--- title: "b2 - Setting up and using rix on macOS" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{b2-setting-up-and-using-rix-on-macos} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r, include = FALSE} knitr::opts_chunk$set( collapse = TRUE, comment = "#>" ) ``` ```{r, include=FALSE} library(rix) ``` *This vignette will discuss macOS-specific topics. If you're not using macOS, you can ignore this vignette, and read the `vignette("b1-setting-up-and-using-rix-on-linux-and-windows")` vignette instead.* ## Introduction When it comes to Nix, there are really only two supported operating systems: macOS and Linux distributions. Windows is "supported" because it is actually running Linux thanks to WSL2. In practice this means that Linux distributions and Windows can be considered one system, and macOS another, separate, system, with its own idiosyncrasies. This vignette details these. ## Installing Nix You can use `{rix}` to generate Nix expressions even if you don't have Nix installed on your system, but obviously, you need to install Nix if you actually want to build the defined development environment and use them. Installing (and uninstalling) Nix is quite simple, thanks to the installer from [Determinate Systems](https://determinate.systems/posts/determinate-nix-installer), a company that provides services and tools built on Nix. Simply open a terminal and run the following line: ```{sh parsermd-chunk-1, eval = FALSE} curl --proto '=https' --tlsv1.2 -sSf \ -L https://install.determinate.systems/nix | \ sh -s -- install ``` Once you have Nix installed, you can build the expressions you generate with `{rix}`! ## What if you don't have R already installed? If you have successfully installed Nix, but don't have yet R installed on your system, you could install R as you would usually do on your operating system, and then install the `{rix}` package, and from there, generated project-specific expressions and build them. But you could also install R using Nix. Running the following line in a terminal will drop you in an interactive R session that you can use to start generating expressions: ``` nix-shell --expr "$(curl -sl https://raw.githubusercontent.com/ropensci/rix/main/inst/extdata/default.nix)" ``` This should immediately start an R session inside your terminal. You can now run something like this: ``` rix(r_ver = "latest", r_pkgs = c("dplyr", "ggplot2"), system_pkgs = NULL, git_pkgs = NULL, ide = "other", project_path = ".", overwrite = TRUE) ``` to generate a `default.nix`, and then use that file to generate an environment with R, `{dplyr}` and `{ggplot2}`. If you need to add packages for your project, rerun the command above, but add the needed packages to `r_pkgs`. This is detailed in the vignettes `vignette("d1-installing-r-packages-in-a-nix-environment")` and `vignette("d2-installing-system-tools-and-texlive-packages-in-a-nix-environment")`. ## Generating expressions Once you have R installed, either through the usual installer for your operating system, or through Nix as explained previously, you can now start building project-specific development environments. On macOS, generating expressions works just like on Linux and Windows. Start an R session, and install `{rix}` if that's not already done. Because `{rix}` is not yet on CRAN, the easiest way is to install it from its r-universe: ```{r parsermd-chunk-2, eval = FALSE} install.packages("rix", repos = c( "https://ropensci.r-universe.dev", "https://cloud.r-project.org" )) ``` You can then use the `{rix}` package to generate expressions. Consult the next vignette `vignette("c-using-rix-to-build-project-specific-environments")` to learn more. ## More macOS specificities ### R Support for Apple Silicon in Nixpkgs In Nixpkgs, the M series processors of the ARM64 family, also known under AArch64, supports R since R version 3.5.3. Earlier versions of R will not compile on modern Apple processor architectures for the corresponding Nixpkgs revisions at that time. Hence, the `darwin-aarch64` platform has constrained backwards-compatibility. ### Shared libraries issue When using environments built with Nix on macOS, you might get crashes (segmentation faults) refering to "shared libraries". These indicate that your user library of R packages is interfering with the project-specific Nix environment. The system's user library that Nix packaged R by default includes appears in the search paths (check `libPaths()`). For macOS, the user library is at `/Library/Frameworks/R.framework/Versions/.-/Resources/library`; we have observed crashes with R packages that depend on system libraries, such as {data.table} or {dplyr}, and their (reverse) dependencies. Because user libraries from the system would also appear on the search path in R from `nixpkgs` for Linux, running `rix()` also runs `rix_init()` which creates a custom `.Rprofile` in your project's path. This custom `.Rprofile` ensure that only packages declaratively defined in your `default.nix` and built to be part of the Nix store (one R package is one separate derivation listed in a unique Nix path) appear on the R library path. ### RStudio and other development interfaces on macOS As of writing, RStudio cannot be installed through `nixpkgs` for macOS, and if you wish to use RStudio with a Nix environment, you have to install it through `nixpkgs`. This means that it is impossible to use RStudio and a Nix environment on macOS. When you try to generate an expression with `ide = "rstudio"` on macOS, this will raise a warning. Here are the options you have: - ignore the warning, because the environment will be built on a Linux distribution (even though you generated the expression on macOS) and used on a Linux distribution; - change the `ide =` argument to either `"other"` or `"code"`. Use `"code"` if you want to use VS Code and `"other"` for any other editor, like Vim or Emacs. These other editors don't need to be installed through `nixpkgs` to use Nix environments, unlike RStudio; - if you're working on a pipeline with the `{targets}` package, you could run it on Github Actions. This means you could work on the code on RStudio outside of the Nix environment, as the code will only be executed on Github Actions runners. See this vignette `vignette("z-advanced-topic-reproducible-analytical-pipelines-with-nix")` for further details; - work on your project as usual, using your usual installation of R and RStudio, but generate a `default.nix` at the end with `ide = "other"` with the right version of R for reproducibility purposes; - use *subshells* to execute only the code you need to run in a specific environment. See this vignette `vignette("z-advanced-topic-running-r-or-shell-code-in-nix-from-r")`; - help us package RStudio for macOS on `nixpgs`. See [here](https://github.com/NixOS/nixpkgs/blob/master/pkgs/applications/editors/rstudio/default.nix), the Nix expression for RStudio. We recommend you continue with the next vignette before tackling the more advanced topics listed above: `vignette("c-using-rix-to-build-project-specific-environments")`. ## Why do we need all these special tweaks? ### Path of Nix installation not in `PATH` When using RStudio Desktop on macOS, you typically launch it from the Applications folder. However, RStudio currently lacks an option to start an integrated R session via a shell startup on macOS (see this [issue](https://github.com/rstudio/rstudio/issues/13341). On RStudio for Linux flavors, `PATH` is properly inherited in R sessions. As a result, key environmental variables for UNIX systems, like PATH, are not be properly loaded from your default shell (e.g., `zsh`, via `~/.zshrc`). Also, RStudio overwrites a `PATH` variable set via `.Rprofile` or `.Renviron` with its own defaults, which makes it impossible to find Nix and standard Nix tools like `nix-build`. It's worth noting that this doesn't impact `rix::rix()`, which generates Nix expressions and doesn't require a Nix installation. As a workaround, we have added mechanisms in `nix_build()` and `with_nix()` that append the path of the Nix store to the `PATH` variable in an active session automatically (via `Sys.setenv()`, when you use RStudio on macOS. You don't have to do anything. We have you covered, and you get a friendly message that informs you. special side effect.