Example Shiny app that automatically installs a source package when deployed to shinyapps.io

app.R

# By default, the directories in .libPaths() aren't writable on shinyapps.io, so
# create a subdir where we'll install our package.
if (!file.exists("R-lib")) {
  dir.create("R-lib")
}
# Unfortunately, there's no way to get deployapp() to ignore this directory, so
# make sure to remove it locally before you call deployapp(). This can be done
# with:
#   unlink("pkgInst/R-lib", recursive = TRUE)

# You may also need to restart R before calling deployapp(), because calling
# runApp() will modify your libpath (below), which can confuse deployapp().

# Add ./R-lib/ to the libPaths
.libPaths( c(normalizePath("R-lib/"), .libPaths()) )

# Install the package if needed.
if (!do.call(require, list("myPackage"))) {
  install.packages("myPackage_0.1.tar.gz", repos = NULL, type = "source")
}

# Instead of `library(myPackage)`, we'll use do.call, to evade deployapp's
# checks for packages installed locally from source.
do.call(library, list("myPackage"))


shinyApp(
  ui = fluidPage(
    p(
      "The output of the function is: ",
      verbatimTextOutput("text")
    )
  ),
  server = function(input, output) {
    output$text <- renderPrint({
      # This function is in myPackage
      myFunc()
    })
  }
)

myPackage_0.1.tar.gz

Hi there,

Thanks for this gist! But does this approach also work at present? Nowadays R dependencies are captured with renv, when deploying an app. I don't seem to be able to reproduce the strategy of including a tarball with my app as suggested above...

Cheers

never mind, I had to wrap my requireNamespace in a do.call to evade dependency detection. So it still works!