Hi Nicolas, Thank you very much for your excelent explanation, which has helped me to learn new and valuable things.
I chose `headline' instead of `section' because in the scenario where I want to apply those LaTeX groups I was looking for something like this: {level 1 - group 1 {level 2 group 1 + group 2 {level 3 group 1 + group 2 + group 3 {etc. }}}} I have tested your function and it works fine when all headers are at the same level, but with different levels the filter understands both properties as `font' (?). Anyway, and only for practical purposes, and because I think that my initial idea of 2 properties was uneconomical, I have slightly modified your function with a single property `:fontspec:'. Now, with a couple of marks and some regexp both the \fontspec group and the one with \addfontfeature are extracted: | :fontspec: font ! optional: features | \fontspec{font}[features] | |----------------------------------------+---------------------------| | :fontspec: > features | \addfontfeature{features} | then the variable `fontspec': (let* ((font <font-string-obtained-via-regexp>) (fontfeature <fontfeature-string-obtained-via-regexp>) (fontspec (cond (font font) (fontfeature fontfeature)))) is passed as an argument of `(funcall create-export-snippet ...)' With a single property it works fine, although I have to test more... Thak you very much! Best regards, Juan Manuel Nicolas Goaziou writes: > Hello, > > Juan Manuel Macías <maciasch...@posteo.net> writes: > >> I am writing a custom parse tree filter that does the following (LaTeX >> backend): if a heading has the ':font:' property, the content of that >> heading is enclosed in a LaTeX group. If the property is ':fontfeature:', >> then the content is enclosed in a different group. The filter works fine >> when all the headings are at the same level. But with different levels, >> it does not returns the expected result. It's evident that I'm doing >> something catastrophically wrong :-). I wonder if anyone could put me on >> the track of the origin of my error... >> >> Below, the offender function and a sample. Thanks in advance! > > I think you are operating at the wrong level. Higher level headlines > contain lower level ones. I suggest to operate on sections instead. > > Also, using `org-element-interpret-data' is meh because you're operating > at the parse tree level. You can insert export-snippet objects directly. > > Here's a proposal. This could be refactored, but you get the idea. > > (defun my-custom-filters/fontspec-headline (tree backend info) > (when (org-export-derived-backend-p backend 'latex) > (org-element-map tree 'section > (lambda (section) > (let ((font (org-export-get-node-property :FONT section t)) > (fontfeature (org-export-get-node-property :FONTFEATURE section > t)) > (create-export-snippet > ;; Create "latex" export-snippet with value V. > (lambda (v) > (org-element-create 'export-snippet (list :back-end "latex" > :value v))))) > (cond > (font > (apply #'org-element-set-contents > section > (append (list (funcall create-export-snippet "%font > start\n")) > (org-element-contents section) > (list (funcall create-export-snippet "%font > end\n"))))) > (fontfeature > (apply #'org-element-set-contents > section > (append (list (funcall create-export-snippet "%fontfeature > start\n")) > (org-element-contents section) > (list (funcall create-export-snippet "%fontfeature > end\n")))))))) > info) > tree)) > > Also, when "org-cite-wip" is merged, you will be able to replace, e.g., > > (funcall create-export-snippet "%fontfeature start\n") > > with > > (org-export-raw-string "%fontfeature start\n") > > Regards,