Filtering one file at a time

[rhn]

I'm converting a bunch of files to bind for printing, and I want to process each file separately.

In Pandoc 3.1.3 with --file-scope, each file was wrapped in a pandoc.Div, so I could extract titles:

function Div(elem)
  return pandoc.Div({elem.c[1]}, {class='titlepage'})
end

But in Pandoc 3.6.4, all elements are simply concatenated.

How do I wrap each file in a separate element?

[jgm]

This was changed in 3.2, #8741.

If each file begins with a level-1 heading, and you don't use level-1 headings elsewhere, then you can simply match Header elements with level == 1.

[jgm]

Of course, if you're just generating title pages (ignoring the rest of the content), then you could always just run pandoc separately on each file in a loop rather than using --file-scope.

[rhn]

I'm doing a couple more things than just title pages, but that was the minimal example. I need to insert tags denoting a new document in order to insert page breaks as well.

I could solve it part of the way by converting each document separately into HTML, and concatenating that. But it still leaves me with a problem: how do I process two elements where one is next to another? (I use Mediawiki markup so I can't just wrap the elements needing extra processing in a container.)

When I get pandoc.Div, I can find the first N elements and do special processing. The docs don't list a root Blocks element for the entire document, though. Is that possible without JSON?

[jgm]

You can use Pandoc as the root in a Lua filter, and get elem.blocks and elem.meta.

I need to insert tags denoting a new document in order to insert page breaks as well.

If the Header elements that start each chapter can be identified, then you can insert them with a filter.

[rhn]

Sadly, the Wiki is not as organized as I would like, so there may be multiple toplevel header elements anywhere.

Regarding wrapping documents in a Div, I would welcome it as an option. Doing a single pass in Lua is a lot easier compared to two passes plus shell scripting, and <article> or <section> is a natural way to deal with multiple documents in HTML.

Thanks for the help!