Parent axis axis is simple to handle, by returning the longest prefix of your current paths if it matches the NodeTest, with special attention to descendant axis prefixes which need a little extra handling. Ancestor axis is the recursive application of parent axis. Following sibling and preceding sibling are both easy enough to handle. Following and preceding are the only two that cause real problems - and they really just mean that document projection can't cull any of the nodes for a particular document used in the query. Other documents in the same query can still be projected successfully. I often find code easier to understand, so if you're interested drop me an email and I can point you to the DB XML code that already does this.