dgabbe/rprofile.site

rprofile.site

Purpose and Features

Customize your R environment for R scripts and R GUIs (RStudio) independently of your .bash files while keeping the files under source code control. You can easily maintain the same environment across multiple computers. The user profile, .Rprofile, is left available to use in your $HOME directory or your project’s working directory.

Review and edit .First() in Rprofile.site and dot-Renviron to set R’s options to your preferences.

To save disk space and prevent reinstalling libraries when a new version of R is installed, the major versions use the same library, i.e.~/Library/R/3.x/library/, which is defined by R_LIBS_USER.

Jump to the Installation. Read Release Notes.

Diagnosing & Correcting Errors From Minor R Release Updates

Keeping a common library across the same major version of R can create some mischief when you install a minor update. For example, 3.3 to 3.4.0. The error I end up fixing is about loading and the first 2 lines are similar to:

1: dyn.load(file, DLLpath = DLLpath, ...)
2: library.dynam(lib, package, package.lib)

RStudio Hangs With A Blank Screen On Startup

It’s a bit vexing when RStudio hangs on startup with a blank screen. It’s not the fault of RStudio, but almost always of .First() loading devtools. Run R, which on a Mac is /Applications/R.app, to see the error.

Code Throws An Error About Loading

RStudio runs, but your code errors out with a stack trace indicating a loading problem. If you updated R while RStudio was running, try quitting and restarting RStudio. Continue reading if the error still occurs.

How to Fix

The load errors happen because your R library, pointed to by R_LIBS_USER, has packages installed which were built with an earlier version of R. When a new version of R is installed, its internals are newer and the version skew may be causing the error.

If RStudio hung, run R to find which package caused the error and move the offending package folder out of R_LIBS_USER. Repeat until RStudio starts.

Run .libPaths() to verify your library path is as intended. If R created ~/Library/R/3.4/library, delete it now. Now copy the package folder back. Run update.packages(ask = FALSE, checkBuilt = TRUE). Quit RStudio, and then restart it to confirm the error is fixed. More details can be found in this excellent post, https://shiny.rstudio.com/articles/upgrade-R.html.

Other commands that are helpful:

.Library
Sys.getenv("R_LIBS_USER")
sessionInfo()
readRenviron(path)

Installation

Follow these steps to install:

mkdir ~/rprofile.site
cd ~/rprofile.site
git init
git clone https://github.com/dgabbe/rprofile.site.git --branch Current

To complete the configuration, continue using the command line with this command

./installer.command

or return to the Finder and double click on installer.command.

If there is an existing Rprofile.site, it is renamed to Rprofile.site.org. Now edit Rprofile.site to make sure it has the options set to your preferences.
Repeat for .Renviron. Startup R or RStudio to verify a message like the one below is displayed:

23-Oct-2016 ~/Rprofile.site .First() starting...

    Option StringsAsFactors:   FALSE
    Option download.file.method:   libcurl
    Option repos:   https://cran.rstudio.com/
    Option defaultPackages: datasets, utils, grDevices, graphics, stats,
        methods

23-Oct-2016 ~/Rprofile.site .First() finished...

If you want to remove this setup:

~/rprofile.site/uninstaller.command
mv Rprofile.site.org Rprofile.site # only if you had one before

Use .First() or Packages to Customize

There are plenty of examples of .Rprofile or Rprofile.site that include functions or more generically, R objects. At first, I thought this was a great idea, but I have found better ways. Here are some points to keep in mind:

When Rprofile.site is executed, any objects are created in the base environment unless coded otherwise.
Storing the objects in .GlobalEnv (RStudio’s Environment tab displays Global Environment) avoids modifying base, but your changes are not permanent. Clearing the Global Environment will delete your objects. Some examples work around this behavior by creating a hidden environment and then adding their objects to it. However, RStudio’s delete all saved objects () default is to include hidden objects.
Don’t assume your R session is initialized only once. R and RStudio have slightly different behaviors. RStudio’s Session->Restart R does not clear out the Global Environment, but does run the R initialization process. The devtools:: functions do as well.

The idiom option("defaultPackages" = c(getOption("defaultPackages"), "package1", "package2"")) ends up appending your packages each time R initializaton runs. Since R will only load a library once, there is no change to your environment, but it’s poor technique.

If you still want default packages loaded, try this:

.First <- function() {
  suppressMessages(require("package1", quietly = TRUE))
  suppressMessages(require("package2", quietly = TRUE))
}

If you use library(), instead of require(), any error will stop the initialization process. It should take more than a missing package to bring the initialization to a halt.

R’s package facility provides a great solution. Writing a package will make you a better R programmer, your code will have fewer collisions with other R code, and RStudio’s deployment features will manage the packages needed.

Remember, loading is different from attaching packages. Once a package is installed, install.packages(), its functions can be called with package::function() syntax. The library function will attach the package to the current environment and make it part of the namespaces that are searched.

dgabbe / rprofile.site