Determine current file being processed in lua filter (when input_files > 1)

[dude-at-RA]

tl;dr, Is there a way to know which file in input_files a lua filter is currently operating on?

I am attempting to use pandoc to build multiple markdown files into a revealjs presentation. The presentation has a visible doc info footer with copyright info and a document ID. Each markdown file has a document ID in the YAML frontmatter, e.g:

---
doc:
  id: PRS123
  revDate: '2025-04-01'
title: Test Presentation
---

## Presentation content

in Lua, I'm adding the doc.id to each slide as it's built, essentially:

function Div(el)
  if (#findElementClassWithPrefix(el, 'slide') > 0) then
    table.insert(el.content, pandoc.Span(yamlVars['%doc.id%'], {class='doc_id'}))
    return el
  end
end

-- here's how yamlVars gets populated
function loadYamlToVars(meta)
    yamlVars = get_vars(meta, yamlVars, '')
end

function get_vars (meta, t, keylead)
    for k, v in pairs(meta) do
        vType = pandoc.utils.type(v)
        if vType == 'Inlines' then
            t["%" .. keylead .. k .. "%"] = {table.unpack(v)}
        elseif vType == 'table' then
            get_vars(v, t, keylead .. k..'.')
        end
    end
    return t
end

I've written the necessary javscript to update the footer div when a new slide is revealed that has doc ID, and it works for a single document (when doc id is added to the revealjs html template). As expected, when passing in multiple files, the last file wins and instead of getting changing doc ids as I navigate through the slidedeck, it's whatever the last markdown file's ID was.

I'm not sure if there's a better way to use the revealjs html template to accomplish this; it looks like all of the converted html content (minus the title content) gets put in to the html document in the $body$ placeholder - so I'm not sure how to pull each file's doc id there.

If I could see which file is being processed, then I could side load that markdown file temporarily in the lua filter and get the metadata out of that file.

[jgm]

If your document can be parsed as commonmark or a variant (gfm, commonmark_x) then you could enable the sourcepos extension, which would make source positions available on AST elements as an attribute, and they'd be accessible in Lua filters. Note that sometimes the positions are added to Span or Div wrappers, since not every element can take attributes.

[dude-at-RA]

Thanks for the quick response!

I'll have to look at the extensions available to commonmark_x / gfm and see if they have the extensions we're using so far with the standard pandoc flavor (I think this is doable with pandoc --list-extensions=markdown). Alternatively, we could change our build system to build input files individually and concatenate them together, but I'd like to avoid that if possible.

Is there a technical limitation that prevents sourcepos from being implemented in the standard pandoc markdown flavor?

Thanks again!

[jgm]

--list-extensions=commonmark

[jgm]

Is there a technical limitation that prevents sourcepos from being implemented in the standard pandoc markdown flavor?

It's not impossible, of course, but the parser wasn't designed with this in mind.