one big picture approach to try: - use normal libraries to turn xml into a tree of clojure data (eg. clojure.data.xml), then use tree-seq plus filter to find subtrees matching some pattern. Using the args to tree-seq you can "prune" subtrees (eg. if the parent is a comment form, not search into the child nodes for data, only look for certain tags inside specific relevant parent tags...)

that said I find xml frustrating (I think we all do), and I've never been fully satisfied with how I've handled xml data

You might find Tim Baldridge's Odin useful

Thanks for the tips. Last idea I had was using a zipper to walk depth first through the doc, keeping track of where you are in the tree and matching on patterns like you said @U051SS2EU. That way I can build up the simpler tree I want to get out of the data.

yeah - the difference between a zipper and tree-seq is with tree-seq you get every subtree in a lazy-seq, and can filter out the ones you think are relevant based on the full path, and with a zipper you have a place oriented navigation API, both can accomplish the same thing, I am prejudiced toward values over places

I think I see what you mean using tree-seq, although I guess it could be quite finicky to rebuild parts of the tree. But I will try again tomorrow

yeah - that approach is about extracting values, rather than transforming the tree

Using XPath (e.g. via https://github.com/eerohele/sigel) might also be an option.