update multiple R versions doc

achubaty · Jun 29, 2023 · 6ee8328 · 6ee8328
1 parent d7b3629
commit 6ee8328
Show file tree

Hide file tree

Showing 2 changed files with 114 additions and 30 deletions.
diff --git a/using-multiple-R-versions-on-linux.Rmd b/using-multiple-R-versions-on-linux.Rmd
@@ -1,7 +1,7 @@
 ---
 title: "Using multiple R versions on Linux"
 author: "Alex M. Chubaty"
-date: "20 Sep 2022"
+date: "29 Jun 2023"
 output: 
   pdf_document: 
     highlight: pygments
@@ -15,23 +15,29 @@ editor_options:
 knitr::opts_chunk$set(echo = FALSE)
 ```
 
+# Overview
+
 Unlike on Windows, most R users on linux tend to work with a single version of R -- whichever version is available in their software repositories.
 With a little bit of work, it is possible to use multiple versions of R on Linux; however, you will need an account with `sudo` privileges to set this up.
 
 The most common use case is to install R-devel alongside the current R version in order to facilitate package checking and testing prior to CRAN submission.
-Additionally, you may wish to maintain older versions of R to maintain compatibility and reproducibility of old code/scripts, or to rerun old analyses.
+Additionally, you may wish to maintain older versions of R to maintain compatibility and reproducibility of old projects, scripts, etc.
+
+This document provides an overview of several approaches to working with multiple versions of R on Linux - although these can also be used on macOS and Windows.
+
+\newpage
 
-# Using Docker
+# 1. Using Docker
 
 This is the preferred solution, as it is self-contained.
 Every time you start an instance of R in a Docker container, it starts in a "factory fresh" state, which means you will need to install additional packages into each new instance or save (commit) your changes for later reuse.
 The advantages here are that packages installed for each project are kept separate, ensuring proper reproducibility and portability.
 (Although [`renv`](https://rstudio.github.io/renv/) is also good for maintaining separate package libraries per project, it doesn't ensure system dependency separation.)
 
 Please note that this guide is intended only to get you started with using R with Docker, not to be a full tutorial for more advanced Docker use.
-See the official [Docker documentation](https://docs.docker.com/) and the [rocker project](https://github.com/rocker-org/rocker) for more advanced usage, including using containers with Rstudio Server installed[^1] or using a virtual X server.
+See the official [Docker documentation](https://docs.docker.com/) and the [rocker project](https://github.com/rocker-org/rocker) for more advanced usage, including using containers with RStudio Server installed[^1] or using a virtual X server.
 
-[^1]: https://github.com/rocker-org/rocker-versioned
+[^1]: <https://github.com/rocker-org/rocker-versioned2>
 
 ## Older R versions
 
@@ -52,44 +58,118 @@ See the official [Docker documentation](https://docs.docker.com/) and the [rocke
     \* NOTE: the `--rm` flag tells docker to remove the container after use.
     If you wish to keep the container for reuse later (*e.g.*, so you don't need to reinstall packages, etc.), then omit this.
 
-### Docker images wth Rstudio
+## R-devel
+
+1. Install Docker following the instructions [here](https://docs.docker.com/install/linux/docker-ce/ubuntu/).
+
+2. Download the latest R-devel image:
+
+    ```bash
+    docker pull rocker/drd
+    ```
+
+3. Start a container running R-devel:
+
+    ```bash
+    docker run -it --rm rocker/drd Rdevel
+    ```
+    
+    \* NOTE: the `--rm` flag tells docker to remove the container after use.
+    If you wish to keep the container for reuse later (*e.g.*, so you don't need to reinstall packages, etc.), then omit this.
+
+## Using RStudio
 
-Alternatively, there are docker images that include Rstudio server built-in, so you can run a container and connect to it through your browser instead of having to work in a terminal.
+Alternatively, there are docker images that include RStudio server built-in, so you can run a container and connect to it through your browser instead of having to work in a terminal.
 
-1. *E.g.*, start a versioned container running Rstudio that includes geospatial libraries:
+1. *E.g.*, start a versioned container running RStudio that includes geospatial libraries from the unstable ubuntugis repo:
 
     ```bash
-    docker run --rm -ti -p 8787:8080 rocker/geospatial:4.0.3
+    docker run --rm -ti -p 127.0.0.1:8080:8787 rocker/geospatial:4.2.3-ubuntugis
     ```
 
 2. Look for the user password to appear (in red) once it's done setting up the container.
 
-3. Open a web browser and go to `localhost:8080` to get to the Rstudio instance running in the docker container.
+3. Open a web browser and go to `localhost:8080` to get to the RStudio instance running in the docker container.
 
 4. Log in with username `rstudio` and the password from step 2.
 
-## R-devel
+\newpage
 
-1. Install Docker following the instructions [here](https://docs.docker.com/install/linux/docker-ce/ubuntu/).
-2. Download the latest R-devel image:
+# 2. Using `rig`
+
+Using the R installation manager `rig` (<https://github.com/r-lib/rig>) is the easiest approach, as it automates the installation an linking of different R versions, and is cross-platfrom (works with Windows, macOS, and Linux).
+Although each version of R maintains separate package libraries, users should strive to use per-project package libraries, e.g., using [`renv`](https://rstudio.github.io/renv/).
+
+1. Install `rig` using the platform-specific installation instructions are provide at <https://github.com/r-lib/rig#id-installation>;
+
+2. Install the R versions you need:
 
     ```bash
-    docker pull rocker/drd
+    rig add release
+    rig add devel
     ```
-3. Start a container running R-devel:
-
+  
     ```bash
-    docker run -it --rm rocker/drd Rdevel
+    rig add 4.1
+    rig add 4.2
+    rig add 4.3
     ```
+  
+    You can see the available versions (and the default, starred) by using `rig list`.
     
-    \* NOTE: the `--rm` flag tells docker to remove the container after use.
-    If you wish to keep the container for reuse later (*e.g.*, so you don't need to reinstall packages, etc.), then omit this.
+    Set the default version using the names from `rig list`:
+    
+    ```bash
+    rig default 4.3-arm64
+    ```
+
+3. Setup symlinks to allow easy launch of versioned R sessions:
+
+  ```bash
+  rig system make-links
+  ```
+  
+  Use e.g., `R-4.3-arm64` on macOS.
+
+4. On macOS, restrict access to system package directories:
+
+    ```bash
+    rig system fix-permissions
+    ```
+
+5. On macOS can also use the menu bar app (<https://github.com/r-lib/rig#id-macos-menu-bar-app>) to manage their R versions:
+
+    ```bash
+    open -a Rig
+    ```
+
+Be sure to open the Rig app preferences to allow launch at startup.
+
+## Switching R versions using `rig`
+
+### Setting the default version
+
+```bash
+## e.g., on macOS:
+rig default 4.3-arm64
+```
+
+Users on macOS can also use the menu bar app.
+
+### Selectively launching an RStudio seesion using specific version of R
+
+```bash
+## e.g., on macOS:
+rig rstudio 4.2-arm64
+```
+
+Users on macOS can also use the menu bar app.
 
 \newpage
 
-# Metal installation
+# 3. Manual build + installation
 
-As noted above, using Docker is likely a better solution, as this approach doesn't maintain separation between packages installed for different minor versions.
+As noted above, using Docker or `rig` are likely better solutions, as manual installations require manual upkeep and manual *uninstallation*.
 
 ## Older R versions
 
@@ -120,7 +200,7 @@ As noted above, using Docker is likely a better solution, as this approach doesn
     
     ```bash
     cd ~/R/src/
-    RVERSION=4.0.5
+    RVERSION=4.2.3
     wget https://cran.r-project.org/src/base/R-4/R-${RVERSION}.tar.gz
     tar -xvzf R-${RVERSION}.tar.gz
     ```
@@ -147,9 +227,11 @@ As noted above, using Docker is likely a better solution, as this approach doesn
     ```bash
     R-3.6.3
     # or
-    R-4.0.5
+    R-4.2.3
     ```
 
+\newpage
+
 ## R-devel
 
 1. Install prerequisites for building R from source:
@@ -264,20 +346,20 @@ As noted above, using Docker is likely a better solution, as this approach doesn
 
 \newpage
 
-## Switching R versions for use with RStudio
+## Manually switching R versions for use with RStudio
 
 RStudio will check for an environment variable, `RSTUDIO_WHICH_R`, which can be set to override using the installed system version of R.
 
-### Rstudio desktop
+### RStudio desktop
 
 ```bash
-export RSTUDIO_WHICH_R=/usr/local/bin/R-3.6.3
+export RSTUDIO_WHICH_R=/usr/local/bin/R-4.2.3
 rstudio &
 ```
 
-### Rstudio server
+### RStudio server
 
-As `root` user:
+As `sudo` user:
 
 ```bash
 umask 0022
@@ -289,7 +371,9 @@ echo "export RSTUDIO_WHICH_R=/usr/local/bin/R-3.6.3" > /etc/profile.d/rstudio.sh
 - <https://stackoverflow.com/a/24019938/1380598>
 - <https://stat.ethz.ch/pipermail/r-sig-debian/2012-August/001937.html>
 - <http://singmann.org/installing-r-devel-on-linux/>
+- <https://github.com/rocker-org/rocker-versioned2>
 - <https://hub.docker.com/r/rocker/drd/~/dockerfile/>
-- <https://support.rstudio.com/hc/en-us/articles/218004217-Building-R-from-source>
+- <https://support.posit.co/hc/en-us/articles/218004217-Building-R-from-source>
 - <https://community.rstudio.com/t/use-a-different-r-version-temporarily-in-rstudio/20848/8>
-- <https://solutions.rstudio.com/sys-admin/advanced-config/environment-variables/>
+- <https://solutions.posit.co/envs-pkgs/environment-variables/>
+- <https://github.com/r-lib/rig>
diff --git a/using-multiple-R-versions-on-linux.pdf b/using-multiple-R-versions-on-linux.pdf