Web
Article

Creating PDFs from Markdown with Pandoc and LaTeX

By Chris Ward

creating PDFs

If you've read some of my previous posts on SitePoint or elsewhere, you may know that I'm working on a board game. In the game, called Chip Shop, you get to run a computer company in 1980s America.

As part of the project, I'm attempting to open source the entire game as much as possible. After several false starts, I've decided on a basic framework of Markdown for most of the game components—especially cards and the manual.

As the game's website uses Jekyll, the website for the game is generated from the Markdown files. I intend to have premium pre-boxed and print-yourself versions of the game, and to achieve this I need to generate PDFs from the Markdown files.

What I'm Trying to Accomplish

My ideal workflow is to generate the PDF files at the same time as generating the website, rather than generate the files as visitors request them. This rules out my usual option for PDF generation, wkhtmltopdf, as it generates PDFs from already generated HTML. Another reason it's not an option is that I want the PDF card versions to look different from the HTML pages, and Jekyll lacks any kind of view mode feature to accomplish this without resorting to complex CSS rules.

The Markdown template file for cards in the Chip Shop game contains a lot of Markdown front matter fields for game mechanics. Not all are used on every card. For convenience during printing, I need to fit as many cards on an A4 page as possible—in this case, a 3×3 grid. Eventually the pages will need to be double-sided, but I haven't implemented that yet.

Enter Pandoc and LaTeX

Any internet search looking for solutions to generating PDFs from Markdown will lead you down the Pandoc path. Pandoc is an open-source, Swiss Army knife markup conversion tool that supports a wide and growing variety of input and output markup formats.

To generate PDFs with Pandoc, LaTeX is needed. LaTeX has its roots in the scientific research community, and is a document declaration and layout system. Combining Pandoc and LaTeX allows us to use variables, and thus to generate PDFs from a series of Markdown files and support Markdown front matter.

Despite the power of Pandoc and LaTeX, I couldn't find any way of combining multiple PDFs (cards) onto one page, especially when using variables from Markdown files. After much research, I settled on PDFJam, a simple command line tool for this requirement.

Installing Dependencies

Markdown

You need no extra software for Markdown, except maybe an editor and there are so many of those, I suggest you read a few SitePoint posts to make your choice.

Jekyll

I'll continue to use Jekyll in my examples taken from my game to illustrate the build process, but it isn't an essential part of PDF generation if you don't need a website.

Pandoc

On my Mac, I installed Pandoc with Homebrew, but there are options for all operating systems.

LaTeX

There are lots of opinions on the best way to install LaTeX, depending on what you need or intend to do with it. A full installation of its common tools and libraries can near 2GB, but for most purposes a minimal installation will be enough. Read the project's download page to find the best option for you.

For this tutorial, we'll be using the xelatex engine, as I use custom fonts. But you can select any engine that supplies specific features you require.

PDFJam

Depending on how you installed LaTeX, you may have PDFJam installed already. (Check by typing which pdfjam in the terminal.) If you haven't, then find details on installation here.

The Build Process

After some consideration, a bash script that I run locally seemed the best option for now. There are better methods, but it works, and I can improve the process later, transferring it to a Continuous Integration system or Git Hooks.

View the bash script on GitHub.

Let's now step through this script.

Setup

bundle install
bundle update

rm -dfr _site
rm -dfr pod

These commands ensure that the Ruby dependencies Jekyll needs to build the website are up-to-date, and that we remove any existing website and print folders.

Build the Website

jekyll build
mkdir -p pod/pdf/cards

Next we build the website and create a folder for the print versions of the cards.

Generating PDFs from Markdown

Let's create a folder containing a PDF version of each Markdown file:

for filename in _cards/*.md; do
  echo $filename
  pandoc --from=markdown+yaml_metadata_block --template _layouts/cards.latex -o pod/pdf/cards/"$(basename "$filename" .md)".pdf --latex-engine=xelatex $filename
done

The script processes every Markdown file in the _cards directory, ensuring that the Markdown front matter fields are observed. Using the cards.latex template (we'll look at that next), the correct LaTeX engine outputs a PDF with an appropriate name.

The LaTeX File

A lot of the magic for generating the card files from Pandoc takes place in a LaTeX template.

View the LaTeX template on GitHub.

LaTeX is new to me, but it isn't too complex. I'll explain what I changed from the default LaTeX file (found in Pandoc_install_dir/data/templates/default.latex) to make the cards work. I recommend sharelatex.com for previewing LaTeX files as you edit them.

\usepackage[paperheight=9.0cm,paperwidth=5.7cm,margin=0.5cm]{geometry}
% Set page size
\usepackage{multicol}
% We need column layouts
\usepackage{amsthm,amsmath,amssymb}
\usepackage{graphicx}
% We want images in our layout
\graphicspath{{/Users/chrisward/Workspace/cs_jk/_site/assets/images/cards/}}
% Where are images located
\usepackage{float}
\usepackage[utf8]{inputenc}
\usepackage{fontspec}
\setmainfont{VT323}
% We want to use a custom font installed on our local system, so add that package and select the font

We need a specific page size, and we'll use columns later for the costs and scores of the cards. We're using graphics and custom fonts, so we need those packages.

We're attempting to create a simple layout that's clear and uncluttered. Here's how we accomplished it:

\begin{document}
  \begin{flushright}
    {$title$}
  \end{flushright}
  $if(image)$
  \begin{figure}[H]
    \centering
    \includegraphics[height=2cm]{$image$}
  \end{figure}
  $endif$
  \begin{flushleft}
    \scriptsize{$body$}
  \end{flushleft}
  \scriptsize
  \begin{tabular}{ l l }
    Costs & Scores \\
    $if(staffcost)$ Staff: {$staffcost$} $endif$ & $if(loyaltyscore)$ Loyalty: {$loyaltyscore$} $endif$ \\
    $if(rdcost)$ RandD: {$rdcost$} $endif$ & $if(profitscore)$ Profit: {$profitscore$} for {$profitlength$} turns $endif$ \\
    $if(marketingcost)$ Marketing: {$marketingcost$} $endif$ & $if(longevityscore)$ Longevity: {$longevityscore$} $endif$ \\
    $if(longevitycost)$ Longevity: {$longevitycost$} $endif$ &  \\
    $if(moneycost)$ Money: {$moneycost$} $endif$ &
  \end{tabular}
  \begin{flushleft}
    $if(specialscore)$
    Special: {$specialscore$}
    $endif$
  \end{flushleft}
  \begin{center}
    $if(legal)$
    \tiny{$legal$}
    $endif$
  \end{center}
\end{document}

I feel a lot of the above is reasonably self explanatory for anyone used to code or markup. We're creating the elements of the card, aligning them, setting font sizes and checking if there are values before outputting them, so that the card doesn't end up with empty fields.

We resize the image to a particular size and centre it. The costs and score values are in a two-column layout, set with the begin{tabular} command and the column quantity with the number of ls.

Single Card

Combining Cards Onto One Page

We use PDFJam to create a large PDF file combining each of the individual PDF cards:

pdfjam pod/pdf/cards/*.pdf --no-landscape --frame true --nup 3x3 --suffix complete --outfile ./cards.pdf
mv cards pod/cards_complete.pdf

With this command, we specificy the following:

  • that the page orientation should always be portrait
  • that each individual PDF should be framed
  • the grid size
  • a file name suffix
  • a file name.

PDFJam can give an error if you don't output into its working directory, so I move the file to where I actually want it (hopefully that's solvable in the future). Here we could also delete the individual PDF files if we don't want them.

And that's it—we have a website and printable PDF of the game cards.

Cards 9up

Running the Script

I run the build script with ./build.sh. As there's a lot of image and PDF processing, it takes about five to ten minutes. I then have a separate script that deploys these folders to a web server.

What's Next

This process has taken me a while to get right, but it's now good enough to move forwards and finesse the process and layouts after play testing.

I hope you find my research and experiments useful for your projects. Please let me know if you have any comments or suggestions.

  • http://atravellingcook.com Cate

    nice article, very easy to follow

Recommended

Learn Coding Online
Learn Web Development

Start learning web development and design for free with SitePoint Premium!

Instant Website Review

Use Woorank to analyze and optimize your website to improve your website to improve your ranking!

Run a review to see how your site can improve across 70+ metrics!

Get the latest in Front-end, once a week, for free.