ProcessedDocument currently allows users to do limited manipulation of documents at a low level (ie, by adding terms and values), but doesn't allow "fields" to be added or removed from the document.

Implementing add_field() should be very easy, and should certainly be done. (The only difficulty could involve spelling correction, where a connection to the database is required by the term generator - but possibly this is a bad approach, anyway.)

There should also be remove_field() (which takes the value to remove, generates the appropriate terms, and removes them from the document instead of adding them), and clear_field() (which removes all information about a particular field from the document).