Skip to main content

English

Installing R packages​

Installing packages without internet

It's possible to install R packages without internet by downloading the source code of a package compressed in tar.gz, copying it to the machine via scp/other methods, and running (having previously loaded an R module):

$> R CMD INSTALL {package}.tar.gz

Requirements to install packages​

Requirements to install packages networked (must have Internet access)​

  • Connect to a login4 of MareNostrum5:

    mylaptop$> ssh {username}@glogin4.bsc.es
    mylaptop$> ssh {username}@alogin4.bsc.es
    IMPORTANT

    The MareNostrum5 login4 are restricted to BSC staff and are only accessible from the BSC internal network or the Virtual Private Network (VPN).

  • Check the Internet connectivity from glogin4/alogin4, for example:

    $> wget --tries=3 --timeout=5 -q --spider google.com && echo "Networked" || echo "Non-networked"
    Networked

Install and manage packages​

Install packages​

  • Install a package from CRAN:

    $> install.packages("SomePackage")

  • Install multiple packages at once:

    $> install.packages(c("A", "B", "C", "D"))

Update packages​

  • Update one or more specific packages:

    $> update.packages(oldPkgs = c("A", "B"))

  • Update all packages:

    $> update.packages()

  • Bioconductor

    $> BiocManager::install()

Remove packages​

  • Remove a package:

    $> remove.packages("SomePackage")

Working with environments​

It's possible to install packages outside of the main R stack, allowing for unprivileged installations:

  • Create a file called ~/.Renviron containing the following line:

    R_LIBS=/path/to/my/R/local/project

  • Install the desired package:

    R> install.packages("package_name")

This should print a message similar to:

Installing package into '/path/to/my/R/local/project'
(as 'lib' is unspecified)

  • You can then load the package as usual:

    R> library(package_name)

The directory specified by the R_LIBS variable can be anywhere, so it's possible to share packages between group members if it's located, for example, in a directory under /gpfs/projects.

Multiple version support​

  • Specific versions of a package can be installed using the devtools::install_version() function:

    devtools::install_version("package_name", version = "1.2.3")

Using this method, you can install and use multiple versions of the same package, which is useful in cases such as needing an older version of an already isntalled package. Packages installed under the directory defined by R_LIBS are priorized when loading packages with multiple versions, but different versions can be loaded as follows:

R> library(package_name, lib.loc="/path/to/package/location")

For example, the directory containing the packages installed under the default version of R in the Marenostrum5 GPP partition is:

R> library(package_name, lib.loc="/apps/GPP/R/4.3.2/INTEL/lib64/R/library")

Finally, if you need multiple versions of R, you can compile a version in one of your directories, or use conda for an easier installation.