![](https://secure.gravatar.com/avatar/d4db251f6da966f841fc5b1a9822eb56.jpg?s=120&d=mm&r=g)
On 4 Aug 2022, at 16:07, Hans Hagen
wrote: On 8/4/2022 12:22 PM, Taco Hoekwater via ntg-context wrote:
On 3 Aug 2022, at 15:54, Pablo Rodriguez via ntg-context
wrote: BTW, is there any way to rephrase "{h3 + blockquote}" in proper Lua? I was wondering about that as well, and I really had no clue how to do that. Some reading and studying later, I realised that there is a preceding-sibling:: axis. That is not documented in xml-mkiv.pdf I think, but it inherited from xpath, and that means this works: \xmlsetsetup{#1}{blockquote/preceding-sibling::h3[-1]/} {xml:section} “Take all blockquotes, then tests their immediate preceding siblings (index [-1]) that are h3." But the CSS version is nicer. Still, both solutions fail on generic input. Sorry, out of clues \startbuffer[demo] ...
That puts the whole <div> content inside the \section, and I don’t think that was the intent. I understood that Pablo only wants the “h3” node. In the original request, the CSS “{h3 + blockquote}” matches a “blockquote” anyplace where it is immediately preceded by a “h3” (and nowhere else). After that, some trickery was needed to try and get back to the actual “h3”. My trick above with preceding-sibling above also gets the “h3” that is immediately followed by a “blockquote”, but it got easily confused if there were multiple “h3” or “blockquote” objects in the input. Best wishes, Taco PS This sort of confusing filtering is why I do most of my xml processing from the lua end, where it is much clearer to me what is what. — Taco Hoekwater E: taco@bittext.nl genderfluid (all pronouns)