Dockerfiles for various pandoc images
This repo contains a collection of Dockerfiles to build various pandoc container images.
Contents
Docker images hosted here have a "core" version and a "latex" version:
pandocand
pandoc-citeproc, as well as the appropriate backend for the full lua filtering backend (lua filters can call external modules).
pandocmight use, and any libraries needed by these packages (such as image libraries needed by the latex graphics packages).
From there, the tagging scheme is either
X.Y,
X.Y.Z,
latest, or
edge.
X.Yor
X.Y.Z: an official
pandocrelease (e.g.,
2.6). Once an
X.Ytag is pushed, it will not be re-built (unless there is a problem). Pandoc releases versions such as
2.7or
2.7.1(there is no
2.7.0), which is where the optional
.Zcomes from.
latest: the
latesttag points to the most recent
X.Yrelease. For example, if tags
2.5and
2.6were available online,
latestwould be the same image as
2.6.
edge: the "bleeding edge" tag clones the
masterbranch of
pandocand
pandoc-citeproc. This tag is a moving target, and will be re-built at least once a month. The CI scripts have a cron job to build each image stack on the first of the month. However, changes to the
masterbranch of this repository may also result in the
edgetag being updated sooner.
latestTag
The current
latesttag for all images points to
pandocversion
2.11.4.
pandoc/core
make alpine
pandoc/latex
make alpine-latex
Note: this section describes how to use the docker images. Please refer to the
pandocmanual for usage information aboutpandoc.
Docker images are pre-provisioned computing environments, similar to virtual machines, but smaller and cleverer. You can use these images to convert document wherever you can run docker images, without having to worry about pandoc or its dependencies. The images bring along everything they need to get the job done.
Install Docker if you don't have it already.
Start up Docker. Usually you will have an application called "Docker" on your computer with a rudimentary graphical user interface (GUI). You can also run this command in the command-line interface (CLI):
open -a Docker
cd path/to/source/dir
You can always run
pwdto check whether you're in the right place.
Let's say you have a
README.mdin your working directory that you'd like to convert to HTML.
docker run --rm --volume "`pwd`:/data" --user `id -u`:`id -g` pandoc/latex:2.6 README.md
The
--volumeflag maps some directory on your machine (lefthand side of the colons) to some directory in the container (righthand side), so that you have your source files available for pandoc to convert.
pwdis quoted to protect against spaces in filenames.
Ownership of the output file is determined by the user executing pandoc in the container. This will generally be a user different from the local user. It is hence a good idea to specify for docker the user and group IDs to use via the
--userflag.
pandoc/latex:2.6declares the image that you're going to run. It's always a good idea to hardcode the version, lest future releases break your code.
It may look weird to you that you can just add
README.mdat the end of this line, but that's just because the
pandoc/latex:2.6will simply prepend
pandocin front of anything you write after
pandoc/latex:2.6(this is known as the
ENTRYPOINTfield of the Dockerfile). So what you're really running here is
pandoc README.md, which is a valid pandoc command.
If you don't have the current docker image on your computer yet, the downloading and unpacking is going to take a while. It'll be (much) faster the next time. You don't have to worry about where/how Docker keeps these images.
Pandoc commands have a way of getting pretty long, and so typing them into the command line can get a little unwieldy. To get a better handle of long pandoc commands, you can store them in a script file, a simple text file with an
*.shextension such as
#!/bin/sh pandoc README.md
The first line, known as the shebang tells the container that the following commands are to be executed as shell commands. In our case, we really don't use a lot of shell magic, we just call pandoc in the second line (though you can get fancier, if you like). Notice that the
#!/bin/shwill not get you a full bash shell, but only the more basic ash shell that comes with Alpine linux on which the pandoc containers are based. This won't matter for most uses, but if you want to write writing more complicated scripts you may want to refer to the
ashmanual.
Once you have stored this script, you must make it executable by running the following command on it (this may apply only to UNIX-type systems):
chmod +x script.sh
You only have to do this once for each script file.
You can then run the completed script file in a pandoc docker container like so:
docker run --rm --volume "`pwd`:/data" --entrypoint "/data/script.sh" pandoc/latex:2.6
Notice that the above
script.shdid specify
pandoc, and you can't just omit it as in the simpler command above. This is because the
--entrypointflag overrides the
ENTRYPOINTfield in the docker file (
pandoc, in our case), so you must include the command.
GitHub Actions is an Infrastructure as a Service (IaaS) from GitHub that allows you to automatically run code on GitHub's servers on every push (or a bunch of other GitHub events).
Such continuous integration and delivery (CI/CD) may be useful for many pandoc users. Perhaps, you're using pandoc convert some markdown source document into HTML and deploy the results to a webserver. If the source document is under version control (such as git), you might want pandoc to convert and deploy on every commit. That is what CI/CD does.
To use pandoc on GitHub Actions, you can leverage the docker images of this project.
To learn more how to use the docker pandoc images in your GitHub Actions workflow, see these examples.
The official images are bare-bones, providing everything required to use pandoc and Lua filters, but not much more. Often, one will want to have additional software available. This is best achieved by building custom Docker images.
For example, one may want to use advanced spellchecking as demonstrated in the [spellcheck] in the Lua filters collection. This requires the aspell package as well as language-specific packages. A good solution would be to define a new Dockerfile and to use
pandoc/coreas the base package:
FROM pandoc/core:latest RUN apk --no-cache add aspell aspell-en aspell-fr
Create a new image by running
docker build --tag=pandoc-with-aspell .in the directory containing the Dockerfile. Now you can use
pandoc-with-aspellinstead of
pandoc/coreto get access to spellchecking in your image.
See Docker documentation for more details, for example part 2 of the Get Started guide.
When
pandochas a new official release, the following steps must be performed in this exact order:
Current `latest` Tagsection to include the new
pandocrelease number. Suppose we are releasing image stacks for
pandocversion 9.8:
$ git checkout -b release/9.8 # ... edit current :latest ... $ git add README.md $ git commit -m 'release=9.8' $ git push -u origin release/9.8
The important part is the commit message. The commit message is checked for
release=X.Y/
release=X.Y.Z, the diff does not really matter, just the message.
Create a pull request first to make sure all image stacks build as expected. 2. Assuming the pull request build succeeds, merge to
masterbranch. The only time that
docker pushis performed is when a commit hits the
masterbranch of this repository.
Code in this repository is licensed under the GNU General Public License Version 2.