[Mifluz-dev] Question re. word data

[Steven J. DeRose]

I've been looking at the MiFluz doc, and it looks quite nice. I'm 
wondering what it would take to enhance it to support storing the 
scope or extent of a token, rather than just a single integer

The goal for me would be to make it able to do containment and 
structure queries on hierarchical data, particularly mbox files and 
XML. By storing a "word" record with start and end offset for each 
XML element, or each MIME mail message (and maybe each MIME header 
line), mifluz could find words only when they occur in a certain 
context.

A quick look suggests this would mainly involve adding a new record 
type besides DATA and STRING, and adding the necessary APIs to get 
them in, found, and out.

I've built systems like this before that scaled into the 100s of MB 
per document, so I know most of the general constraints, but I don't 
know anything about the internals of Mifluz (yet). Does this sound 
like a feasible enhancement and approach?

Any advice appreciated.

Thanks!

--
Steve DeRose