Best Practices for Managing R Libraries

R Software Modules

User are advised to utilize these modules when developing workflows on MSI systems

15:34:44 [vega0051@ahl03 ~ ]$ ml avail R/*rocky
------------------------------ /common/software/modulefiles/manual/common ------------------------------
R/4.2.0-openblas-rocky8  R/4.2.0-rocky8  R/4.3.0-openblas-rocky8  R/4.4.0-openblas-rocky8  R/4.4.2-openblas-rocky8

Workflows that utilize R>=4.4.2 can make use of the newly added r-rich software module collection. These include many of the dependencies that would otherwise need to be loaded in addition to R.

15:34:54 [vega0051@ahl03 ~ ]$ ml avail r-rich
------------------------------ /common/software/modulefiles/bundle/views/common ------------------------------  
r-rich/4.4.2_msi1.2  r-rich/4.6.0_msi1.2

Choosing R Library Installation Locations

When you load an R module on the MSI system, a base set of R libraries will be available to you from the MSI R module. However, you will likely need to install your own R libraries to accomplish your work. By default, MSI R modules will ask your permission to install libraries in your user home directory.

MSI recommends installing your R libraries for a project into a directory within that MSI project space rather than your MSI user home directory for the following reasons:

Maintaining your R libraries in a directory within a shared project space means that you can share those libraries with other members of your project team if desired.
You will be able to control your R library versions for that specific project.
Using your user home directory for R libraries across multiple projects may lead to unintentional software updates and conflicts.
You have more storage available to you in a project space than in your private user home directory.

Please note that R libraries are specific to the module version you have loaded on an MSI system; the default version of R (what is loaded with module load R) can and will change. It is best practice to always specify which version of R you are loading with module load so that you can ensure proper version control of your libraries.

Installing R Libraries

R libraries (also known as packages) can be installed either using command-line R or from within an RStudio interactive job in Open OnDemand (OOD). For challenging library installations, MSI recommends using command-line R rather than RStudio.

For a particular project, you should use the same version of R for your work with command-line R and RStudio. When installing R libraries via command-line R, you should request an interactive compute job first and then install your packages inside the compute job to avoid CPU timeouts.

Tip

There are common package install instructions on our R package reference page with the commond dependencies that need to be loaded for certain packages to successfully install

Example steps for installing R libraries via the command line:

Request an interactive job. See Interactive Computing with srun for more details.
```
srun -n 10 --mem 16G -t 1:00:00 -p interactive --pty bash
```
Load the version of R that you will be using for your project.
Make a new directory for your R libraries.
```
mkdir $SHARED/MyRlibs_$USER
```
Start R.
```
R
```
From within R, set your .libPaths() to your local installation for your project.
```
.libPaths("$SHARED/MyRlibs_$USER")
```

Install your package or packages from within R.

Example 1: CRAN package

install.packages("ggplot2")

Example 2: Bioconductor package

install.packages("BiocManager")
BiocManager::install("GenomicRanges")

Load your library using the library() function.
```
library("ggplot2")
```

Troubleshooting Challenging R Package Installations

Installation of some R packages may require loading system modules. These will need to be loaded when you install your R packages and whenever you want to use your R packages. For interactive OOD RStudio sessions, you can include module loads in the custom environment option box.

Common Warning Messages and Issues

You may see a warning, not an error, that a system folder located in /common/software/ is not writable. This is expected behavior: it means you have not defined a user library you have permission to modify. You can either follow the previous steps to do so, or follow the prompts provided by the warning to automatically create one in a standard location.

Another common warning may appear when attempting to install a package not located on CRAN, the default R package repository indexed by the install.packages() function. Even if the message states that the package is unavailable for this version of R, it may simply need to be pulled and installed from a different location. Package documentation should tell you whether packages need to be installed via Bioconductor or directly from a GitHub repository.