❓
FAQ
No.
rcli makes use of the R binaries from rstudio/r-builds for the respective underlying distribution.No,
rcli assumes that all runtime dependencies are installed. The easiest way to do so is to install the respective distribution packages first (e.g. r-base-core on Ubuntu or R on Fedora) so that all dependencies are installed and then invoke rcli to install custom versions.The environment variable
PATH lists all directories in which your shell (e.g. zsh or bash) looks for executables. If some paths are missing, your shell will not be able to find (possibly existing) executables. There are many different standard locations where executables can be found and many more reasons why some of these paths are not set within your PATH env var.PATH is usually defined by your respective shell profile, e.g. .bashrc , .bash_profile , .zprofile and so on. If for some reason rcli reported that you're missing a certain entry in your PATH env var, you can add it in your respective profile by appending your missing path entry to the existing PATH as follows: PATH=$PATH:<the missing path> , e.g. PATH=$PATH:/usr/local/bin.radian is a Python library and if you're using Python3 on an arm64 machine, you will have the arm64 build installed. Hence, radian will only work with the R arm64 interpreters.To use
radian with an x86_64 installation of R, Python (and radian) for the x86_64 architecture would need to be installed.Yes,
x86_64 is supported by macOS via the "Rosetta 2" translation environment. By default, rcli will install the arm64 version of R (if possible) (R >= v4.1.0). Otherwise the x86_64 version will be installed.To force the installation of
x86_64 versions, pass the --arch x86_64 flag to rcli install.On a M1 machine one can install both the
arm64 and x86_64 versions of R (the latter supported via Rosetta). If you do so and plan to switch between both architectures, it is recommended not to use ccache to speed up source installations as the cache created by one of the respective R interpreters will also attempted to be used for the respective other architecture. This will not work and lead to loading failures during load-time, i.e. when calling library(<package>).The official macOS installation instructions from CRAN state that only one patch version per minor release can be installed. Is this also true for
rcli?This is one point where
rcli tries to help: rcli enables you to install and switch (between) any patch version of an R minor version (e.g. 4.1.1 and 4.1.2). This is done by applying some symlinking magic behind the scences and outsourcing the system lib of each patch version to /opt/R/<R version>.rcli installs the selected R version via the official CRAN installer and moves it Resources/ directory to /opt/R/<R version>. When switching, rcli modifies some symbolic links pointing to the "current" version of R specified at /Library/Frameworks/R.framework and links to static assets of each R version residing at /opt/R.Unfortunately some paths in the
.dylib files of R on macOS are hardcoded and point to /Library/Frameworks/R.framework resources. It is therefore not possible to have multiple "active" and "working" installations side-by-side. This is also stated and explained in the CRAN macOS manual. Every time an R version is switched, some symlink switching is happening which takes only a few seconds. However, due to the limitations mentioned above, one cannot run a different R version than the currently active one in a background job on macOS (this is possible on Windows and Linux).
Last modified 11mo ago