Dear John,

Le lundi 29 avril 2019 à 16:57 -0400, John Kitchin a écrit :
For org-ref, there isn't much magic on what happens on export. LaTeX is certainly the most well supported, and it seems like org -> latex -> pandoc is the only way that makes sense to get to docx to me.  Using pandoc on org files directly is probably hopeless unless you can get pandoc to include some definitions for the org-ref links.

This might be difficult : the development of ox-pandoc seems to not be very active at the moment...

Some of the link types in org-ref have some exports defined for org, html, latex, sometimes ascii. If one of these works well with pandoc we could try to make them output something useful for them, or at least make sure that org->org export turns them into something useful.

I'm currently looking at the ox-latex exporter in order to understand what it does for source blocks with org's names and captions (and try to fix the fact that they are labeled and nubered as figures...). Theis understanding might help me to go in the direction you suggest.

Getting figure/table numbers has always been tricky; I don't think this worked well with pandoc, and handling it on the org side requires some preprocessing to add numbers. For now, the ox-word exporter in scimax comes closest, but it isn't a feature I use a lot, so it hasn't been improved in a while.

Again, looking at what ox-latex does for org's names and captions might be helpful. Ox-pandoc seems to do a decent job on docx output.


John

-----------------------------------
Professor John Kitchin 
Doherty Hall A207F
Department of Chemical Engineering
Carnegie Mellon University
Pittsburgh, PA 15213
412-268-7803


On Mon, Apr 29, 2019 at 1:06 PM Emmanuel Charpentier <emm.charpentier@free.fr> wrote:
Dear list,

one of my uses od org-mode is to prepare documents wrapping R (and
sometimes Sagemath) call results in interpretation text. My reference
output is .pdf documents, but I *have* to prepare a .docx version (for
use in managerial spheres, where computer literacy is *very* low.
Cross-references and citations are a sine qua non, maths are useful.

I have been annoyed by a couple of deficiencies and inconsistencies
between exporters, so I prepared a test document testing various cases.
This documents and some exports are attached (NE = Native exporter, PE
= ox-pandoc exporter).

TL;DR :

  * I tested the built-in latex/pdf exporter as well as ox-pandoc, the
latter both for .pdf and .docx export. The built-in ODT exporter
doesn't export citations ; therefore, I didn't test it further.


  * org-ref's :labels and :refs do not export to anything but the
built-in latex exporter. The native system of #+NAME:s and #+CAPTION:s,
a bit on the heavy side, seems not to fail (except that they do not
expand in a caption...).

  * Maths, tables, figures are unproblematic.

  * The requirements of org-reftex, the built-in latex exporter and ox-
pandoc being mutually incompatible, and some ingenuity is required. see
the attached org source. Org-ref's requirements do not simplify the
situation...

  * Code snippets (i. e. source blocks exporting code) have a
captioning/numbering problem :

    - With the built-in latex exporter, they are numbered and labeled
as figures.

    - The pandoc latex exporter numbers them separately (as seen by
referencing them), but do not output this number (nor the category)
before the caption.

    - The pandoc .docx exporter works as advertised.

So I have a couple of questions:

  * What can be done to reconcile org-ref's, latex-exporter's and ox-
pandoc's requirements for bibliographies ?

  * How to fix the pdf exporters' quirks with code snippets ?

HTH,

--
Emmanuel Charpentier