d1 - Installing R and Python packages in a Nix environment

In rix: Reproducible Data Science Environments with 'Nix'

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(rix)

Introduction

You now know how to declare and build reproducible development environments using {rix} and Nix. This vignette will explain how to install specific versions of CRAN packages and how to install packages from GitHub.

A word of caution

It is important at this stage to understand that you should not call install.packages() from a running Nix environment. Doing so will raise an error to avoid issues. If you want to add packages while analyzing data, you need to add it the default.nix expression and rebuild the environment. The same goes for installing packages from GitHub; use the method described in this vignette instead of using something like remotes::install_github().

We recommend you create a script called create_env.R or similar, and add the call to rix() there:

library(rix)

rix(r_ver = "4.4.0",
    r_pkgs = c("dplyr", "ggplot2"),
    system_pkgs = NULL,
    git_pkgs = NULL,
    ide = "code",
    project_path = path_default_nix,
    overwrite = TRUE,
    print = TRUE)

Then, add the packages you need to r_pkgs and run the script again. Then, build the environment using nix-build again, and drop into it using nix-shell. Calling install.packages() is a bad idea for several reasons:

• it goes against the idea of defining an environment in a declarative way. If you were able to add packages using install.packages(), your environment would end up in a state where the default.nix definition of the environment and the actual environment don't match anymore.
• using install.packages() would likely simply not work, and if it would work, it would cause issues. For example, if you call install.packages("ggplot2") from one Nix shell, it will not install {ggplot2} "inside" the Nix shell, but will install it on your user's system library of packages. This is because the Nix shell cannot be changed at run-time, and so, R will instead install the packages in the user's library. This version of {ggplot2}, because it is in that system-wide library of packages, will be available to any other Nix shells. If you call install.packages("ggplot2") again from another Nix shell, say 6 months later, this will replace the first version of {ggplot2} with the latest version.

Ideally, you should only manage R versions and R packages using Nix, and uninstall any system-managed version of R and R packages. But if you do wish to keep a system-managed version of R and R packages, rix::rix() also runs rix::rix_init() automatically which generates an .Rprofile file that avoids any clashes between your global user library and Nix-managed libraries of R packages.

Installing old packages archived on CRAN

It is possible to install an arbitrary version of a package that has been archived on CRAN:

path_default_nix <- tempdir()

rix(
  r_ver = "4.2.1",
  r_pkgs = c("dplyr@0.8.0", "janitor@1.0.0"),
  system_pkgs = NULL,
  git_pkgs = NULL,
  ide = "none",
  project_path = path_default_nix,
  overwrite = TRUE
)

#> let
#>  pkgs = import (fetchTarball "https://github.com/NixOS/nixpkgs/archive/79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a.tar.gz") {};
#>
#>   git_archive_pkgs = [
#>     (pkgs.rPackages.buildRPackage {
#>       name = "dplyr";
#>       src = pkgs.fetchzip {
#>        url = "https://cran.r-project.org/src/contrib/Archive/dplyr/dplyr_0.8.0.tar.gz";
#>        sha256 = "sha256-f30raalLd9KoZKZSxeTN71PG6BczXRIiP6g7EZeH09U=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages)
#>           assertthat
#>           glue
#>           magrittr
#>           pkgconfig
#>           R6
#>           Rcpp
#>           rlang
#>           tibble
#>           tidyselect
#>           BH
#>           plogr;
#>       };
#>     })
#>
#>     (pkgs.rPackages.buildRPackage {
#>       name = "janitor";
#>       src = pkgs.fetchzip {
#>        url = "https://cran.r-project.org/src/contrib/Archive/janitor/janitor_1.0.0.tar.gz";
#>        sha256 = "sha256-3NJomE/CNbOZ+ohuVZJWe2n1RVGUm8x8a0A0qzLmdN4=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages)
#>           dplyr
#>           tidyr
#>           snakecase
#>           magrittr
#>           purrr
#>           rlang;
#>       };
#>     })
#>   ];
#>
#>   system_packages = builtins.attrValues {
#>     inherit (pkgs)
#>       R
#>       glibcLocales
#>       nix;
#>   };
#>
#> in
#>
#> pkgs.mkShell {
#>   LOCALE_ARCHIVE = if pkgs.system == "x86_64-linux" then  "${pkgs.glibcLocales}/lib/locale/locale-archive" else "";
#>   LANG = "en_US.UTF-8";
#>    LC_ALL = "en_US.UTF-8";
#>    LC_TIME = "en_US.UTF-8";
#>    LC_MONETARY = "en_US.UTF-8";
#>    LC_PAPER = "en_US.UTF-8";
#>    LC_MEASUREMENT = "en_US.UTF-8";
#>
#>   buildInputs = [ git_archive_pkgs   system_packages   ];
#>
#> }

The above expression will install R version 4.2.1, and {dplyr} at version 0.8.0 and {janitor} at version 1.0.0. This can be useful, especially for packages that have been archived, but otherwise, this feature should ideally be used sparingly, for two reasons. First, if you want to reconstruct an environment as it was around 2019, use the version of R that was current at that time using the date argument in rix(). This will ensure that every package that gets installed is at a version compatible with that version of R, which might not be the case if you need to install a very old version of one particular package. Second, doing so will install the package from source. For packages that don't require compilation, this should be fine, but packages that require compilation will likely fail to compile successfully. We are working on handling this better for future versions of {rix}.

Installing packages from GitHub

It is also possible to install packages from GitHub:

path_default_nix <- tempdir()

rix(
  r_ver = "4.2.1",
  r_pkgs = c("dplyr", "janitor"),
  git_pkgs = list(
    list(
      package_name = "housing",
      repo_url = "https://github.com/rap4all/housing/",
      commit = "1c860959310b80e67c41f7bbdc3e84cef00df18e"
    ),
    list(
      package_name = "fusen",
      repo_url = "https://github.com/ThinkR-open/fusen",
      commit = "d617172447d2947efb20ad6a4463742b8a5d79dc"
    )
  ),
  ide = "none",
  project_path = path_default_nix,
  overwrite = TRUE
)

#> # This file was generated by the {rix} R package v0.7.1 on 2024-07-01
#> # with following call:
#> # >rix(r_ver = "79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a",
#> #  > r_pkgs = c("dplyr",
#> #  > "janitor"),
#> #  > git_pkgs = list(list(package_name = "housing",
#> #  > repo_url = "https://github.com/rap4all/housing/",
#> #  > commit = "1c860959310b80e67c41f7bbdc3e84cef00df18e"),
#> #  > list(package_name = "fusen",
#> #  > repo_url = "https://github.com/ThinkR-open/fusen",
#> #  > commit = "d617172447d2947efb20ad6a4463742b8a5d79dc")),
#> #  > ide = "none",
#> #  > project_path = path_default_nix,
#> #  > overwrite = TRUE)
#> # It uses nixpkgs' revision 79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a for reproducibility purposes
#> # which will install R version 4.2.1.
#> # Report any issues to https://github.com/ropensci/rix
#> let
#>  pkgs = import (fetchTarball "https://github.com/NixOS/nixpkgs/archive/79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a.tar.gz") {};
#>
#>   rpkgs = builtins.attrValues {
#>     inherit (pkgs.rPackages)
#>       dplyr
#>       janitor;
#>   };
#>
#>   git_archive_pkgs = [
#>     (pkgs.rPackages.buildRPackage {
#>       name = "housing";
#>       src = pkgs.fetchgit {
#>         url = "https://github.com/rap4all/housing/";
#>         rev = "1c860959310b80e67c41f7bbdc3e84cef00df18e";
#>         sha256 = "sha256-s4KGtfKQ7hL0sfDhGb4BpBpspfefBN6hf+XlslqyEn4=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages)
#>           dplyr
#>           ggplot2
#>           janitor
#>           purrr
#>           readxl
#>           rlang
#>           rvest
#>           stringr
#>           tidyr;
#>       };
#>     })
#>
#>
#>     (pkgs.rPackages.buildRPackage {
#>       name = "fusen";
#>       src = pkgs.fetchgit {
#>         url = "https://github.com/ThinkR-open/fusen";
#>         rev = "d617172447d2947efb20ad6a4463742b8a5d79dc";
#>         sha256 = "sha256-TOHA1ymLUSgZMYIA1a2yvuv0799svaDOl3zOhNRxcmw=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages)
#>           attachment
#>           cli
#>           desc
#>           devtools
#>           glue
#>           here
#>           magrittr
#>           parsermd
#>           roxygen2
#>           stringi
#>           tibble
#>           tidyr
#>           usethis
#>           yaml;
#>       };
#>     })
#>    ];
#>
#>   system_packages = builtins.attrValues {
#>     inherit (pkgs)
#>       R
#>       glibcLocales
#>       nix;
#>   };
#>
#> in
#>
#> pkgs.mkShell {
#>   LOCALE_ARCHIVE = if pkgs.system == "x86_64-linux" then  "${pkgs.glibcLocales}/lib/locale/locale-archive" else "";
#>   LANG = "en_US.UTF-8";
#>    LC_ALL = "en_US.UTF-8";
#>    LC_TIME = "en_US.UTF-8";
#>    LC_MONETARY = "en_US.UTF-8";
#>    LC_PAPER = "en_US.UTF-8";
#>    LC_MEASUREMENT = "en_US.UTF-8";
#>
#>   buildInputs = [ git_archive_pkgs rpkgs  system_packages   ];
#>
#> }

This will install two packages from GitHub: the {housing} package and more specifically the code as it is in the fusen branch; however the branch name is not required, as the commit is enough to pin the exact version of the code needed. The {fusen} package is also installed, as of commit d617172447d.

If you want to install a package from GitHub, which store the R package in a subfolder, you should specify the subfolder in the repo_url argument. For example, if you want to install the R version of the package {BPCells}:

path_default_nix <- tempdir()

rix(
  r_ver = "4.5.0",
  r_pkgs = c("dplyr"),
  system_pkgs = NULL,
  git_pkgs = list(
    package_name = "BPCells",
    repo_url = "https://github.com/bnprks/BPCells/r",
    commit = "16faeade0a26b392637217b0caf5d7017c5bdf9b"
  ),
  ide = "none",
  project_path = ".",
  overwrite = TRUE,
  print = TRUE
)

Installing local archives

It is also possible to install packages from a local tar.gz file. For this, place the package in the same folder where the default.nix will be generated, and write something like this:

rix(
  r_ver = "4.3.1",
  local_r_pkgs = c("chronicler_0.2.1.tar.gz", "knitr_1.43.tar.gz"),
  overwrite = TRUE
)

This assumes that both chronicler_0.2.1.tar.gz and knitr_1.43.tar.gz have been downloaded beforehand. If you want to include a local package that you are developing, make sure to first build it using devtools::build() to build the .tar.gz archive, but if you can, consider uploading the source code to your package on GitHub and installing it from GitHub instead.

Installing Python packages

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

rix(
  date = "2025-02-17",
  r_pkgs = "ggplot2",
  py_conf = list(
    py_version = "3.12", 
    py_pkgs = c("polars", "great-tables")
  ),
  overwrite = TRUE
)

This will install Python 3.12, the polars and great-tables packages, but also ipykernel and pip to ensure the Python interpreter works correctly with IDEs such as Positron.

Installing Python packages not available via nixpkgs (impure)

Not all Python packages can be installed through Nix; unlike CRAN, Pypi doesn’t get automatically mirrored and individual packages fixed by volunteers. Instead, specific Python packages get packaged individually for Nix. Thus, it could very well be the case that a specific Python package (or version of a Python package) that you need for a project is not available via nixpkgs.

In this case, it is still possible to use Python-specific package managers, like uv, to install packages. This is also useful if you work on a project with colleagues that use uv and that don’t want (yet) to use Nix. uv, is 10-100x faster than pip and also generates a lock file for improved reproducibility.

The idea is to install uv in your shell (but not any Python nor Python packages):

rix(
  ...
  system_pkgs = c("uv"),
  ...
)

And then use uv from your shell as you would usually. We recommend specifying Python packages in a requirements.txt file, and specifying explicit versions (e.g., scanpy==1.11.4). Finally, we also recommend setting a shell hook to set up the virtual environment and install the packages from the requirements.txt when entering the shell (mind the quotes):

rix(
  ...
  system_pkgs = c("uv"),
  shell_hook = "
      if [ ! -f pyproject.toml ]; then
        uv init --python 3.13.5 # or whichever Python version you need
      fi
        uv add --requirements requirements.txt
      # Create alias so python uses uvs environment
      alias python='uv run python'
  ",
)

After running nix-shell, uv should initalize a Python project with the specified Python version and install the packages listed in requirements.txt within the nix environment. This will take place each time nix-shell is called, however this will be cached and not installed each time.

To make sure everything works fine, you could simply start a Python interpreter and try to load numpy. This should work fine, but if it doesn’t, the following error could be raised:

ImportError:

IMPORTANT: PLEASE READ THIS FOR ADVICE ON HOW TO SOLVE THIS ISSUE!

Importing the numpy C-extensions failed. This error can happen for
many reasons, often due to issues with your setup or how NumPy was
installed.

We have compiled some common reasons and troubleshooting tips at:

    https://numpy.org/devdocs/user/troubleshooting-importerror.html

Please note and check the following:

  * The Python version is: Python3.13 from "/home/user/projects/rix_uv/.venv/bin/python3"
  * The NumPy version is: "2.2.6"

and make sure that they are the versions you expect. Please carefully study the
documentation linked above for further help.

Original error was: libstdc++.so.6: cannot open shared object file: No such file or directory

That is an issue when using wheels (wheels are binaries of Python packages that get installed by default using uv). These wheels expect certain libraries to be in certain places. One way to solve this is to add the following to your shell hook:

shellHook = ''
   # Export LD_LIBRARY is required for python packages that dynamically load libraries, such as numpy 
   export LD_LIBRARY_PATH="${pkgs.lib.makeLibraryPath (with pkgs; [ zlib gcc.cc glibc stdenv.cc.cc ])}":LD_LIBRARY_PATH;
   ...

Your environment should now work.

If this seems complicated: yes, and that is actually exactly the type of problems that Nix aims to solve. However, there are just too many Python packages to automate their inclusion into nixpkgs like how it’s done for R. If you can, prefer using the Python packages included in nixpkgs.

Installing Julia packages

Unlike with R packages, date‑specific snapshots of the Julia registry only become available starting on the 2025-09-04. If you choose an earlier date, Nixpkgs will fall back to whatever registry snapshot was bundled at that time—which will be an older version, and it may not install successfully. So if you need Julia, use a date after the 2025-09-04.

To add Julia packages, you can use the jl_conf argument of rix(). This list needs to first specify a Julia version, and then an atomic character vector of packages:

rix(
  date = "2025-09-04",
  r_pkgs = "ggplot2",
  jl_conf = list(
    jl_version = "1.11",
    jl_pkgs = c("TidierData", "GLM")
  ),
  project_path = ".",
  overwrite = TRUE
)

This will install Julia 1.11, the TidierData and GLM packages. You can also install the long-term support version of Julia by using "lts" as the jl_version.

Converting from an renv.lock file

{rix} also includes an renv2nix() function that converts an renv.lock file into a valid Nix expression. Read the vignette vignette("f-renv2nix") to learn more.

A complete example

This example shows how to install packages from CRAN, from the CRAN archives and from GitHub:

path_default_nix <- tempdir()

rix(
  r_ver = "4.2.1",
  r_pkgs = c("dplyr", "janitor", "AER@1.2-8"),
  git_pkgs = list(
    list(
      package_name = "housing",
      repo_url = "https://github.com/rap4all/housing/",
      commit = "1c860959310b80e67c41f7bbdc3e84cef00df18e"
    ),
    list(
      package_name = "fusen",
      repo_url = "https://github.com/ThinkR-open/fusen",
      commit = "d617172447d2947efb20ad6a4463742b8a5d79dc"
    )
  ),
  ide = "none",
  project_path = path_default_nix,
  overwrite = TRUE
)

#> # This file was generated by the {rix} R package v0.7.1 on 2024-07-01
#> # with following call:
#> # >rix(r_ver = "79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a",
#> #  > r_pkgs = c("dplyr",
#> #  > "janitor",
#> #  > "AER@1.2-8"),
#> #  > git_pkgs = list(list(package_name = "housing",
#> #  > repo_url = "https://github.com/rap4all/housing/",
#> #  > commit = "1c860959310b80e67c41f7bbdc3e84cef00df18e"),
#> #  > list(package_name = "fusen",
#> #  > repo_url = "https://github.com/ThinkR-open/fusen",
#> #  > commit = "d617172447d2947efb20ad6a4463742b8a5d79dc")),
#> #  > ide = "none",
#> #  > project_path = path_default_nix,
#> #  > overwrite = TRUE)
#> # It uses nixpkgs' revision 79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a for reproducibility purposes
#> # which will install R version 4.2.1.
#> # Report any issues to https://github.com/ropensci/rix
#> let
#>  pkgs = import (fetchTarball "https://github.com/NixOS/nixpkgs/archive/79b3d4bcae8c7007c9fd51c279a8a67acfa73a2a.tar.gz") {};
#>  
#>   rpkgs = builtins.attrValues {
#>     inherit (pkgs.rPackages) 
#>       dplyr
#>       janitor;
#>   };
#>  
#>   git_archive_pkgs = [
#>     (pkgs.rPackages.buildRPackage {
#>       name = "housing";
#>       src = pkgs.fetchgit {
#>         url = "https://github.com/rap4all/housing/";
#>         rev = "1c860959310b80e67c41f7bbdc3e84cef00df18e";
#>         sha256 = "sha256-s4KGtfKQ7hL0sfDhGb4BpBpspfefBN6hf+XlslqyEn4=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages) 
#>           dplyr
#>           ggplot2
#>           janitor
#>           purrr
#>           readxl
#>           rlang
#>           rvest
#>           stringr
#>           tidyr;
#>       };
#>     })
#> 
#> 
#>     (pkgs.rPackages.buildRPackage {
#>       name = "fusen";
#>       src = pkgs.fetchgit {
#>         url = "https://github.com/ThinkR-open/fusen";
#>         rev = "d617172447d2947efb20ad6a4463742b8a5d79dc";
#>         sha256 = "sha256-TOHA1ymLUSgZMYIA1a2yvuv0799svaDOl3zOhNRxcmw=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages) 
#>           attachment
#>           cli
#>           desc
#>           devtools
#>           glue
#>           here
#>           magrittr
#>           parsermd
#>           roxygen2
#>           stringi
#>           tibble
#>           tidyr
#>           usethis
#>           yaml;
#>       };
#>     })
#>  
#>     (pkgs.rPackages.buildRPackage {
#>       name = "AER";
#>       src = pkgs.fetchzip {
#>        url = "https://cran.r-project.org/src/contrib/Archive/AER/AER_1.2-8.tar.gz";
#>        sha256 = "sha256-OqxXcnUX/2C6wfD5fuNayc8OU+mstI3tt4eBVGQZ2S0=";
#>       };
#>       propagatedBuildInputs = builtins.attrValues {
#>         inherit (pkgs.rPackages) 
#>           car
#>           lmtest
#>           sandwich
#>           survival
#>           zoo
#>           Formula;
#>       };
#>     })
#>   ];
#>    
#>   system_packages = builtins.attrValues {
#>     inherit (pkgs) 
#>       R
#>       glibcLocales
#>       nix;
#>   };
#>   
#> in
#> 
#> pkgs.mkShell {
#>   LOCALE_ARCHIVE = if pkgs.system == "x86_64-linux" then  "${pkgs.glibcLocales}/lib/locale/locale-archive" else "";
#>   LANG = "en_US.UTF-8";
#>    LC_ALL = "en_US.UTF-8";
#>    LC_TIME = "en_US.UTF-8";
#>    LC_MONETARY = "en_US.UTF-8";
#>    LC_PAPER = "en_US.UTF-8";
#>    LC_MEASUREMENT = "en_US.UTF-8";
#>
#>   buildInputs = [ git_archive_pkgs rpkgs  system_packages   ];
#>
#> }

The next vignette, vignette("d2-installing-system-tools-and-texlive-packages-in-a-nix-environment"), explains how you can install tools such as text editors, git, Quarto, TexLive packages, and any other tool available through nixpkgs for your development environments.

Package installation issues

Some R packages are quite difficult to install: that is usually not an issue for most users that use either Windows or macOS as their operating systems, because when calling install.packages() a compiled binary gets downloaded from CRAN and installed in a matter of seconds. On Ubuntu, likely the most popular Linux distribution, binary packages for R packages are also available via the r2u repository. However, if you need to install old packages, these instead will need to be installed from source, as binaries for old packages are not kept. For most packages, this is not an issue, but some packages require compilation and this is where issues start.

Nix solves this, because all packages must have their dependencies also declared, so installing old packages should not be an issue. However, it can happen that one particular package that you want to install may not build. This can happen because, even though we spend a lot of time making sure R packages work flawlessly with Nix, there are many R packages (almost 30'000 between CRAN and Bioconductor) and there's not many of us (R contributors for Nix). Should you have trouble installing a package, feel free to open an issue and we'll do our best to fix it!

We also made sure that old packages would work by backporting many fixes, and actually building many old versions of popular packages for all the dates included in available_dates().

Here is the list of packages that were built and tested (but keep in mind that this list doesn't show all the dependencies of all the packages that also have to work, and that just because a package isn't listed, doesn't mean it's not going to work!):

Click to show package list

DBI

R6

RColorBrewer

RCurl

RSQLite

Rcpp

RcppEigen

arrow

askpass

backports

base64enc

bit

bit64

blob

broom

bslib

cachem

callr

cellranger

cli

clipr

collapse

colorspace

cpp11

crayon

curl

data_table

dbplyr

devtools

digest

dplyr

duckdb

evaluate

fansi

farver

fastmap

fontawesome

forcats

fs

gargle

generics

ggplot2

glue

gtable

haven

highr

hms

htmltools

htmlwidgets

httr

icosa

igraph

isoband

jquerylib

jsonlite

kit

knitr

labeling

languageserver

later

lifecycle

lubridate

magrittr

memoise

mime

modelr

munsell

nloptr

openssl

openxlsx

pillar

pkgconfig

prettyunits

processx

progress

promises

ps

purrr

rJava

ragg

rappdirs

readr

readxl

rematch

rematch2

rlang

rmarkdown

rprojroot

rstan

rstudioapi

rvest

sass

scales

selectr

Seurat

sf

shiny

stars

stringi

stringr

sys

systemfonts

terra

textshaping

tibble

tidyr

tidyselect

tidyverse

timechange

tinytex

tzdb

utf8

vctrs

viridisLite

withr

xfun

xlsx

xml2

yaml

zoo

rix documentation built on Sept. 11, 2025, 5:12 p.m.