And you should then use XSLT to transform the docbook-generated html to Qt html. It is designed for that task, no weird Python regexes and string-splitting and indentation-tracking required. IOW, choosing the right tool for the job.
I should note that XPath is really the key here: it’s an elegant language to get and transform tree-structured and sequential data, so it could be well-used in a Pythonic template pipeline.
XSLT is a functional programming language that leans heavily on XPath to support transformation of tree-based structures. XQuery is another in the same vein. But you can accomplish the same things using an imperative language like Python, calling into XPath for data.
