rix: Reproducible Environments with Nix

• Introduction
• Quick start for returning users
• Getting started for new users
  • Docker
• Why Nix? Comparison with Docker+renv/Conda/Guix
• Contributing
• Thanks
• Recommended reading

Introduction

{rix} is an R package that leverages Nix, a package manager focused on reproducible builds. With Nix, you can create project-specific environments with a custom version of R, its packages, and all system dependencies (e.g., GDAL). Nix ensures full reproducibility, which is crucial for research and development projects.

Use cases include running web apps (e.g., Shiny, {plumber} APIs) or {targets} pipelines with a controlled R environment. Unlike {renv}, which snapshots package versions, {rix} provides an entire ecosystem snapshot, including system-level dependencies.

While Nix has a steep learning curve, {rix}

1. simplifies creating Nix expressions, which define reproducible environments, also from renv.lock files;
2. lets you work interactively in IDEs like RStudio or VS Code, or use Nix in CI/CD workflows;
3. makes it easy to create Docker images with the right packages;
4. provides helpers that make it easy to build those environments, evaluate the same code in different development environments, and finally to deploy software environments in production.

If you want to watch a 5-Minute video introduction, click the image below:

Nix includes nearly all CRAN and Bioconductor packages, with the ability to install specific package versions or GitHub snapshots. Nix also includes Python, Julia (and many of their respective packages) as well as many, many other tools (up to 120’000 pieces of software as of writing). Expressions generated by {rix} point to our fork of Nixpkgs which provides improved compatibility for older versions of R and R packages, especially for Apple Silicon computers.

If you have R installed, you can start straight away from your R session by first installing {rix} from CRAN:

install.packages("rix")

or the development version from R-universe (development versions will generally have fixes and newer features):

install.packages("rix", repos = c(
  "https://ropensci.r-universe.dev"
))

Now try to generate an expression using rix():

library("rix")

# Choose the path to your project
# This will create two files: .Rprofile and default.nix
path_default_nix <- "."

rix(
  r_ver = "4.3.3",
  r_pkgs = c("dplyr", "ggplot2"),
  system_pkgs = NULL,
  git_pkgs = NULL,
  ide = "code",
  project_path = path_default_nix,
  overwrite = TRUE,
  print = TRUE
)

This will generate two files, default.nix and .Rprofile in project_default_nix. default.nix is the environment definition written in the Nix programming language, and .Rprofile prevents conflicts with library paths from system-installed R versions, offering better control over your environment and improving isolation of Nix environments. .Rprofile is created by rix_init() which is called automatically by the main function, rix().

It is also possible to provide a date instead of an R version:

# Choose the path to your project
# This will create two files: .Rprofile and default.nix
path_default_nix <- "."

rix(
  date = "2024-12-14",
  r_pkgs = c("dplyr", "ggplot2"),
  system_pkgs = NULL,
  git_pkgs = NULL,
  ide = "code",
  project_path = path_default_nix,
  overwrite = TRUE,
  print = TRUE
)

It is also possible to add Python (and Python packages) and/or Julia (and Julia packages) to an environment, by passing a list of two elements to the py_conf argument of rix() (or jl_conf for Julia). This list needs to first specify a Python or Julia version, and then an atomic character vector of Python (Julia) packages:

rix(
  date = "2025-09-04",
  r_pkgs = "ggplot2",
  py_conf = list(
    py_version = "3.12",
    py_pkgs = c("polars", "great-tables")
  ),
  jl_conf = list(
    jl_version = "1.10",
    jl_pkgs = c("Arrow", "TidierData")
  ),
  project_path = path_default_nix,
  overwrite = TRUE
)

It should be noted that while we offer the guarantee that most CRAN and Bioconductor packages are supported and installable through Nix (less than 5% of packages aren’t currently supported), the same cannot be said about Python or Julia packages. Should you require a Python or Julia but cannot successfully bulid the environment, do open an issue and we’ll see what we can do.

The table below illustrates this all the different types of environment you can generate:

r_ver or date	Intended use	State of R version	State of CRAN packages	State of Bioconductor packages	State of other packages in Nixpkgs
r_ver = "latest-upstream"	Start of new project where versions don’t matter	Current or previous	Outdated (up to 6 months)	Outdated (up to 6 months)	Current at time of generation
r_ver = "4.4.2" (or other)	Reproducing old project or starting a new project where versions don’t matter	Same as in `r_ver`, check `available_r()`	Outdated (up to 2 months if using latest release)	Outdated (up to 2 months if using latest release)	Potentially outdated (up to 12 months)
date = "2024-12-14"	Reproducing old project or starting a new project using the most recent date	Current at that date, check `available_dates()`	Current at that date, check `available_dates()`	Current at that date, check `available_dates()`	Potentially outdated (up to 12 months)
r_ver = "bleeding-edge"	To develop against the latest release of CRAN	Always current	Always current	Always current	Always current
r_ver = "frozen-edge"	To develop against the latest release of CRAN, but manually manage updates	Current at time of generation	Current at time of generation	Current at time of generation	Current at time of generation
r_ver = "r-devel"	To develop/test against the development version of R	Development version	Always current	Always current	Always current
r_ver = "r-devel-bioc-devel"	To develop/test against the development version of R and Bioconductor	Development version	Always current	Development version	Always current
r_ver = "bioc-devel"	To develop/test against the development version of Bioconductor	Always current	Always current	Development version	Always current

If you want to benefit from relatively fresh packages and have a stable environment for production purposes, using a date for r_ver is likely your best option.

Quick-start for returning users

Click to expand

If you’re already familiar with Nix and {rix}, install Nix using the Determinate Systems installer: (if you’re using WSL, do check out the detailed installation instructions though):

curl --proto '=https' --tlsv1.2 -sSf -L https://install.determinate.systems/nix | sh -s -- install

Then, configure the rstats-on-nix binary cache to speed up builds. The recommended approach is to use rix::setup_cachix() from R:

nix-shell -p R rPackages.rix  # Start R with rix (may take a few minutes on first use)

rix::setup_cachix()

Then add yourself as a trusted user:

# Linux:
echo "trusted-users = root $USER" | sudo tee -a /etc/nix/nix.custom.conf && sudo systemctl restart nix-daemon

# macOS:
echo "trusted-users = root $USER" | sudo tee -a /etc/nix/nix.custom.conf && sudo launchctl kickstart -k system/org.nixos.nix-daemon

Many thanks to Cachix for sponsoring the rstats-on-nix cache! See vignette("setting-up-linux-windows") or vignette("setting-up-macos") for alternative approaches and NixOS instructions.

You can then use {rix} to build and enter a Nix-based R environment:

library(rix)

path_default_nix <- "."

rix(
  r_ver = "4.3.3",
  r_pkgs = c("dplyr", "ggplot2"),
  system_pkgs = NULL,
  git_pkgs = NULL,
  ide = "code",
  project_path = path_default_nix,
  overwrite = TRUE,
  print = TRUE
)

To build the environment, call nix_build()

# nix_build() is a wrapper around the command line tool `nix-build`
nix_build(project_path = ".")

If you don’t have R installed, but have the Nix package manager installed, you can run a temporary terminal session which includes R and the released version of {rix}:

nix-shell -p R rPackages.rix

or if you prefer the development version of {rix}:

nix-shell --expr "$(curl -sl https://raw.githubusercontent.com/ropensci/rix/main/inst/extdata/default.nix)"

You can then create new development environment definitions, build them, and start using them.

Getting started for new users

New to {rix} and Nix? Start by reading the vignette("getting-started") (online documentation). to learn how to set up and use Nix smoothly.

Docker

Try Nix inside Docker by following this vignette("nix-inside-docker") vignette.

How is Nix different from Docker+renv/{groundhog}/{rang}/(Ana/Mini)Conda/Guix? or Why Nix?

Docker + {renv}

Docker and {renv} provide robust reproducibility by combining package snapshots with system-level dependencies. However, for long-term reproducibility, Nix offers a simpler approach by bundling everything (R, packages, and dependencies) in a single environment.

Ana/Miniconda & Mamba

Conda is similar to Nix, but Nix offers immutable environments, making it more reliable for preventing accidental changes. Nix also supports nearly all CRAN and Bioconductor packages, which Conda lacks.

Nix vs. Guix

Guix, like Nix, focuses on reproducibility, but Nix supports more CRAN/Bioconductor packages and works across Windows, macOS, and Linux.

Is {rix} all there is?

No, there are other tools that you might want to check out, especially if you want to set up polyglot environments (even though it is possible to use {rix} to set up an environment with R and Python packages for example).

Take a look at https://devenv.sh/ and https://prefix.dev/ if you want to explore other tools that make using Nix easier!

What’s the recommended workflow?

Ideally, you shouldn’t be using a system-wide installation of R, and instead use dedicated Nix environments for each of your projects.

Start a new project by writing a file called generate_env.R and write something like:

library(rix)

path_default_nix <- "."

rix(
  r_ver = "4.3.3",                # Change to whatever R version you need
  r_pkgs = c("dplyr", "ggplot2"),  # Change to whatever packages you need
  system_pkgs = NULL,
  git_pkgs = NULL,
  ide = "code",
  project_path = path_default_nix,
  overwrite = TRUE,
  print = TRUE
)

Then use the following command to bootstrap an enivronment with R and {rix} only (from the same directory):

nix-shell --expr "$(curl -sl https://raw.githubusercontent.com/ropensci/rix/main/inst/extdata/default.nix)"

and then simply run Rscript generate_env.R which will run the above script, thus generating the project’s default.nix. If you need to add packages, open the generate_env.R file again, modify it, and run it again, do not edit the default.nix directly. Also, commit all the files to version control to avoid any issues.

Reproducible pipelines using rixpress

{rix} is getting a little sister package, {rixpress}, which is heavily inspired by {targets}. If you want to build polyglot data science pipelines with the power of Nix, give {rixpress} a try!

Contributing

Refer to Contributing.md to learn how to contribute to the package.

Please note that this package is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

LLM Context File

The file rix.pkgctx.yaml contains an LLM-optimised API specification for {rix}, generated by pkgctx. You can paste this file directly into an LLM conversation to give it comprehensive knowledge of {rix}’s functions, their signatures, arguments, and usage examples. This enables LLMs to write correct {rix} code without hallucinating function names or arguments. The file is automatically regenerated on each push to stay in sync with the package API. To use it, simply copy the contents into your chat with Claude, ChatGPT, or any other LLM when asking it to help you write {rix} code.

If you want to use pkgctx for your own package, just re-use the pkgctx.yaml action file. You can also use pkgctx to generate context files for any package available on CRAN, PyPi and GitHub.

Thanks

Thanks to the Nix community for making Nix possible, and thanks to the community of R users on Nix for their work packaging R and CRAN/Bioconductor packages for Nix (in particular Justin Bedő, Rémi Nicole, nviets, Chris Hammill, László Kupcsik, Simon Lackerbauer, MrTarantoga and every other person from the Matrix Nixpkgs R channel: https://matrix.​to/#/%23r:nixos.org.

Finally, thanks to David Solito for creating {rix}’s logo!

Recommended reading

• NixOS’s website
• Nixpkgs’s GitHub repository
• Nix for R series from Bruno’s blog. Or, in case you like video tutorials, watch this one on Reproducible R development environments with Nix
• nix.dev tutorials
• Nix pills
• Nix for Data Science
• NixOS explained: NixOS is an entire Linux distribution that uses Nix as its package manager.
• Blog post: Nix with R and devtools
• Blog post: Statistical Rethinking and Nix
• Blog post: Searching and installing old versions of Nix packages