Software Exercise 2.3: Apptainer examples¶

Objective: Customize an existing apptainer definition file to create a conda, Python, R or Julia environment.

Why learn this?: A large amount of research software is based on existing base languages (like R or Python), and these examples should be useful for getting started with building a custom container environment.

Overview¶

In this set of exercises, we have provided sample definition files as a starting point for building a container. In all these examples, each container starts with an existing container with the base language or package manager.

After each example, we will include links to other, similar definition files, stored in CHTC's Recipes Repository.

If you want to learn more about the components of a definition file, see the examples in Example 2.4.

What to do¶

Find the definition file that looks most similar to your software installation instructions.
(Optional) Look at other examples in the CHTC recipes repository to see if there is another example that might be a better fit.
Edit the definition file to install your needed packages or libraries.
Try to build and test a container.
1. If you need to review the process for building and testing a container see Exercise 1.4

Sample definition files¶

Conda¶

The placeholder libraries in this example are cowsay and fortunes; replace the names of these libraries with the ones you want to install.

Bootstrap: docker
From: continuumio/miniconda3:latest

%post
    conda install python=3.10

See other conda definition file examples here: Conda definition files

R / tidyverse¶

The placeholder libraries in this example are cowsay and fortunes; replace the names of these libraries with the ones you want to install.

Most R programs rely on libraries that are part of the tidyverse; if you are using any of the common tidyverse programs, you should use a definition file like this:

Bootstrap: docker
From: rocker/tidyverse:4.3.1

%post
    R -e "install.packages(c('cowsay','fortunes'), dependencies=TRUE, repos='http://cran.rstudio.com/')"

If you aren't using any of the tidyverse packages, you can build on a base R package with this definition file:

Bootstrap: docker
From: rocker/r-ver:4.3.1

%post
    R -e "install.packages(c('cowsay','fortunes'), dependencies=TRUE, repos='http://cran.rstudio.com/')"

See other R definition file examples here: R definition files

Python / pip¶

The placeholder libraries in this example are scipy and cowsay; replace the names of these libraries with the ones you want to install.

Bootstrap: docker
From: python:3.11

%post
    python3 -m pip install scipy cowsay

See other Python definition file examples here: Python definition files

Julia¶

The placeholder libraries in this example are Cowsay and DataFrames; replace the names of these libraries with the ones you want to install.

Bootstrap: docker
From: julia:1.10

%post
        export JULIA_DEPOT_PATH="/opt/julia"

        # Install your packages using the following command:
        julia -e 'using Pkg; Pkg.add(["Cowsay", "DataFrames"]); Pkg.instantiate(); Pkg.precompile()'

%environment
        export JULIA_DEPOT_PATH=":/opt/julia"

See other Julia definition file examples here: Julia definition files