• ~30m Packamon presentation: package overview, latest changes and features (Daan, Jasper, Marvin, Maxim)
• ~15m Real-world use case (Laure & Michela)
• ~15m Discussion

Continuous Integration is a software development practice where members of a team integrate their work frequently

Integration:

• push to a shared (git) repository
• build and compile code
• verify, check and test

Continuous/Frequently:

• as often as possible: with each commit
• automation is a necessity!

• Shorter development cycles

  • Faster release of new features
  • Frequent testing reduces the complexity of errors

• Automates and enforces good practices

• Reduces manual work

  • Saves (waiting) time
  • Reduces errors

• Early discovery of bugs

The name is a portmanteau of package and monitoring: if there are problems with your package, packamon will help you catch ’em all.

packamon contains tools for R package quality control:

• detect dependencies in local environment
• generate Dockerfiles to record dependencies
• generate Jenkinsfiles to automate quality control steps

Jenkins is a self-contained, open source automation server which can be used to automate all sorts of tasks related to building, testing, and delivering or deploying software.

(source: https://www.jenkins.io/doc/)

The OA jenkins can be found at https://ci.openanalytics.eu

A Jenkinsfile describes a sequence of steps to be run by Jenkins every time one or more commits are pushed to a repository.

These steps are grouped into stages.

Stages run on a particular agent.

pipeline {
    agent {
        # agent description
    }
    stages {
        stage('build') {
            steps {
                # steps go here
            }
        }
        stage('check') {
            steps {
                # steps go here
            }
        }
    }
}

What steps should be run for an R package?

What steps should be run for an R package?

• (Rcpp)
• roxygen
• R CMD build
• R CMD check
• tests
• … others?

For reproducibility, we use docker containers as ephemeral agents.

The image used to run the container needs to have everything to run the steps we defined:

• (a specific version of) R installed
• all R package dependencies (Imports, Suggest, Depends)
• all system dependencies

First stage in the Jenkinsfile: run docker build with the Dockerfile included in the repository.

All remaining stages are run using in a container created from the resulting image.

(source: https://xkcd.com/1629/)

1. run packamon::init() in a new project

2. (optional) run packamon::writeJenkinsfile() or packamon::writeDockerfile() for more flexibility

3. commit and push the generated files:

   • Dockerfile
   • Jenkinsfile
   • template.Dockerfile

4. continue working on your package as normal

5. if your dependencies change, run packamon::writeDockerfile() again

The best way to see if your code is of good quality is to test it. Therefore we added a default step that runs all testthat tests and reports the code coverage using covr.

JUnit export: Keeps track of which tests succeeded/skipped/failed for each commit. This way you can easily detect at which point code got broken.

Coverage: Shows how much code of your package is captured within the tests. One should try to get the coverage as high as possible.

Problem: after Dockerfile has been generated, difficult to add extra non-autodetected dependencies since next generate will overwrite them.

Solution: Templates!

Templates are regular Dockerfiles with an extra #include directive.

Lines of the form

#include packamon.<step>

will be replaced in the final Dockerfile by packamon::writeDockerfile() or packamon::init()

Other lines (docker command or comments) will be copied to the generated Dockerfile.

The default template:

#include packamon.disclaimer

#include packamon.from

#include packamon.system-dependencies

#include packamon.r-repos

#include packamon.r-dependencies

#include packamon.local-r-dependencies

#include packamon.runtime-settings

Using a different base image and extra sysdeps:

#include packamon.disclaimer

FROM openanalytics/my-super-cool-image:latest

#include packamon.system-dependencies
RUN apt update && apt install libfunny

#include packamon.r-repos

#include packamon.r-dependencies

#include packamon.local-r-dependencies

#include packamon.runtime-settings

Continuous Delivery (CD): automated release of CI results

• automatically push new software versions to a repository
• simplifies release of new software versions
• reduces manual work

Continuous Deployment (also CD):

• automated deployment of the CI/CD output

Compact Disc

Yes!

• Docker images can be used to run beyond CI context
• R packages in RDepot (next packamon release)

Install from RDepot:

install.packages("packamon", repos = c(
    rdepot = "https://rdepot-dev.openanalytics.eu/repo/public", 
    getOption("repos")))

Alternatively pull from https://scm.openanalytics.eu/git/packamon and install:

install.packages("~/git/packamon/packamon", repos = NULL)

You can take a look at the getting-started vignette if you get lost.