Skip to content

R notes

Installing and updating

On Debian

To install R:

sudo apt-get install r-base r-base-dev

On MacOS

brew install --cask r-app # --cask isn't strictly necessary

Using formulae (like the r formula) from Homebrew is not recommended in most cases. See below.

Package management

Install packages using install.packages(). Whenever R has a new update in the distribution (4.4 to 4.5, for example), packages will generally need to be reinstalled also. The location they are installed to can vary and R may ask.

Often R complains about missing Debian packages (curl, ssl) and may fail if miniconda/anaconda is already installed (may want to change dir name).

On Debian packages are installed from a CRAN repository to a local library directory. The default users library must be created at ~/R/x86_64-pc-linux-gnu-library/{version#}, but packages can also be installed to /usr/local/lib/R/site-library if permissions allow.

On MacOS packages are installed to /Library/Frameworks/R.framework/Versions/<version num>/Resources/library

On a fresh install, I usually just start with installing tidyverse since it gets used so much, then pick and choose additional packages depending on use cases.

install.packages(c('tidyverse', 'xts', 'forecast')) # for example if I had some timeseries forecasting to do...

Sometimes the core R packages on Debian go out of date, usually after a new version of R is installed and need to be updated. To update all installed packages, start R with sudo and run:

update.packages(ask=False) # set ask to false if you have a big list of packages

Jupyter R notebooks

  • To run R in Jupyter notebooks install the IRKernel package using the appropriate MacOS or Linux method (in linux check for libzmq3-dev first).

  • The kernel "spec" must be installed or made available in your environment with IRkernel::installspec() into an environment that also has jupyter. Usually this is a conda environment or similar, unless you've installed jupyter globally. This involves starting the environment with jupyter

    conda activate jupyter-env-name

Then starting R and running IRkernel::installspec(). Do this for any environment you want to run jupyter and R in.

Compiling packages in MacOS/Homebrew

Its common for R to lose track of compilers or miscellaneous system files needed to build packages when using MacOS and Homebrew, especially when the source formula for R was installed. To fix compile errors you may be needed to point R to the Homebrew-installed compiler files as shown here. Installing the r-app cask, instead of the formula (just called r) should help since it uses already-compiled packages.

It might also be advisable to use the recommended R binary download method on https://r-project.org.