Thanks, @Enrico Parisi, that looks a good starter for the kind of thing we are exploring.  

Thanks for the link, but as we are stuck on Ensemble (not Iris), we can't make use of this.

Thanks, that's helpful. Our end goal was the perennial XML to JSON conversion conundrum, or HL7 to JSON, and as we're currently stuck on Ensemble 2018, we've not got access to %JSON.Adapter, so we're... limited.

Our immediate goal was, given a particular message/document, visit all nodes holding data within the document (traverse the tree), and do some processing for each node - eg output a JSON representation of the data held at that node so that a JSON serialisation of the whole document can be produced; but the general case, not just the JSON case, is of interest.

Current application is taking an HL7 feed and sending (a subset of) the data in the HL7 and sending to be consumed by a downstream system via a REST web-service. The downstream in this case is being developed by another team within our organisation, and they can take XML - we've built a transformation to extract the relevant data into an EnsLib.EDI.XML.Document, and we've developed an XML schema for the data so we can validate it before sending on. But it would have been easier for them to consume JSON, so we were exploring how to do that - and are conscious that future requirements might include sending JSON rather than XML...