Pandoc and foreign characters

Question

I've been trying to use Pandoc to convert some Markdown into a PDF file. This is a sample that Pandoc will not convert for me:

# Header!

## Sub Header

themselves derived respectively from the Greek ἀναρχία i.e. 'anarchy'

That's just something I grabbed from the top of the wikipedia database dump. Pandoc doesn't like that at all. This is the error message it gives me:

pandoc: Error producing PDF from TeX source.
! Package inputenc Error: Unicode char \u8:ἀ not set up for use with LaTeX.

See the inputenc package documentation for explanation.
Type  H <return>  for immediate help.
 ...                                              

l.53 ...es derived respectively from the Greek ἀ

Is there a command switch I can give it to get around this? I tried following the advice to do something like this, but it failed:

iconv -t utf-8 test.md | pandoc -o test.pdf

Update Before following John's advice below, see this.

Update 2 This is the command that ultimately got it working. Hopefully this will help someone:

pandoc test2.md -o test2.pdf --latex-engine=xelatex --template=my.latex --variable mainfont="DejaVu Serif" --variable sansfont=Arial

And this is the contents of my.latex:

\documentclass[$if(fontsize)$$fontsize$,$endif$$if(lang)$$lang$,$endif$$if(papersize)$$papersize$,$endif$]{$documentclass$}
\usepackage[T1]{fontenc}
\usepackage{lmodern}
\usepackage{amssymb,amsmath}
\usepackage{ifxetex,ifluatex}
\usepackage{fixltx2e} % provides \textsubscript
% use microtype if available
\IfFileExists{microtype.sty}{\usepackage{microtype}}{}
% use upquote if available, for straight quotes in verbatim environments
\IfFileExists{upquote.sty}{\usepackage{upquote}}{}
\ifnum 0\ifxetex 1\fi\ifluatex 1\fi=0 % if pdftex
  \usepackage[utf]{inputenc}
  \usepackage{ucs}
$if(euro)$
  \usepackage{eurosym}
$endif$
\else % if luatex or xelatex
  \usepackage{fontspec}
  \ifxetex
    \usepackage{xltxtra,xunicode}
  \fi
  \defaultfontfeatures{Mapping=tex-text,Scale=MatchLowercase}
  \setromanfont{TeX Gyre Pagella}
  \newcommand{\euro}{€}
$if(mainfont)$
    \setmainfont{$mainfont$}
$endif$
$if(sansfont)$
    \setsansfont{$sansfont$}
$endif$
$if(monofont)$
    \setmonofont{$monofont$}
$endif$
$if(mathfont)$
    \setmathfont{$mathfont$}
$endif$
\fi
$if(geometry)$
\usepackage[$for(geometry)$$geometry$$sep$,$endfor$]{geometry}
$endif$
$if(natbib)$
\usepackage{natbib}
\bibliographystyle{plainnat}
$endif$
$if(biblatex)$
\usepackage{biblatex}
$if(biblio-files)$
\bibliography{$biblio-files$}
$endif$
$endif$
$if(listings)$
\usepackage{listings}
$endif$
$if(lhs)$
\lstnewenvironment{code}{\lstset{language=Haskell,basicstyle=\small\ttfamily}}{}
$endif$
$if(highlighting-macros)$
$highlighting-macros$
$endif$
$if(verbatim-in-note)$
\usepackage{fancyvrb}
$endif$
$if(tables)$
\usepackage{longtable}
$endif$
$if(graphics)$
\usepackage{graphicx}
% We will generate all images so they have a width \maxwidth. This means
% that they will get their normal width if they fit onto the page, but
% are scaled down if they would overflow the margins.
\makeatletter
\def\maxwidth{\ifdim\Gin@nat@width>\linewidth\linewidth
\else\Gin@nat@width\fi}
\makeatother
\let\Oldincludegraphics\includegraphics
\renewcommand{\includegraphics}[1]{\Oldincludegraphics[width=\maxwidth]{#1}}
$endif$
\ifxetex
  \usepackage[setpagesize=false, % page size defined by xetex
              unicode=false, % unicode breaks when used with xetex
              xetex]{hyperref}
\else
  \usepackage[unicode=true]{hyperref}
\fi
\hypersetup{breaklinks=true,
            bookmarks=true,
            pdfauthor={$author-meta$},
            pdftitle={$title-meta$},
            colorlinks=true,
            urlcolor=$if(urlcolor)$$urlcolor$$else$blue$endif$,
            linkcolor=$if(linkcolor)$$linkcolor$$else$magenta$endif$,
            pdfborder={0 0 0}}
\urlstyle{same}  % don't use monospace font for urls
$if(links-as-notes)$
% Make links footnotes instead of hotlinks:
\renewcommand{\href}[2]{#2\footnote{\url{#1}}}
$endif$
$if(strikeout)$
\usepackage[normalem]{ulem}
% avoid problems with \sout in headers with hyperref:
\pdfstringdefDisableCommands{\renewcommand{\sout}{}}
$endif$
\setlength{\parindent}{0pt}
\setlength{\parskip}{6pt plus 2pt minus 1pt}
\setlength{\emergencystretch}{3em}  % prevent overfull lines
$if(numbersections)$
$else$
\setcounter{secnumdepth}{0}
$endif$
$if(verbatim-in-note)$
\VerbatimFootnotes % allows verbatim text in footnotes
$endif$
$if(lang)$
\ifxetex
  \usepackage{polyglossia}
  \setmainlanguage{$mainlang$}
\else
  \usepackage[$lang$]{babel}
\fi
$endif$
$for(header-includes)$
$header-includes$
$endfor$

$if(title)$
\title{$title$}
$endif$
\author{$for(author)$$author$$sep$ \and $endfor$}
\date{$date$}

\begin{document}
$if(title)$
\maketitle
$endif$

$for(include-before)$
$include-before$

$endfor$
$if(toc)$
{
\hypersetup{linkcolor=black}
\setcounter{tocdepth}{$toc-depth$}
\tableofcontents
}
$endif$
$body$

$if(natbib)$
$if(biblio-files)$
$if(biblio-title)$
$if(book-class)$
\renewcommand\bibname{$biblio-title$}
$else$
\renewcommand\refname{$biblio-title$}
$endif$
$endif$
\bibliography{$biblio-files$}

$endif$
$endif$
$if(biblatex)$
\printbibliography$if(biblio-title)$[title=$biblio-title$]$endif$

$endif$
$for(include-after)$
$include-after$

$endfor$
\end{document}

Solution 1

Use the --pdf-engine=xelatex option.

Solution 2

By default, Pandoc use the pdflatex engine when converting markdown file to pdf files. pdflatex can not handle Unicode characters very smoothly as xelatex. You should try xelatex instead. But, merely using xelatex command is not enough. As is often the case, you need to choose a proper font which contains glyphs for the Unicode characters your want to typeset.

I am a Chinese user, so take Chinese for example. If you have a test.md which contains the following content:

你好汉字

you can use the following command to compile this markdown file:

pandoc --pdf-engine=xelatex -V CJKmainfont="KaiTi" test.md -o test.pdf

In the above command, --pdf-engine=xelatex is used to select the LaTeX engine (for the new version of Pandoc, --latex-engine option is deprecated). -V CJKmainfont="KaiTi" is used to select the proper font which support Chinese. For other languages, you may use the flag -C mainfont="<FONT_NAME>".

How to find a font which support your language

In order to find a font which supports your language, you need to know your language code. Then, if you are on Linux system or on Windows systems with TeX Live installed. You can use the following command to find a valid font for you language:

fc-list :lang=zh #find the font which support Chinese (language code is `zh`)

The output on my Linux system is shown below

If you choose to use, e.g. the font Source Han Serif CN, then use the following command to compile your markdown file:

 pandoc --pdf-engine=xelatex -V CJKmainfont="Source Han Serif CN" test.md -o test.pdf

Solution 3

UPDATE: the answer below seems to be valid for pandoc 1.x but with later versions the syntax has changed

---

Coming back to this post in five years time and the issue is still there. The command

pandoc -s test.md -t latex -o test.pdf

fails when test.md contains text with non-latin characters, Greek, Cyrillic, CJK, Hebrew and Arabic included.

LaTeX was designed before Unicode and its support for different character sets is robust in some areas but far from comprehensive, so the advice to use XeLaTeX is valid yet requires one to choose the main font carefully, since there is no automatic choice.

Below is a small taxonomy of possible issues and some solutions. All tested with Pandoc 1.19.

Cyrillic

Support for Cyrillic alphabet in LaTeX is provided via T2A font encoding.

Consider a small sample:

# Header

## Subheader

Tetris (Russian: Тетрис) quoting Wikipedia is a tile-matching puzzle 
video game

Running this example with pandoc would fail with:

! Package inputenc Error: Unicode char Т (U+422)
(inputenc)                not set up for use with LaTeX.

See the inputenc package documentation for explanation.

A fix is available as fontenc option is a predefined variable in default.latex template.

Running this example with

pandoc -t latex -o tetris.pdf -V fontenc=T2A cyrillic.md

would produce correct rendering

This however would not handle other language features correctly such as hyphenation. A better way would be to use Babel and have it select the correct font encoding.

pandoc -t latex -o tetris.pdf -V lang -V babel-lang=russian cyrillic.md

Or to switch languages with Babel commands inside Markdown

# Header

## Subheader

Tetris (Russian: \foreignlanguage{russian}{Тетрис}) quoting Wikipedia 
is a tile-matching puzzle video game

And run with

pandoc -t latex -o tetris.pdf -V lang -V babel-lang=english \
       -V babel-otherlangs=russian cyrillic2.md

Greek

The example in the original post contains characters both from the main and extended Greek Unicode codepages.

Anyway, the widely used LGR greek font encoding is not covered by LaTeX 3 project and is classified as a local encoding, i.e. it may vary from site to site and from system to system according to the LaTeX Encoding Guide.

On TeX Live the following packages need to be installed: texlive-greek-inputenc, texlive-greek-fontenc and texlive-cbfonts. Note that you need Babel 3.9 or later. However the result of

pandoc -t latex -o anarchy.pdf -V fontenc=LGR greek.md

may appear unexpected.

In order to correct this issue one has to setup LaTeX Babel package correctly. And insert commands to switch between the languages in the original text:

# Header!

## Sub Header

themselves derived respectively from the Greek \textgreek{ἀναρχία} 
i.e. 'anarchy'

Compiling this with the following command

pandoc -s greek2.md -t latex -V fontenc=T2A -V lang -V babel-lang=english \
    -V babel-otherlangs=greek -o greek.pdf

would produce the output exactly as you would expect it to be:

XeLaTeX

All of this would not be needed if we were using XeLaTeX.

Just running the original example with

pandoc -s greek.md --latex-engine=xelatex -t latex -o greek.pdf

would produce

Because the font does not contain anything in the greek character positions the output contains some white space instead.

Selecting one of the popular fonts as the new mainfont would help a bit

pandoc -s greek.md --latex-engine=xelatex \
    -V mainfont="Liberation Serif" -t latex -o greek.pdf

However characters from the extended Greek codepage such as the small letter alpha with psili accent are not rendered.

The Font Setup for Greek with XeTeX/LuaTeX Guide suggests to use DejaVu, Libertine or Free font families.

Indeed with DejaVu Serif, Linux Libertine O as well as Tempora and perhaps some other fonts, the result would be as expected. See below the rendering with XeLaTeX and Linux Libertine fonts.

pandoc -s greek.md --latex-engine=xelatex -V mainfont="Linux Libertine O" \
      -t latex -o greek.pdf

Solution 4

Works for Cyrillic characters

pandoc myfile.md --pdf-engine=xelatex -V mainfont=Arial

Solution 5

You can use --latex-engine=xelatex, as said before, but the best I have found is to use the lang variable to specify the document language in the header, like this: lang: ru-RU. A working example on my debian workstation:

---
title: Lady Macbeth de Mzensk (Chostakovitch, livret d'Alexandre Preis, 1934)
lang: ru-RU
---

# Acte I / Tableau 1

*[Народ ненадежный]*  
Ха, ха, ха, ха, ха, ха, ха. *[...]* Чуыствуем  
На кого ты нас покидаешь?  
Без хозяина будет скучно,  
скучно, тоскливо, безрадостно.

Не работа. Без тебя невеселье. Воз вращайся  
Как можно скорей, скорей !

Then you can launch:

$ pandoc -o your-file-output.pdf your-source-file.md

View more solutions

Comments

[Mike Thomsen]

That created the document, but now I have a bunch of blank characters where the greek word is supposed to be. I think it is not recognizing the characters in question.

[z--]

The blank characters show up when you have a font selected which does not contain greek glyphs. Use the --variable mainfont="..." option on the command line. See johnmacfarlane.net/pandoc/demos.html Example 14 (Xe)Latex

[J3soon]

See Pandoc with Chinese for more info.

[J3soon]

XD. However, for me I need to use mainfont instead of CJKmainfont.

[Dmitri Chubarov]

It is not true that pdflatex can not handle Unicode characters.

[jdhao]

@DmitriChubarov, is that true? I am not an expert at LaTeX. But xelatex seems the option when dealing with Unicode characters.

[Dmitri Chubarov]

LaTeX was designed before Unicode, so it had only support for 256-character coding tables, however as it was possible to switch coding tables and fonts dynamically under the hood, it was completely normal to supply UTF-8 encoded input to LaTeX compiler, translation was performed by packages such as inputenc and fontenc.

[jdhao]

@DmitriChubarov, thanks for the information. I will revise my answer.

[jdhao]

For CJK language, you might want to use -V CJKmainfont="..." instead. See the answer here.

[Sardathrion - against SE abuse]

--pdf-engine is the new option …

[Creasixtine]

You can also use the lang variable as I have described in my answer.

[Roman Golyshev]

That 'babel-lang' option it the only thing that worked for me, thanks! It is not directly mentioned in the official pandoc documentation

[Merchako]

This answer is incomplete (doesn't mention fonts) and no longer correct (should say --pdf-engine).

[Nazar Paruna]

I don't know why but for some reason after checking updates and installing packages this command stopped work. I have removed Pandoc (v2.16.2) and MikTeX (v21.8) and re-installed again. After that I check for updates, but stop to install new packages. And it works again. Maybe my comment will be useful for someone.