Preserving <i>, <b> and <cite> in non-HTML formats

[snan]

So the position seems to be that em and strong are OK for * and ** but then how about this idea, as per discussed on commonmark's issue tracker: that we can write <i>, <cite> and <b> and have them be understood by pandoc (as an exception to most other HTML tags just being passed through or dropped) and turned to italics and bold in other formats, like ContTeXt and LaTeX?

[mb21]

For HTML input this already works...

echo '<i>foo</i>' | pandoc -f html -t native
[Plain [Emph [Str "foo"]]]

But you're talking about markdown input, or..?

echo '<i>foo</i>' | pandoc -f markdown -t native
[Para [RawInline (Format "html") "<i>",Str "foo",RawInline (Format "html") "</i>"]]

true, there it's converted to raw HTML, which will only work for HTML output and be dropped otherwise... but why would you want to use HTML <i> inside markdown?

[snan]

Yes, using a markdown source document to generate html and other output formats, including some custom lua ones.

[snan]

I'm way OK with another solution to create b and i and cite too. We've previously argued for having * and ** always be i and b. I'm not invested in one particular solution, I just want to find some way to not generate In <em>Hamlet</em>, Shakespeare says…

[mb21]

I just want to find some way to not generate In <em>Hamlet</em>

ah, you can write a lua filter to change the native Emph [Str "foo"] into RawInline (Format "html") "<i>",Str "foo",RawInline (Format "html") "</i>"]... or post-process the HTML...

[snan]

What I want is the opposite; to write markdown that fits on all kinds of markdown sites (that don't all use pandoc)—and I can do that with the `<i>` and `<b>` most places—and then still have pandoc turn `<i>` into italic or `<b>` into bold when I want to generate ConTeXt. So I guess one workaround there is to preprocess it before I pass it to pandoc. Also I love markdown but this is such a weird part of the language.

[mb21]

haha, I don't get it. If you want to use portable markdown, just use * and **...? some markdown processors will convert to <em>, and a few weird ones might convert to <i>... so yes, you can post-process the HTML then I guess..?

[snan]

If I write `_Taraxacum officinale_ grows from unbranched taproots` and that gets turned into `<em>Taraxacum officinale</em> grows from unbranched taproots`, that's wrong. It's not emphasis.

[mb21]

Ah, so this is a duplicate of #4297 ?

[snan]

It's another solution to the same problem!

[tarleb]

If I understand correctly, then you want to write <i>Saccharomyces cerevisiae</i> and have it appear verbatim in HTML, but as "normal" emphasized text in other formats? This would be doable with a Lua filter. Another filter-based solution would be to write [Escherichia coli]{.species}. The corresponding filter would be

function Span (span)
  if not span.classes:includes 'species' then
    return nil
  elseif FORMAT:match 'html' then
    return {pandoc.RawInline('html', '<i>')} .. span.content .. {pandoc.RawInline('html', '</i>')}
  else
    return pandoc.Emph(span.content)
  end
end

(Untested)

[tarleb]

For completeness, here's a filter which allows using <i> as an alternative to * or _.

if FORMAT:match 'html' then
  return {}
end

local List = require 'pandoc.List'

function is_italics_start(inln)
  return inln.t == 'RawInline'
    and inln.format:match 'html'
    and inln.text:match '<i>'
end

function is_italics_end(inln)
  return inln.t == 'RawInline'
    and inln.format:match 'html'
    and inln.text:match '</i>'
end

function Inlines (inlns)
  local result = List()
  local italics = nil
  for _, inln in ipairs(inlns) do
    if is_italics_start(inln) then
      italics = List()
    elseif is_italics_end(inln) then
      result:insert(pandoc.Emph(italics))
      italics = nil
    elseif italics then
      italics:insert(inln)
    else
      result:insert(inln)
    end
  end
  return result
end

Any additional discussions should happen on the pandoc-discuss mailing list.