Thanks

(I'm only talking about text indices here.)

Note that the Catalog is actually a tree of database/cache objects. There are a series of buckets and sub-buckets that end in a word object. The buckets, sub-buckets, and words are all database objects and can be cached/updated separately.

Put another way, every branch and leaf on the tree in a Catalog is a separately stored and cached database object.

The 1,027 words in the document (and the branches on the tree to get to them) that was updated would obey the cache parameters.

Hmm, that is a significant optimization, and one I was not aware of.

What problem are you trying to solve -- response time, memory usage, disk usage?

All of the above :)