Comment by alaroldai

For data that isn't part of the document, you could use a write-through wrapper around the rope, with a secondary data structure mapping ranges of the document to this extra data.

From the wrapper's point of view, there's no difference between character and non-character data, and the whole buffer can be modeled as a collection of indices mapping ranges of the document to different kinds of data.

One of those indices could be a rope (mapping document ranges to character data, for the document text). Other kinds of indices could also be used. The important thing is that all edits go through the wrapper so that all the relevant indices get updated.