pairDelim in metadata seems to have no effect

[malcolmsailor]

I can't get pairDelim or lastDelim to have any effect when compiling pdfs via tex.

For example, I am copying the contents of https://github.com/lierdakil/pandoc-crossref/blob/7aa8e9eccaff0c457b28e4a30b290699fb922518/test/m2m/delim/input.md and compile it on my Mac with pbpaste | pandoc --filter pandoc-crossref with either --output ~/tmp/test.tex or --output ~/tmp/test.md.

When I compile to markdown, I see the pairDelim value on certain lines, e.g.,

References with `-` prepended won't have prefix at all: 1pair2,
eqns. 3pair4

But when I compile to tex, the pairDelim value seems to be absent:

References with \texttt{-} prepended won't have prefix at all:
\ref{eq:0}, \ref{eq:1}, eqns.~\ref{eq:2}, \ref{eq:3}

I can see from https://github.com/lierdakil/pandoc-crossref/blob/7aa8e9eccaff0c457b28e4a30b290699fb922518/test/m2m/delim/expect.tex that this seems to be the expected behavior, but it seems like a bug. I note that the documentation says "none of the *Delim options are honored with cleveref output. Use cleveref’s customization options instead", but I don't believe I am using cleveref here.

Is this actually the intended behavior? If so, do the docs need to be updated? Is there a way we can use pairDelim in tex output? Thanks!

[lierdakil]

I guess that's a shortcoming of the documentation... Since actual numbering is offloaded to LaTeX, we can't collapse ranges as we don't know a priori which numbers those will translate to (there are heuristics that could get us about 80% there, but since TeX is a Turing-complete language, the remaining 20% are basically insurmountable).

If you don't care about producing valid TeX references and only care about getting a reasonable-looking PDF, there's a hack you could do: instead of pandoc -F pandoc-crossref -t pdf invoke it as a pipe pandoc -t json | pandoc-crossref --pipe | pandoc -f json -t pdf -- this will sidestep all special handling wrt LaTeX. However, you'll have to also either modify the template or inject header-includes to suppress LaTeX titles on figures/tables/listings, otherwise you'll get Table N: Table M: <caption> type issues (where N and M might or might not coincide depending on LaTeX numbering config vs pandoc-crossref config). I may add an option to make such a hack more convenient.

Alternatively, if you don't particularly care about LaTeX itself, cleveref is a very viable option, it's basically the inspiration for how pandoc-crossref handles reference collapsing &c. You'll have to configure it via header-includes or a raw LaTeX block at the start of the document, though.

[malcolmsailor]

OK, thanks for the suggestions. I think for my current purposes, the simplest solution will be just to manually enter the delimiters myself as in "see figures [-@fig:1] and [-@fig:2]", but perhaps I will try one of the other solutions down the road.

If you're inclined to update the docs, it could be helpful to anyone who finds themselves in my position wondering why it doesn't seem to be working. I would submit an update myself but I think you'd be much better situated to explain succinctly and accurately.