inverted question or exclamation mark in pdf output after quote at the end of citation

[korakinos]

$ pandoc --version
pandoc 2.7
Compiled with pandoc-types 1.17.5.4, texmath 0.11.2.2, skylighting 0.7.7

This is likely a variation of #1164, triggered in the special case of a citation. I haven't been able to find a way to work around it.

Observed behaviour: If the letter combinations ?", ?“, !" and !“ appear at the end of a citation, the question/exclamation mark is inverted and the quotation mark changed to a single ‘ in pdf output.

Steps to reproduce:

1. Create the file temp.md, containing just the line:
   [@Lenz, "What is happening with the question mark end the closing quotation mark here?“]

2. Create any bogus empty bibliography file in the same directory:
   $ touch bogus.bib

3. Run pandoc:
   pandoc --output="temp.pdf" --bibliography="bogus.bib" "temp.md"

The resulting PDF shows the described behaviour:

temp.pdf

[jgm]

Let's see:

% pandoc  --biblio biblio.bib -t latex
[@item1, "What is happening with the question mark end the closing quotation mark here?“]
^D
(Doe 2005, "What is happening with the question mark end the closing
quotation mark here?``)

Here the `` is pandoc's way of rendering a left double quote, since this is the normal latex ligature for that symbol. But it turns out that the sequence ?` is a LaTeX ligature for the Spanish inverted question mark, so in this context you get bad results!

For your purposes, one way around this is to insert a zero-width unicode space between the question mark and the ``. Another way around it is to disable the use of tex ligatures by specifying -t latex-smart. In this case you'd also want to use a curly quote for the opening quote, presumably, and you'd want to specify --pdf-engine=xelatex so that the unicode quotes will be processed properly.

[jgm]

We can try to make pandoc's latex writer smarter, so that it doesn't inadvertently create the sequence ?`.

[korakinos]

Thanks a lot for both the zero-width whitespace quick fix and for the actual fix! This was fast!

(In case anybody else is in need of the quick fix before a new release of pandoc is made, the unicode code point for a zero-width space is U+200B. Edit: However, the zero-width space doesn't prevent line breaking. The "word joiner" character, code point U+2060, is thus more suitable.)