2.9 R Startup

Sys.getenv(x) get the values of the environment variables. Returns a vector of the same length as x.

  • x a character vector

Environment Variables examples:

> Sys.getenv(c("HOME", "R_HOME", "R_PAPERSIZE", "R_PRINTCMD"))
           HOME                                      R_HOME 
"/Users/menghan" "/Library/Frameworks/R.framework/Resources" 
    R_PAPERSIZE                                  R_PRINTCMD 
           "a4"                                       "lpr"

Rstudio doesn’t load Rprofile or Renviron

I store my Rprofile and Renviron in non-default places (i.e. ~/.config/R). When opening R in a normal shell, my environment is loaded perfectly fine. When opening Rstudio, it doesn’t load my options, settings or paths.

usethis is a workflow package: it automates repetitive tasks that arise during project setup and development, both for R packages and non-package projects.

2.9.1 .Rprofile

What is .Rprofile?

.Rprofile is a startup file to set options and environment variables. .Rprofile files can be either at the user or project level.

  • User-level .Rprofile files live in the base of the user’s home directory, and
  • project-level .Rprofile files live in the base of the project directory.

R will source only one .Rprofile file. If there is a project-level .Rprofile, the user-level file will NOT be sourced, i.e., the project-level config file takes priority.

So if you have both a project-specific .Rprofile file and a user .Rprofile file that you want to use, you explicitly source the user-level .Rprofile at the top of your project-level .Rprofile with source("~/.Rprofile").

.Rprofile files are sourced as regular R code, so setting environment variables must be done inside a Sys.setenv(key = "value") call.

Here is an example project-level .Rprofile file that

  • sources the user-level .Rprofile file,
  • load some commonly used packages, and
  • optionally, sets some options and environment variables specific to the project.
cat("This is the local user .Rprofile file\n")
# Source the global user .Rprofile
source("~/.Rprofile")

# Add project-specific configurations below
library(knitr)
library(tidyverse)

This can be convenient for loading commonly used packages in some documentation projects for yourself.

  • Note that it is recommended to load packages individually and explicitly if you are working on a research project, where reproducibility is important and in case of collaboration.

Change default settings globally for all R sessions

Quitting R will erase the default theme setting. If you load ggplot2 in a future session it will revert to the default gray theme. If you’d like for ggplot2 to always use a different theme (either yours or one of the built-in ones), you can set a load hook and put it in your .Rprofile file. For example, the following hook sets the default theme to be theme_minimal() every time the ggplot2 package is loaded.

setHook(packageEvent("ggplot2", "onLoad"), 
        function(...) ggplot2::theme_set(ggplot2::theme_bw()))

Of course, you can always override this default theme by adding a theme object to any of your plots that you construct in ggplot2.

Set stargazer default output type to identify output type automatically

stargazer use type="latex" by default, which will throw an error when you render your document to HTML. But if you specify stargazer(..., type = "html"), it will have problems when you render your document to PDF. So it is better to set the output type automatically based on the output format of the document.

Set type = ifelse(knitr::is_html_output(), "html", "latex") by default.

# Set stargazer to automatically detect output type (html or latex) 
# based on the output format of the document
setHook(
  packageEvent("stargazer", "attach"),
  function(...) {
    assign(
      "stargazer",
      function(...,
               type = if (knitr::is_html_output()) "html" else "latex") {
        stargazer::stargazer(..., type = type)
      },
      envir = .GlobalEnv
    )
  }
)

An easy workaround is to set type="text", which will work for both HTML and PDF outputs, but the table will not be as pretty as the HTML or LaTeX output. text output is a simple ASCII table (monospaced) that is suitable for console output. Good readability but moderate aesthetics. type="text" is a good option for quick previews, but NOT suitable for publication.

2.9.2 .Renviron

.Renviron is a user-controllable file that can be used to create environment variables. This is especially useful to avoid including credentials like API keys inside R scripts. This file is written in a key-value format, so environment variables are created in the format:

Key1=value1
Key2=value2
...additional key=value pairs

And then Sys.getenv("Key1") will return "value1" in an R session.

Like with the .Rprofile file, .Renviron files can be at either the user or project level. If there is a project-level .Renviron, the user-level file will not be sourced. The usethis package includes a helper function for editing .Renviron files from an R session with usethis::edit_r_environ().

The .Renviron file is most useful for defining sensitive information such as API keys (such as GitHub, Twitter, or Posit Connect) as well as R specific environment variables like the history size (R_HISTSIZE=100000) and default library locations R_LIBS_USER.

Rcpp compilation breaks in R 4.1.0

devtools::build("my_package")
Error: Could not find tools necessary to compile a package
Call `pkgbuild::check_build_tools(debug = TRUE)` to diagnose the problem.

  • In RStudio, I am continually prompted to install additional build tools and I can’t install the build tool. \(\rightarrow\) Bypass the option options(buildtools.check = function(action) TRUE).

  • Turns out R was pointing to an old clang version in my Makevars.

    I just deleted it using in Terminal

    sudo rm ~/.R/Makevars

Install SDK command line tool

Download from developer.apple.com. Software development kit.

https://developer.apple.com/download/all/

R compiler tools for cpp on MacOS