How to make a filter work post xml render?

[fredguth]

Description

I am trying to build a Scielo template to be used with Quarto for a while (#12251)[ˆ1].

The footnotes are being generated as:

<fn id="fn1">This is a footnote</fn>

instead of:

<fn fn-type="other' id="fn1">This is a footnote</fn>

For some reason, the scielo format checker complaints if there is no fn-type.

I tried to solve with the footnote-type.lua filter, but it is not working. It seems like the <fn> tag is being generated after the filter is applied.

Here is how I am calling the filter:

 ---
format:
  jats: 
    template: scielo.template
    output-ext: "scielo.xml"
    citeproc: true
    filters: 
      - at: post-render
        path: footnote-type.lua
  ---

Strangely, I also noticed that the el.format in the filter is html even though I haven't asked for a html format.
I guess I don't understand how the Pandoc processing happens inside Quarto. I imagined that pandoc was called and transformed qmd ---> AST --->xml.

---

[ˆ1]: I have built a mwe repo with the most minimal example I could think. It will also be used to another problem I found while trying to create this custom format.

[cderv]

This is your filter

function RawBlock(el)
    print(':::', el.format)
    if el.format == "html" then
        print(':::', el.text)
      el.text = el.text:gsub('<fn%s+id="([^"]+)">', '<fn fn-type="other" id="%1">')
      print(':::', el.text)
    end
    return el
  end

What you are doing it probably not what you think.
You are

• catching any RawBlock in the document => el.format is the raw format used on this rawblock.
• printing to console some content
• modifying the text of the Raw Block to be different with a gsub
• Returning the raw block

Your idea of qmd -> AST -> XML is a good one. This means your filter shouldn't consider the footnote has been transform to xml yet. at: post-render means it will come last in the Quarto rendering process, but still before Pandoc writer applies

For some reason, the scielo format checker complaints if there is no fn-type.

It seems Pandoc writers does not consider fn-type (https://github.com/jgm/pandoc/blob/3acd8859bd59b865631707f785207b43cda8314f/src/Text/Pandoc/Writers/JATS.hs#L512-L519) so that is why you don't see it

You could open an issue in Pandoc if this is supposed to be in the JATS spec I guess.

To catch the Note in Lua you would need to use the Note node in AST, but I don't think you can control how it written it seems as it does not have any attributes.

• https://pandoc.org/lua-filters.html#type-note
• https://pandoc.org/lua-filters.html#pandoc.Note

So I guess only post processing your XML is necessary 😞

[fredguth]

Thanks. So, post-render is only related to quarto render, right? And pandoc transforms happen after that. But, how do I change how Quarto calls pandoc? I notice that for JATS format, for example, it is using an extension that I may not want to use. Is this something we can configure in the quarto document yaml?

[cderv]

So, post-render is only related to quarto render, right? And pandoc transforms happen after that.

This how filters works at https://quarto.org/docs/extensions/filters.html#overview

They all apply to modify the AST. You can read about post-render at https://quarto.org/docs/prerelease/1.4/lua_changes.html#more-precise-targeting-of-ast-processing-phases

This is just about where you filter does get inserted in the Quarto Lua processing for the AST before Pandoc writers does apply

But, how do I change how Quarto calls pandoc? I notice that for JATS format, for example, it is using an extension that I may not want to use. Is this something we can configure in the quarto document yaml?

Not sure to see what you are referring too. If you want to write how Jats writer in Pandoc does the output, you need to either open an issue with them, or use Lua fitler to create the raw xml from the node (output a rawBlock("jats", ...) for example; But this won't work for all, it depends on the writers.

If the JATS writer from pandoc is not adapted to what you want to do, tweaking Quarto may not be enough, and pandoc itself needs to be made more generic.

And I don't see what the extensions refers to .