Force fenced-div syntax in markdown writer even when no attributes are set?

[cscheid]

I was slightly surprised by the fact that the Markdown writer can emit <div> when -t markdown+fenced_divs, due to a check for empty attributes in addition to the check for Ext_fenced_divs.

Would you consider accepting a PR from us that adds an additional Markdown writer extension Ext_force_fenced_divs, which would bypass that check? It would be useful for us over at the Quarto project to be able to specify a set of extensions that guarantees that no "naked" html elements are emitted by Pandoc's Markdown writer. I can come up with a workaround involving the injection of a UUID-named class and removal, but it would be a lot cleaner to just add +force_fenced_divs to our Markdown Writer calls.

[jgm]

Would consider it. I'm not quite sure what the purpose of that check was.

[cscheid]

Talking with @tarleb earlier today, he suggested I start with a PR that just removes that check.

[jgm]

OK, I merged the PR, but still thinking about this issue.

Currently we handle the cases of a div with empty attr and a span with empty attr differently.

A div with empty attr (after the PR) turns into

::: {}
contents
:::

While a span with empty attr turns into just its contents.

It seems odd to use different approaches in these two cases. Since the motivation for removing the empty attr check was to avoid raw HTML, what if we just do what we did with spans, and print the contents of the div without any marker?

The alternative would be to use [contents]{} for the span-with-empty-attr case, but I'm somewhat concerned that this will lead to a lot of spurious markup, esp. when converting from LaTeX.

[jgm]

Confirmed my fear about lots of spurious markup. Will try the other approach.

[tarleb]

I note that HTML writer preserves empty spans. So from a consistency perspective, it might be better to remove the empty spans in the LaTeX reader.

[jgm]

For now I'm going to stay with the mixed approach. I think that we really want the empty div structures, because (e.g.) someone may be using flexbox for layout and it matters how things are grouped. It's less crucial for spans. But, comments welcome.

[jgm]

Just saw your comment @tarleb. What you say is true. So, maybe we should make the change in the LaTeX reader. I'll make a note of this on the other issue.