Note that this is applicable to any problem where you want to perform updates on a key value store, and the updates are associative (aka a semigroup). So you can do fast wordcount this way on massive documents. Or your updates might be updates to individual fields of a row of a table. Or your updates might add and remove items from a sorted list, allowing documents to be deleted from as well as inserted into an inverted index.

The nice thing is that you don’t have to retrieve a key in order to update it. You defer the update to the most convenient time.

http://www.logarithmic.net/pfh/blog/01244327900