21 GitHub Actions

Published

2024-09-11

TLDR

CI/CD

Continuous Integration (CI) and Continuous Deployment (CD) help automate software development tasks, especially testing and deployment. In the context of a Shiny app-package, CI/CD usually refers to GitHub Actions or Travis CI.¹

Accessing the code examples

I’ve created the shinypak R package In an effort to make each section accessible and easy to follow:

Install shinypak using pak (or remotes):

# install.packages('pak')
pak::pak('mjfrigaard/shinypak')

Review the chapters in each section:

library(shinypak)
list_apps(regex = '^16')
## # A tibble: 3 × 2
##   branch          last_updated       
##   <chr>           <dttm>             
## 1 16.1_gha-style  2024-08-23 07:17:12
## 2 16.2_gha-shiny  2024-08-23 07:13:54
## 3 16.3_gha-docker 2024-08-01 12:49:40

Launch an app:

launch(app = "16.1_gha-style")

Continuous integration

Automated Testing: CI can automate the tests (testthat and shinytest2) for the code in your app-package, which is helpful when pushing a new feature or bug fix.
Quality Assurance: CI helps maintain the code quality by running a series of ‘checks and balances’ to confirm updates and changes don’t break anything. Should something go wrong, CI will alert developers so they can fix it promptly.
Change Management: CI systems work hand-in-hand with version control systems (like Git), which makes it possible to track all the changes within a project. This makes sure every modification is checked and is safe to be included in the app-package pushed to the main branch.

Continuous deployment

Automated Deployment: Once CI confirms the changes are safe, CD can automatically deploy your updated Shiny app-package to a package management system or Shiny server, which means that the updated version becomes available to users without any manual intervention.
Consistent Updating: CD ensures that users experience new features and fixes much faster. Improvements are deployed as soon as they are made and tested, which leads to a more responsive development process.
Streamlined Development: CI and CD streamline the software development process so developers can focus more on building and less on testing and deploying.

Properly implemented CI/CD is like a well-oiled machine, where each part works seamlessly with the others, which saves time, reduces errors, and makes the process efficient and reliable.

GitHub Actions

GitHub Actions (or Actions, for short) is a feature provided by GitHub that enables automation of various development workflows. These workflows are defined in YAML files and can automate some of the repetitive tasks in the development lifecycle, right within a GitHub repository.

Actions allow us to create custom workflows to automatically build, test, and deploy our code. Workflows can be triggered by a push to a branch, a pull request, a specific time schedule, or another GitHub event.

The workflow file

The YAML workflow file configures and defines the automated tasks run in a GitHub repository. Workflows typically include fields for building, testing, and deploying code, and are structured into several key sections:

`on`

“An event is a specific activity in a repository that triggers a workflow run.”

on:
  push:
    branches: [ main ]
  pull_request:
    branches: [ main ]

on defines the event(s) that trigger the workflow.²

`name`

name: shiny

name is an optional field to identify a workflow in the Actions tab of the GitHub repository.³

`jobs`

“A job is a set of steps in a workflow that is executed on the same runner.”

jobs:
  check:
    runs-on: ubuntu-latest

The job and it’s identifier (check) are used to reference the runner, in this example it’s ubuntu-latest.

jobs:
  check:
    runs-on: ${{ matrix.config.os }}

matrix runs jobs across different operating systems, programming language versions, etc.

    name: ${{ matrix.config.os }} (${{ matrix.config.r }})

    strategy:
      fail-fast: false
      matrix:
        config:
          - {os: macos-latest,   r: 'release'}
          - {os: windows-latest, r: 'release'}
          - {os: ubuntu-latest,   r: 'release'}

fail-fast: false means the workflow will continue running even if one of the matrix jobs fails.

`steps`

“Steps can run commands, run setup tasks, or run an action in your repository, a public repository, or an action published in a Docker registry. Not all steps run actions, but all actions run as a step.”

steps:
    - uses: actions/checkout@v2

    - name: Set up R
      uses: r-lib/actions/setup-r@v2

    - name: Install package
      run: |
        install.packages('remotes')
        remotes::install_local('.')
      shell: Rscript {0}

steps define a series of tasks or actions that the job will execute. These steps check out the code, set up an R environment, and installs a local package

What is shell: Rscript {0}?

  name: Install packages
  run: |
    pkgs <- c('glue', 'cli')
    install.packages(pkgs)
  shell: Rscript {0}

The command install.packages(pkgs) is run as an R script:

shell specifies which command-line interpreter (bash, pwsh, python, or Rscript) to use for the run commands
Rscript is the command-line tool provided by R to execute R scripts and commands in a non-interactive environment.
{0} is a placeholder for the R commands written in the run section.

Comments

Lines beginning with # will not be executed. It’s common to provide 1) a reference to the workflow source (r-lib/actions in this case) and 2) a link for help with debugging build failures.

# Workflow derived from https://github.com/r-lib/actions/tree/v2/examples
# Need help debugging build failures? Start at https://github.com/r-lib/actions#where-to-find-help

Workflow permissions

In order for workflows to run, we’ll need to make sure Actions have read and write permissions. We can do this by clicking on the repository’s Settings tab, then expand the Actions menu and select General:

In the General settings, we want to confirm the workflow has Read and write permissions permissions (if this needs to be been changed, be sure to click Save).

Now our workflows can commit and push changes to the repository.

21.1 Code styling and linting

In R packages, we can set up the GitHub Action infrastructure with usethis::use_github_action(). The name can be any of the workflow files available at r-lib/actions. The first workflow we’ll be demonstrating can be used to automate the code style in an app-package.

Launch app with the shinypak package:

launch('16.1_gha-style')

usethis::use_github_action("style")

✔ Setting active project to '/projects/apps/sap'
✔ Creating '.github/'
✔ Adding '*.html' to '.github/.gitignore'
✔ Creating '.github/workflows/'
✔ Saving 'r-lib/actions/examples/style.yaml@v2' to '.github/workflows/style.yaml'

The output tells us a local .github/workflows/style.yaml file has been copied from the r-lib/actions/examples/style.yaml@v2 into the .github/workflows folder:

.github/
  └── workflows
    └── style.yaml

2 directories, 1 file

The documentation for the style.yaml@v2 file tells us this workflow,

“styles the R code in a package, then commits and pushes the changes to the same branch.”

The code styling in this workflow is handled by the styler package, which “formats your code according to the tidyverse style guide.”

Three functions in the 16.1_gha-style branch have altered to have zero style (R/display_type.R, R/mod_var_input.R, and R/test_logger.R). You can copy the code for these functions in the chunk below:⁴

show/hide no-style code

# display_type ----
display_type <- function(run = "w") {
  if(Sys.getenv("RSTUDIO") == "1"){
    switch(run,
      p=options(shiny.launch.browser=.rs.invokeShinyPaneViewer),
      b=options(shiny.launch.browser=.rs.invokeShinyWindowExternal),
      w=options(shiny.launch.browser=.rs.invokeShinyWindowViewer),
      NULL=options(shiny.launch.browser=NULL))
    environment <- "RStudio"
    shinyViewerType <- getOption('shiny.launch.browser') |> 
                        attributes() |> unlist() |> unname()
    cli::cli_alert_info("App running in {environment}")
    cli::cli_alert_info("shinyViewerType set to {shinyViewerType}")
  }else{
    environment <- "RStudio"
    cli::cli_alert_info("App not running in {environment}")
  } 
}

# mod_var_input_ui ----
mod_var_input_ui <- function(id) {
  ns <- NS(id)
  tagList(
    selectInput(
      inputId = ns("y"),
      label = "Y-axis:",
      choices = c("IMDB rating" = "imdb_rating",
        "IMDB number of votes" = "imdb_num_votes",
        "Critics Score" = "critics_score",
        "Audience Score" = "audience_score",
        "Runtime" = "runtime"), selected = "audience_score"
    ),
    selectInput(inputId = ns("x"),
      label = "X-axis:",
      choices = c("IMDB rating" = "imdb_rating",
        "IMDB number of votes" = "imdb_num_votes",
        "Critics Score" = "critics_score",
        "Audience Score" = "audience_score",
        "Runtime" = "runtime"), selected = "imdb_rating"
    ),
    selectInput(inputId = ns("z"),
      label = "Color by:",
      choices = c("Title Type" = "title_type",
        "Genre" = "genre",
        "MPAA Rating" = "mpaa_rating",
        "Critics Rating" = "critics_rating",
        "Audience Rating" = "audience_rating"), selected = "mpaa_rating"
    ),
    sliderInput(inputId = ns("alpha"),
      label = "Alpha:",
      min = 0, max = 1, step = 0.1, value = 0.5
    ),
    sliderInput(
      inputId = ns("size"),
      label = "Size:", min = 0, max = 5, value = 2
    ),
    textInput(inputId = ns("plot_title"),
      label = "Plot title",
      placeholder = "Enter plot title"
    )
  )
}
# test_logger ----
test_logger <- function(start = NULL, end = NULL, msg) {
  
  if (is.null(start) & is.null(end)) {
    
    cat("\n")
    
    logger::log_info("{msg}")
    
    } else if (!is.null(start) & is.null(end)) {
    
    cat("\n")
    
    logger::log_info("\n[ START {start} = {msg}]")
  
    } else if (is.null(start) & !is.null(end)) {
      
    cat("\n")
      
    logger::log_info("\n[ END {end} = {msg}]")
    
    } else {
      
    cat("\n")
      
    logger::log_info("\n[ START {start} = {msg}]")
    
    cat("\n")
    
    logger::log_info("\n[ END {end} = {msg}]")
    
    }
  
}

We’ll make one small change to style.yaml before pushing it to GitHub. Instead of having the workflow automatically commit and push the styled code changes to the same branch, we’ll limit the code styling to the 16.1_gha-style branch:

on:
  push:
    branches: [16.1_gha-style]
    paths: ["**.[rR]", "**.[qrR]md", "**.[rR]markdown", "**.[rR]nw", "**.[rR]profile"]

This change ensures our style workflow will be triggered only for pushes to the specified branches (and when changes are made to files with the specified extensions in path).

After saving these changes to .github/workflows/style.yaml, we’ll add, commit, and push the changes to GitHub,

git add .
git commit -m "updates to style workflow"

[16.1_gha-style 899bd38] updates to style workflow
 4 files changed, 43 insertions(+), 45 deletions(-)

git push

Enumerating objects: 17, done.
Counting objects: 100% (17/17), done.
Delta compression using up to 12 threads
Compressing objects: 100% (8/8), done.
Writing objects: 100% (9/9), 950 bytes | 25.00 KiB/s, done.
Total 9 (delta 6), reused 0 (delta 0), pack-reused 0
remote: Resolving deltas: 100% (6/6), completed with 6 local objects.
To https://github.com/mjfrigaard/sap.git
   d366e0f..899bd38  16.1_gha-style -> 16.1_gha-style

If we view the Actions tab, we’ll see the workflow listed with the name of our commit message:

Clicking on the workflow reveals the following:

The style widget contains each step in the workflow file. If we click on it we can see each step executed in real-time. Below we can see the dependencies installed from the pak lockfile:

Click on the style widget — Click on the **style** widget

View the Install dependencies workflow step running — View the **Install dependencies** workflow step running

The actual styling comes is in the Style step:

Style workflow step running — **Style** workflow step running

When the workflow has finished, we can pull the style changes to our local branch and view the styled files:⁵

git pull

remote: Enumerating objects: 4, done.
remote: Counting objects: 100% (4/4), done.
remote: Total 4 (delta 3), reused 4 (delta 3), pack-reused 0
Unpacking objects: 100% (4/4), 450 bytes | 56.00 KiB/s, done.
From https://github.com/mjfrigaard/sap
   899bd38..0ad97cf  16.1_gha-style -> origin/16.1_gha-style
Updating 899bd38..0ad97cf
Fast-forward
 R/display_type.R  | 13 +++----------
 R/mod_var_input.R | 30 ++++++++++++++++++++----------
 R/test_logger.R   | 30 ++++++++++--------------------
 3 files changed, 33 insertions(+), 40 deletions(-)

21.2 Shiny app deployment

In this section we’ll be adapting the GitHub Action workflow file provided for deploying a Shiny applications. Before we get to the workflow file, we need to address a few settings and configurations that need to take place outside of the workflow file.

Launch app with the shinypak package:

launch('16.2_gha-shiny')

21.2.1 Creating repository secrets

We need to make sure our rsconnect secrets are stored with the GitHub repository (so we can access them from within the workflow file). To do this, we’re going to return to the Settings tab on the sap repository, expand Secrets and variables under Security, and select Actions:⁶

Secrets and variables for sap repository — **Secrets and variables** for `sap` repository

Use New repository secret to create three new secrets: RSCONNECT_USER, RSCONNECT_TOKEN, and RSCONNECT_SECRET. You can access these in your shinyapps.io account:

Token and Secret from shinyapps.io — **Token** and **Secret** from shinyapps.io

When you’re finished, should see the following three Repository secrets:

Repository secrets in sap — **Repository secrets** in `sap`

21.2.2 Enable private repositories

We also need to make sure our shinyapps.io account allows us to install from private GitHub repositories during a workflow deployment. We can do this in the Profile menu item in our dashboard:

Update Authentication in shinyapps.io profile

Clicking Update Authentication will open your Posit profile settings, where you can enable private repositories.

21.2.3 Capture dependencies

The documentation for ‘Shiny App Deployment’ indicates we need to create a renv lockfile in the 16.2_gha-shiny branch.⁷

We can create the lockfile with the code below:

install.packages('renv')
renv::init(force = TRUE)

This project contains a DESCRIPTION file.
Which files should renv use for dependency discovery in this project? 

1: Use only the DESCRIPTION file. (explicit mode)
2: Use all files in this project. (implicit mode)

We’re going to use the explicit mode, because the dependencies in the DESCRIPTION should be all that’s needed to run and deploy the application.

Selection: 1

- Using 'explicit' snapshot type. Please see `?renv::snapshot` for more details.

This project already has a private library. What would you like to do? 

1: Activate the project and use the existing library.
2: Re-initialize the project with a new library.
3: Abort project initialization.

We will start fresh and create a new project library:

Selection: 2
- Resolving missing dependencies ...

After renv has discovered and installed the dependencies, the R session will restart and we’ll see the following new folder/files:

├── .Rprofile
├── renv/
│   ├── activate.R
│   └── settings.json
└── renv.lock

1: .Rprofile contains a call to source("renv/activate.R"), which manages the project-level dependencies
2: renv settings
3: renv lockfile

When using renv in package development, we want to be sure to run renv::install() and renv::update():

renv::install()

- There are no packages to install.

renv::update()

- Checking for updated packages ... Done!
- All packages appear to be up-to-date.

21.2.4 `shiny-deploy.yaml`

We can include the ‘Shiny App Deployment’ workflow file in sap with usethis::use_github_action("shiny-deploy"):⁸

usethis::use_github_action("shiny-deploy")

✔ Creating '.github/'
✔ Adding '^\\.github$' to '.Rbuildignore'
✔ Adding '*.html' to '.github/.gitignore'
✔ Creating '.github/workflows/'
✔ Saving 'r-lib/actions/examples/shiny-deploy.yaml@v2' to '.github/workflows/shiny-deploy.yaml'

This example workflow file is designed to deploy a Shiny application to a server. We’re going to make a few changes to shiny-deploy.yaml so it will deploy the application stored in the 16.2_gha-shiny branch.

The example shiny workflow includes calls to rsconnect::setAccountInfo() and rsconnect::deployApp(). We’ll perform a sanity check and confirm we can deploy the application using the information we’ve collected:

Start by entering your rsconnect info:

install.packages('rsconnect')
rsconnect::setAccountInfo(
  name = "mjfrigaard",
  token = "YOUR TOKEN", 
  secret = "YOUR SECRET")

Then try to deploy the application using rsconnect::deployApp():

library(sap)
rsconnect::deployApp(
  appName = "shinyAppPkgsCICD",
  account = "mjfrigaard",
  server = "shinyapps.io",
  forceUpdate = TRUE)

Reading the deployment log will tells us if the secret, token, configuration, and lockfile are all working:

── Preparing for deployment ──────────────────────────────────────────────────
✔ Deploying "shinyAppPkgsCICD" using "server: shinyapps.io / username: mjfrigaard"
ℹ Bundling 50 files: .github/workflows/shiny-deploy.yaml, .Rbuildignore, ..., 
tests/testthat.R, and vignettes/test-specs.Rmd
ℹ Capturing R dependencies with renv
✔ Found 99 dependencies
✔ Created 2,431,325b bundle
ℹ Uploading bundle...
✔ Uploaded bundle with id 8130675
── Deploying to server ─────────────────────────────────────────────────────────
Waiting for task: 1372980209
  building: Processing bundle: 8130675
  building: Building image: 9770352
  building: Fetching packages
  building: Installing packages
  building: Installing files
  building: Pushing image: 9770352
  deploying: Starting instances
  unstaging: Stopping old instances
── Deployment complete ─────────────────────────────────────────────────────────
✔ Successfully deployed to <https://mjfrigaard.shinyapps.io/shinyAppPkgsCICD/>

Making sure my application will deploy locally with the code I plan on putting in a workflow file makes me confident it run when it’s triggered. Below we’ll adapt the .github/workflows/shiny-deploy.yaml file to deploy sap.

21.2.4.1 Trigger

on:
  push:
    branches: [16.2_gha-shiny]
    
name: shiny

The event we want the workflow triggered on is a push to the 16.2_gha-shiny branch
We’ll change the name to shiny

21.2.4.2 Jobs

jobs:
  shiny-deploy:
    runs-on: ubuntu-latest
    env:
      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}

jobs remains unchanged, but the identifier should match the name. ⁹

21.2.4.3 Steps

    steps:
      - uses: actions/checkout@v3

Check out the repository code

      - uses: r-lib/actions/setup-pandoc@v2

Set up Pandoc for document conversions

      - uses: r-lib/actions/setup-r@v2
        with:
          use-public-rspm: true

Set up R environment
a. use-public-rspm is the public RStudio package manager

      - uses: r-lib/actions/setup-renv@v2

Set up renv to manage the project-specific dependencies captured in renv.lock

      - name: Install rsconnect
        run: install.packages("rsconnect")
        shell: Rscript {0}

Install rsconnect the step to deploy our app to the Shiny server
a. run installs the rsconnect package

In the following steps, we’ll provide our account username (ACCOUNT) and a name for our app (shinyAppPkgsCICD). This will deploy an application at https://<ACCOUNT>.shinyapps.io/<APPNAME>/

    - name: Authorize and deploy app
      env: 
        APPNAME: shinyAppPkgsCICD
        ACCOUNT: mjfrigaard
        SERVER: shinyapps.io

Authorize and deploy app defines the step to deploy our Shiny app
env sets up the environment variables APPNAME, ACCOUNT, and SERVER

The final run step sets the account information using rsconnect::setAccountInfo() and deploys the app to the specified server using rsconnect::deployApp():

    run: |
      rsconnect::setAccountInfo("${{ secrets.RSCONNECT_USER }}", "${{ secrets.RSCONNECT_TOKEN }}", "${{ secrets.RSCONNECT_SECRET }}")
      rsconnect::deployApp(appName = "${{ env.APPNAME }}", account = "${{ env.ACCOUNT }}", server = "${{ env.SERVER }}", forceUpdate = TRUE)
    shell: Rscript {0}

When we add, commit, and push the changes to the repo, we see the following deployment log:

git add .
git commit -m "shiny deploy workflow"
git push

[16.2_gha-shiny 7953a5c] shiny deploy workflow
 10 files changed, 2130 insertions(+)
 create mode 100644 .Rprofile
 create mode 100644 .github/.gitignore
 create mode 100644 .github/workflows/shiny-deploy.yaml
 create mode 100644 renv.lock
 create mode 100644 renv/.gitignore
 create mode 100644 renv/activate.R
 create mode 100644 renv/settings.json
 create mode 100644 rsconnect/shinyapps.io/mjfrigaard/shinyAppPkgsCICD.dcf

We can see the workflow running on the Actions tab of the sap repository:

When it’s finished, we can see the deployment log from the workflow looks similar to the log we saw locally:

You can view the deployed application here: https://mjfrigaard.shinyapps.io/shinyAppPkgsCICD/

21.3 Shiny apps & Docker

We covered Docker in the previous chapter, which builds and application in an image and containerizes it’s deployment. In this section we’ll combine Docker and GitHub Actions to automate deployment of our app. Automating app deployments will minimize the risk of environmental discrepancies (avoiding the “it works on my machine” problem) and manual deployment processes. Plus, if something goes wrong after a workflow run, rolling back to a previous version in Git is straightforward.

Launch app with the shinypak package:

launch('16.3_gha-docker')

21.3.1 Secrets and tokens

Most of the steps below are covered in the excellent blog post by Nicola Rennie ¹⁰ I’ve updated the contents to work with a Shiny app-package.

To make sure our Connect username, token, and secret travels with our Dockerfile, we will need to store them as environmental variables.

Environmental variables can be set/unset with Sys.setenv()/Sys.unsetenv():

Sys.setenv(RSCONNECT_USER = '<username>', 
  RSCONNECT_TOKEN = '<token>', 
  RSCONNECT_SECRET = '<secret>')

These environmental variables are be passed to rsconnect::setAccountInfo() with a deploy.R file.

21.3.2 `deploy.R`

deploy.R contains the calls to rsconnect::setAccountInfo() and rsconnect::deployApp() we had in our previous workflow file. Sys.getenv() will retrieve the username, token, and secret stored with Sys.setenv():

setAccountInfo(name = Sys.getenv("RSCONNECT_USER"),
               token = Sys.getenv("RSCONNECT_TOKEN"),
               secret = Sys.getenv("RSCONNECT_SECRET"))
deployApp(appDir = ".", 
  appName = "shinyAppPkgsDockerCiCd", 
  account = "mjfrigaard", 
    server = "shinyapps.io", 
    forceUpdate = TRUE)

The deploy.R script should be placed in the root folder of sap. By default, this will deploy the application launched with app.R. We’ll change this to the application launched with ggp2_movies_app() by changing the contents of app.R to the following:

show/hide updated app.R

# set option to turn off loadSupport() ----
withr::with_options(new = list(shiny.autoload.r = FALSE), code = {
  if (!interactive()) {
    sink(stderr(), type = "output")
    tryCatch(
      expr = {
        library(sap)
      },
      error = function(e) {
        pkgload::load_all()
      }
    )
    # create shiny object from dev/ ----
    shinyAppDir(appDir = 
                system.file("dev", package = "sap"))
  } else {
    pkgload::load_all()
    sap::ggp2_movies_app(options = list(test.mode = FALSE))
  }
})

21.3.3 `Dockerfile`

The Dockerfile will look similar to the example in the Docker chapter, but with a few important changes:

The /home/shinyAppPkgsDockerCiCd directory is not only the image location we’ll be launching the application from, it’s also the URL for the deployed app:
- https:// username .shinyapps.io/shinyAppPkgsDockerCiCd/
We’ll include rsconnect and bslib in the list of packages to install.
The final command runs the application from the deploy.R file

FROM rocker/shiny
RUN mkdir /home/shinyAppPkgsDockerCiCd
ADD . /home/shinyAppPkgsDockerCiCd
WORKDIR /home/shinyAppPkgsDockerCiCd
RUN R -e 'install.packages(c("rlang", "stringr", "shiny", "ggplot2", "remotes", "rsconnect", "bslib"))'
RUN R -e 'remotes::install_local(upgrade="never")'
EXPOSE 8180
CMD Rscript deploy.R

21.3.4 `docker.yml`

name: docker-shiny, shinyrpkgs

on:
  push:
    branches: [ 16.3_gha-docker ]

jobs:
  docker:
    runs-on: ubuntu-latest
    
    steps:
      - uses: actions/checkout@v3

      - name: Build image
        run: docker build -t shinyrpkgs .

Add a name for the Docker/Shiny workflow
Specify the triggering event the workflow will run on
The job will run on ubuntu-latest (with docker id)
Checkout the code
Build the docker image with docker build

The final execute step is runs docker run and passes our environment variables to the secrets we have stored in GitHub (i.e., with secrets.<SECRET_NAME>):

      - name: execute
        run: >
          docker run -e RSCONNECT_USER=${{ secrets.RSCONNECT_USER }} -e RSCONNECT_TOKEN=${{ secrets.RSCONNECT_TOKEN }}  -e RSCONNECT_SECRET=${{ secrets.RSCONNECT_SECRET }} shinyrpkgs

21.3.5 App dependencies

rsconnect has a handy appDependencies() function that returns a data.frame of packages, versions, and repositories for your application:

head(rsconnect:::appDependencies(appDir = "."))

#>        Package  Version Source               Repository
#> 1  AsioHeaders 1.22.1-2   CRAN https://cran.rstudio.com
#> 2         MASS   7.3-60   CRAN https://cran.rstudio.com
#> 3       Matrix    1.6-4   CRAN https://cran.rstudio.com
#> 4           R6    2.5.1   CRAN https://cran.rstudio.com
#> 5 RColorBrewer    1.1-3   CRAN https://cran.rstudio.com
#> 6         Rcpp   1.0.12   CRAN https://cran.rstudio.com

Passing appDependencies() in your app-package’s root directory will tell you if any packages have missing Source or Repository values (this can cause the deployment to fail).

I also recommend running attachment::att_amend_desc() to capture all the dependencies in the DESCRIPTION. For example, att_amend_desc() added the following packages to the Suggests field in the DESCRIPTION:¹¹

[+] 2 package(s) added: shinytest2, testthat.

21.3.6 Docker build

In the docker workflow log, we can see the commands from our Dockerfile executed to install the necessary packages:

Installing R packages from Dockerfile — Installing R packages from `Dockerfile`

After installing the dependencies, the sap.tar.gz is installed from *source*:

Building our app-package from *source* — Building our app-package from `*source*`

The output above should look familiar–it’s very similar to what we see when we run devtools::load_all().

Capturing and recording dependencies with renv

In the Shiny workflow example above, we captured the application dependencies with renv. However, in this Docker and Shiny workflow, we didn’t create a renv.lock file or a renv/ folder.

We get away with this because when the workflow runs, renv is used to capture the dependencies (line 9-12 in the output below). We see a message about the packages listed in the Suggests field or our DESCRIPTION:

Not having shinytest2 and testthat packages installed won’t interfere with our application deploying because these packages are listed in the Suggests field in the DESCRIPTION file and are used intests/.

After installing the dependencies, we can see the Shiny deployment log execute in the workflow:

View the deployed application here.

Recap

Recap: GitHub Actions

GitHub Actions can be used to set up the necessary R environment and dependencies required to deploy the application from your app-package, while allowing secure account credentials stored in GitHub secrets.
The workflows in this chapter are only triggered by changes to the 16.1_gha-style, 16.2_gha-shiny, and 16.3_gha-docker branches of the sap repository.
GitHub Actions can streamline a secure deployment process, which allows you to focus more on developing your application.
Read about GitHub Actions in the rhino framework in this appendix section

See a typo, error, or something missing?

Please open an issue on GitHub

This chapter will only cover CI/CD with GitHub Actions. Travis-CI has been around longer than GitHub Actions, and if you’d like to compare the two, I recommend this article.↩︎
Triggers can be push and/or pull events to specific branches, creating a release, commenting on an issue, or even configured to run at scheduled times.↩︎
If name is not provided, GitHub will use the file path↩︎
We’ll be covering code styling in the Style chapter.↩︎
You can find more examples of common GitHub Actions for R packages in r-lib/actions/examples.↩︎
GitHub secrets should be specified by the user following the instructions in the Creating secrets for a repository article.↩︎
“This action assumes you have an renv lockfile in your repository that describes the R packages and versions required for your Shiny application.” - Shiny App Deployment ↩︎
Instructions for the GitHub Action Shiny deployment can be found here ↩︎
run-on is the latest Ubuntu runner and env is the GITHUB_PAT environment variable is your GitHub personal access token (access a secret with secrets.<SECRET_NAME>)↩︎
Automatically deploying a Shiny app for browsing #RStats tweets with GitHub Actions. Nicola Rennie. October 3, 2022.↩︎
att_amend_desc() adds a dev folder with a YAML config file, but you can remove this before building/deploying.↩︎

GitHub Actions

The workflow file

on

name

jobs

steps

Comments

Workflow permissions

21.1 Code styling and linting

21.2 Shiny app deployment

21.2.1 Creating repository secrets

21.2.2 Enable private repositories

21.2.3 Capture dependencies

21.2.4 shiny-deploy.yaml

21.2.4.1 Trigger

21.2.4.2 Jobs

21.2.4.3 Steps

21.3 Shiny apps & Docker

21.3.1 Secrets and tokens

21.3.2 deploy.R

21.3.3 Dockerfile

21.3.4 docker.yml

21.3.5 App dependencies

21.3.6 Docker build

Recap

`on`

`name`

`jobs`

`steps`

21.2.4 `shiny-deploy.yaml`

21.3.2 `deploy.R`

21.3.3 `Dockerfile`

21.3.4 `docker.yml`