R Setup in VS Code for macOS/Win10

First, if you want to use RStudio, there doesn't seem to be anything wrong with that. However, if you already use VS Code for C/C++, Javascript, Python, Ruby, Markdown, text files, and everything else, it makes sense to also use it for R.

This should basically work for both Win10 and macOS. Your projects (stored in a version control system, of course) should be portable between the two. The use of renv keeps the package dependencies (and their versions) in sync for your project between multiple machines and users.

This also recommends (at least so far) the addition of plotly to enhance ggplot as it creates interactive plots that can be shared with people not running R. (Plotly also works well with Javascript and Python.)

Some references:

R for Data Science is a very good online book, especially if you are starting from zero in R.
This post and this post discussing some best practices for organizing your data, your projects, and your code.
The original reference used to put together these instructions, although the languageserversetup package did not seem to actually be necessary, so it is not used below.
A post specific to R and VS Code that was also a useful reference to putting together the instructions below. (Linked from the previous bullet, and probably ultimately more useful.)
This post discusses more about package management and why it is important. It goes briefly into even using Docker. However, it also discusses packrat which is out of date. packrat has been essentially replaced by renv. See also here.

Now the basic steps:

Install R

Download the installer package for R (here v4.0.0 is used) for your operating system (here either Win10 or macOS, but there's no reason Linux won't work), and install it.

On Windows, you will also need to install RTools because some R packages need to be natively compiled when you install them, and Windows does not readily offer a command-line compiler (e.g. clang, gcc) like macOS and Linux do.

When you install RTools in Windows, please don't use the default install location. It is no longer the 1980s, and you shouldn't clutter up your drive root. Just install it in Program Files\R (where R itself is installed) or something. Then add an environment variable:

RTOOLS40_HOME: C:\Program Files\R\rtools40

and add C:\Program Files\R\rtools40\usr\bin to your path.

Install Radian

For an improved command-line R experience, install Radian. This does require Python to run, but it improves the command-line experience (much like ipython does for Python), so it is worth it.

See this gist and this gist for Windows for general development system setup including Python (and terminal customization, etc.).

With Python and pip set up, run:

pip install --user radian

Then you should be able to run radian from a terminal prompt. If you cannot, make sure the pip install location is on your system path. For example, if you're on Windows and you used the dev setup instructions linked above that include pyenv, make sure that C:\Users\<username>\.pyenv\pyenv-win\bin is in your path.

Install the `languageserver` package in R Globally

Before setting up VS Code and its R-related extensions, you need to install the languageserver package in R. This provides the hooks that VS Code needs to to add help features, linting and code analysis, code execution hooks, etc. to the user interface.

From a terminal, run R (not Radian in this case if you installed Radian) to open the basic R shell. Then:

install.packages("languageserver")

and follow the instructions to select a CRAN mirror and all that. This should install languageserver and all of its dependencies.

Important: If you are on Win10, run R from an elevated command prompt. That is, run PowerShell as Administrator to do this step. This isn't clearly documented anywhere I could find, but without doing this as admin, the VS Code connection was never able to find the installed package.

Install VS Code and Extensions

Install VS Code in the usual way. Then install the vscode-R and vscode-r-lsp extensions.

In terms of extension configuration, in the R extension I have:

enabled "Bracketed Paste" (recommended when using Radian)
set the R path in macOS to /usr/bin/R
set the Windows path to C:\Users\<username>\.pyenv\pyenv-win\shims\radian.bat (unfortunately I couldn't get this to work using a more portable replacement like $HOME\.pyenv\...)
enabled the (currently experimental) Session Watcher which adds some nice things like being able to get plots and help right in the VS Code environment.

everything else is set to the default. In the LSP (language server protocol) extension, everything is set to the default.

Then reload your VS Code Window, and it should be all set.

Project Setup and Test

renv, gitignore, data folder, plotly, liveserver, etc...

Other Notes

Plot Updating

One thing I've noticed is that if you use ggplot and the plots show up within VS Code, the plot will not update if you "source" an entire script file (Cmd/Ctrl+Shift+S) but it will if you only run the code line that generates the plot (Cmd/Ctrl+Enter).

However, if you source a file so the source(...) command appears in the integrated R terminal, then go to the terminal, edit the command to include the parameter print.eval = TRUE, and execute, the plot will update.

gbingersoll/r_setup.md