Under pandoc 1.15.0.6, some unusual but valid characters in URLs are not escaped by the LaTeX writer. For instance, this produces a broken link:

pandoc -o test.pdf -sS << EOT
> <http://www.digizeitschriften.de/dms/resolveppn/?PID=PPN345858530_0013|log48>
> EOT

It appears in the PDF as http://www.digizeitschriften.de/dms/resolveppn/?PID=PPN345858530_0013\T1\textbar{}log48.

Thread: https://groups.google.com/forum/#!msg/pandoc-discuss/E489etyaKgI/cUFuyicXCAAJ (as far as I can tell, no issue was generated from this).

If I'm understanding the code correctly, am I correct in thinking that the fix for this is as simple as adding not isUrl to https://github.com/jgm/pandoc/blob/master/src/Text/Pandoc/Writers/LaTeX.hs#L239?

It looks as if it's the list of 'unwise' characters that are not handled within a URL. The question, then, might be whether it would be best to escape these characters in all outputs (which has come up before in #1640).

Taking the example from #1640, this also results in an incorrect PDF:

pandoc -F pandoc-citeproc -o test.pdf << EOT
---
references:
- id: test
  DOI: 10.1002/(sici)1097-0185(19990415)257:2<50::aid-ar4>3.3.co;2-n
...

@test

EOT

Thanks for taking care of this!