andykuszyk.github.io

My personal tech blog

View on GitHub

A LaTeX Deployment Pipeline

Outside of my day job, I dabble in writing short science-fiction stories. I recently decided to start publishing these stories on a small site in order to share them with my friends and family and solicit feedback.

During this process, I needed to come up with a good way of transforming my raw LaTeX files into a working site that could be updated automatically. This article takes you through the process of:

The structure of my LaTeX projects

After a bit of experimentation, I’ve settled on a standar project layout which is available at https://gitlab.com/akuszyk/template.

This repo consists of a simple LaTeX book layout, with a main file (template.tex) which is used to configure the layout of the document and include any other files (e.g. chapters). For my shorter work, I simply write directly in this file.

For example, my template.tex file looks like this:

\documentclass{book}
\usepackage[paperwidth=130mm,paperheight=198mm]{geometry}
\title{Title}
\date{}
\author{A. Kuszyk}

\begin{document}
    \maketitle
    \tableofcontents
    \include{chapters/chapter-one}
\end{document}

This begins a book, in A5 format, with a title, author, table of contents and various included files. I find this is useful for breaking down larger documents into smaller, more managable files. A typical chapter would begin with:

\chapter{Chapter One}
This is the first chapter.

Included in the repo is a Makefile, with plenty of handy targets for building PDF, HTML and EPUB files. Most notable, however, is the default target. Simply running make will build a PDF of the document. Here’s the Makefile:

.DEFAULT_GOAL := default

build:
    docker build -t build .

default: build
    docker run --mount type=bind,source=$(PWD),destination=/build build

clean:
    rm *.pdf

pdf:
    pdflatex *.tex && pdflatex *.tex

stamp:
    ./scripts/stamp.sh

watch:
    find . | grep -v git | grep -e 'tex$$' | entr -c make

count:
    wc -w chapters/*.tex

update:
    ./scripts/update.sh

html:
    ./scripts/html.sh

epub:
    ./scripts/epub.sh

reading-time-mins:
    echo $$(($$(cat $$(find . | grep -v .git | grep -e 'tex$$') | wc | awk '{print $$2}') / 200))

As you can see, the default target builds the PDF in Docker container which means that I don’t need a local installation of LaTeX on the machine where I’m working, just Docker, and it also means I get reproducible output files whether I’m building the PDF locally or on a CI platform. Here’s the Dockerfile that’s used:

FROM andykuszyk/latex
WORKDIR build
CMD pdflatex *.tex && pdflatex *.tex

The andykuszyk/latex base image is pretty straightforward:

FROM alpine 
RUN apk update && \
    apk add texlive git python3 curl make bash && \
    wget https://github.com/jgm/pandoc/releases/download/2.9.2.1/pandoc-2.9.2.1-linux-amd64.tar.gz && \
    tar xvf pandoc-2.9.2.1-linux-amd64.tar.gz && \
    cp pandoc-2.9.2.1/bin/* bin && \
    rm -rf pandoc-2.9.2.1 && \
    rm pandoc-2.9.2.1-linux-amd64.tar.gz

As for the rest of the targets, check out the repo for more details on their implementation.

The CI pipeline for my LaTeX projects

Whenever I push to any of my writing projects, the CI will build a PDF, stamp it with part of the commit hash and artifact the PDF. This makes it easy to quickly take a look at previous builds in PDF format and access PDFs on the go.

The GitLab CI configuration for the stage that actually builds the PDF is dead simple:

image: andykuszyk/latex
pdf:
    script:
        - make pdf
        - make stamp
    artifacts:
        paths:
            - '*.pdf'

Some of my projects also publish directly to my static fiction site, and this involves a couple of additional steps. Once the PDF is built, I use the targets in the Makefile to also build a HTML and EPUB files (which is simple enough, using pandoc). I then clone the static site repo, copy the files in and run a site generation make target which transforms these raw files into the static site which is served.

Once this process is complete, I commit and push the changes and let the CI for the static site take care of publishing the new version. This full process for one of my projects looks like this:

image: andykuszyk/latex
pdf:
    script:
        - make pdf
        - make stamp
    artifacts:
        paths:
            - '*.pdf'
publish:
    script:
        - make html
        - make epub
        - make pdf
        - git clone https://$CREDENTIALS@github.com/andykuszyk/akuszyk.com
        - cp *.html akuszyk.com
        - cp *.pdf akuszyk.com
        - cp *.epub akuszyk.com
        - cd akuszyk.com
        - make generate-the-kingdom-of-tharg
        - git add the-kingdom-of-tharg/
        - git commit -m "Auto-update from GitLab"
        - git push

Regenerating my static site from raw LaTeX files

As mentioned above, whenever I push to a LaTeX project that is configured to be published, I generate an HTML file and then re-generate my static site. This process involves using some template files along with a bit of Python to take the HTML file converted by pandoc (from the original LaTeX sources) and transform it into a static site that is a bit more user friendly (with a touch of Bootstrap and Javascript). You can check out the site generator I’ve written at https://github.com/andykuszyk/akuszyk.com.

A key part of the generated site is the ability for readers to leave comments against each story, which I’ll cover in the next section.

A commit to the static site repo results in the site being packaged up into a new Docker image and deployed out to my web server.

Static site comments - the GitLab way

Part of the point of publishing my work in the first place was to solicit feedback from readers. I wanted people to be able to leave anonymous comments easily and quickly, without having to sign-up to any third party services, or sign-in with any social logins. It was also important for the comments to end up somewhere useful, where I could be notified about them, review them and work on them.

Since I store my writing projects in GitLab, the natural place for my to store the comments was as issues in each GitLab project. As a result, I wrote a small project that could create and list issues from a GitLab project, but expose an unauthenticated HTTP interface suitable for use with a public facing blog (or fiction site!). The idea is that this service has a personal access token that it uses to access the GitLab API and simply forwards on comment-style messages to and from GitLab.

Conclusion

With a version of this comments/issues service deployed alongside my static site, I was now in a position to publish my LaTeX projects at the push of a commit in a fully automated way, with the capability for readers to leave comments at the end of it. I hope this helps other budding authors in a similar position to me!