly past the key of its last hit (docId is that hit's document, nodePre 0), so a document inserted since with a key the earlier pages passed isn't returned. Its text is at most about 1,400 characters, as it holds the key.