The renv package is a new effort to bring project-local
R dependency management to your projects. The goal is for
renv to be a robust, stable replacement for the Packrat package, with
fewer surprises and better default behaviors.
Underlying the philosophy of renv is that any of your
existing workflows should just work as they did before –
renv helps manage library paths (and other project-specific
state) to help isolate your project’s R dependencies, and the existing
tools you’ve used for managing R packages (e.g.
install.packages(), remove.packages()) should
work as they did before.
Workflow
The general workflow when working with renv is:
Call
renv::init()to initialize a new project-local environment with a private R library,Work in the project as normal, installing and removing new R packages as they are needed in the project,
Call
renv::snapshot()to save the state of the project library to the lockfile (calledrenv.lock),Continue working on your project, installing and updating R packages as needed.
Call
renv::snapshot()again to save the state of your project library if your attempts to update R packages were successful, or callrenv::restore()to revert to the previous state as encoded in the lockfile if your attempts to update packages introduced some new problems.
The renv::init() function attempts to ensure the
newly-created project library includes all R packages currently used by
the project. It does this by crawling R files within the project for
dependencies with the renv::dependencies() function. The
discovered packages are then installed into the project library with the
renv::hydrate() function, which will also attempt to save
time by copying packages from your user library (rather than
reinstalling from CRAN) as appropriate.
Calling renv::init() will also write out the
infrastructure necessary to automatically load and use the private
library for new R sessions launched from the project root directory.
This is accomplished by creating (or amending) a project-local
.Rprofile with the necessary code to load the project when
the R session is started.
If you’d like to initialize a project without attempting dependency
discovery and installation – that is, you’d prefer to manually install
the packages your project requires on your own – you can use
renv::init(bare = TRUE) to initialize a project with an
empty project library.
Reproducibility
Using renv, it’s possible to “save” and “load” the state
of your project library. More specifically, you can use:
renv::snapshot()to save the state of your project torenv.lock; andrenv::restore()to restore the state of your project fromrenv.lock.
For each package used in your project, renv will record
the package version, and (if known) the external source from which that
package can be retrieved. renv::restore() uses that
information to retrieve and reinstall those packages in your
project.
Caveats
It is important to emphasize that renv is not a panacea for
reproducibility. Rather, it is a tool that can help make
projects reproducible by solving one small part of the problem: it
records the version of R + R packages being used in a project, and
provides tools for reinstalling the declared versions of those packages
in a project. Ultimately, making a project reproducible requires some
thoughtfulness from the user: what does it mean for a particular project
to be reproducible, and how can renv (and other tools) be
used to accomplish that particular goal of reproducibility?
There are a still a number of factors that can affect whether this project could truly be reproducible in the future – for example,
The results produced by a particular project might depend on other components of the system it’s being run on – for example, the operating system itself, the versions of system libraries in use, the compiler(s) used to compile R and the R packages used, and so on. Keeping a ‘stable’ machine image is a separate challenge, but Docker is one popular solution. See also
vignette("docker", package = "renv")for recommendations on how Docker can be used together withrenv.The R packages that the project depends on may no longer be available. If your project depends on R packages available on CRAN, it’s possible those packages may be removed in the future – either by request of the package maintainer, or by the maintainers of CRAN itself. This is quite rare, but needs consideration if reproducibility of a project is paramount.
In addition, be aware that package installation may fail if a package
was originally installed through a CRAN-available binary, but that
binary is no longer available. renv will attempt to install
the package from sources in this situation, but attempts to install from
source can (and often do) fail due to missing system prerequisites for
compilation of a package. The renv::equip() function may be
useful in these scenarios, especially on Windows: it will download
external software commonly used when compiling R packages from sources,
and instruct R to use that software during compilation.
A salient example of this is the rmarkdown package, as
it relies heavily on the pandoc command line utility.
However, because pandoc is not bundled with the rmarkdown
package (it is normally provided by RStudio, or installed separately by
the user), simply restoring an renv project using
rmarkdown may not be sufficient – one also needs to ensure
the project is run in a environment with the correct version of
pandoc available.
Infrastructure
The following files are written to and used by projects using
renv:
| File | Usage |
|---|---|
.Rprofile |
Used to activate renv for new R sessions launched in
the project. |
renv.lock |
The lockfile, describing the state of your project’s library at some point in time. |
renv/activate.R |
The activation script run by the project
.Rprofile. |
renv/library |
The private project library. |
renv/settings.dcf |
Project settings – see ?settings for more details. |
In particular, renv/activate.R ensures that the project
library is made active for newly launched R sessions. This ensures that
any new R processes launched within the project directory will use the
project library, and hence are isolated from the regular user
library.
For development and collaboration, the .Rprofile,
renv.lock and renv/activate.R files should be
committed to your version control system; the renv/library
directory should normally be ignored. Note that
renv::init() will attempt to write the requisite ignore
statements to the project .gitignore.
Dependency Discovery
By default, renv::snapshot() will examine your project’s
R files to determine which packages are used in your project, and will
include only those packages (alongside their recursive dependencies) in
the lockfile. This is done via a call to the
renv::dependencies() function. We call this an “implicit”
snapshot, since the packages your project depends on are implicit based
on how packages appear to be used in your project. renv
uses static analysis to determine which packages appear to be used;
e.g. by scanning your code for calls to library() or
require().
While useful, this approach is not 100% reliable in detecting the
packages required by your project. If you find that renv’s
dependency discovery is failing to discover one or more packages used in
your project, one escape hatch is to include a file called
_dependencies.R with code of the form:
library(<pkg>)
Ignore Files
By default, renv reads the .gitignore files
in your project (if any) to infer which files should be ignored when
scanning for dependencies. If you find that renv’s
dependency discovery is scanning files you don’t want to be scanned, you
can use an .renvignore file to instruct renv
to ignore certain patterns of files in the project. For example, you
might use:
/data
to tell renv not to scan files within the
data folder.
If you’d prefer that renv ignored all folders by
default, except for some subset of folders where you place your code
files, you could use something like:
*
!/code
In this case, renv will only scan your code
folder at the root of the project directory for dependencies.
Explicit Snapshots
If you’d instead prefer to explicitly declare which packages are used
in your project, you can do so by creating a DESCRIPTION
file at your project root. These DESCRIPTION files should
be formatted similarly to those used by default in R package development
– see https://r-pkgs.org/description.html for more
details.
In this case, your DESCRIPTION file might look like:
Type: project
Description: My project.
Depends:
tidyverse,
devtools,
shiny,
data.table
The packages used in your project can be part of either the
Depends or Imports fields.
Collaborating
When sharing a project with other collaborators, you may want to
ensure everyone is working with the same environment – otherwise, code
in the project may unexpectedly fail to run because of changes in
behavior between different versions of the packages in use.
renv can help to make such collaboration easier – see
vignette("collaborating", package = "renv") for more
details.
Package Sources
renv is able to install and restore packages from a
variety of sources, including:
renv uses an installed package’s
DESCRIPTION file to infer its source. For example, packages
installed from the CRAN repositories typically have the field:
Repository: CRAN
set, and renv takes this as a signal that the package
was retrieved from CRAN.
Inferring Package Sources
The following fields are checked, in order, when inferring a package’s source:
The
RemoteTypefield; typically written for packages installed by thedevtools,remotesandpakpackages,The
Repositoryfield; for example, packages retrieved from CRAN will typically have theRepository: CRANfield,The
biocViewsfield; typically present for packages installed from the Bioconductor repositories,
As a fallback, if renv is unable to determine a
package’s source from the DESCRIPTION file directly, but a
package of the same name is available in the active R repositories (as
specified in getOption("repos")), then the package will be
treated as though it was installed from an R package repository.
If all of the above methods fail, renv will finally
check for a package available from the cellar. See here for more details. The package cellar is
typically used as an escape hatch, for packages which do not have a
well-defined remote source, or for packages which might not be remotely
accessible from your machine.
Unknown Sources
If renv is unable to infer a package’s source, it will
inform you during renv::snapshot() – for example, if we
attempted to snapshot a package called skeleton with no
known source:
> renv::snapshot()
The following package(s) were installed from an unknown source:
skeleton
renv may be unable to restore these packages in the future.
Consider reinstalling these packages from a known source (e.g. CRAN).
Do you want to proceed? [y/N]:
While you can still create a lockfile with such packages,
restore() will likely fail unless you can ensure this
package is installed through some other mechanism.
Custom R Package Repositories
Custom and local R package repositories are supported as well. The
only requirement is that these repositories are set as part of the
repos R option, and that these repositories are named. For
example, you might use:
repos <- c(CRAN = "https://cloud.r-project.org", WORK = "https://work.example.org")
options(repos = repos)
to tell renv to work with both the official CRAN package
repository, as well as a package repository you have hosted and set up
in your work environment.
Upgrading renv
After initializing a project with renv, that project
will then be ‘bound’ to the particular version of renv that
was used to initialize the project. If you need to upgrade (or otherwise
change) the version of renv associated with a project, you
can use renv::upgrade(). This will install the
latest-available version of renv from your declared package
repositories. Alternatively, if you’re currently using a development
version of renv as installed from GitHub in your project,
then renv will install the latest-available version of
renv from GitHub.
With each commit of renv, we bump the package version
and also tag the commit with the associated package version. This
implies that you can call, for example:
renv::upgrade(version = "0.15.5")
to request the installation of that particular version of
renv if so required.
Cache
One of renv’s primary features is the use of a global
package cache, which is shared across all projects using
renv. The renv package cache provides two
primary benefits:
Future calls to
renv::restore()andrenv::install()will become much faster, asrenvwill be able to find and re-use packages already installed in the cache.Because it is not necessary to have duplicate versions of your packages installed in each project, the
renvcache should also help you save disk space relative to an approach with project-specific libraries without a global cache.
To understand the renv cache, we need to first
understand what an R library is. An R library is, normally, a
directory of installed R packages which can be loaded and used within an
R session. These are the directories reported by e.g.
.libPaths(), and R uses these directories when searching
for packages to load (e.g. in response to a call to
library(dplyr)).
When using renv with the global package cache, the
project library is instead formed as a directory of symlinks (or, on
Windows, junction points) into the renv global package
cache. Hence, while each renv project is isolated from
other projects on your system, they can still re-use the same installed
packages as required.
In some cases, renv will be unable to directly link from
the global package cache to your project library – for example, if the
package cache and your project library live on different disk volumes.
In such a case, renv will instead copy the package from the
cache into the project library.
By default, renv generates its cache in the following
folders:
| Platform | Location |
|---|---|
| Linux | ~/.local/share/renv |
| macOS | ~/Library/Application Support/renv |
| Windows | %LOCALAPPDATA%/renv |
If you’d like to share the package cache across multiple users, you
can do so by setting the RENV_PATHS_CACHE environment
variable to a shared path. This variable can be set in an R startup file
to make it apply to all R sessions. For example, it could be set
within:
- A project-local
.Renviron; - The user-level
~/.Renviron; - A site-wide file at
$(R RHOME)/etc/Renviron.site.
You may also want to set RENV_PATHS_CACHE so that the
global package cache can be stored on the same volume as the projects
you normally work on. This is especially important when working projects
stored on a networked filesystem.
While we recommend enabling the cache by default, if you’re having
trouble with renv when the cache is enabled, it can be
disabled by setting the project setting
renv::settings$use.cache(FALSE). Doing this will ensure
that packages are then installed into your project library directly,
without attempting to link and use packages from the renv
cache.
If you find a problematic package has entered the cache (for example,
an installed package has become corrupted), that package can be removed
with the renv::purge() function. See the
?purge documentation for caveats and things to be aware of
when removing packages from the cache.
Installation from Source
In the end, renv still needs to install R packages –
either from binaries available from CRAN, or from sources when binaries
are not available. Installation from source can be challenging for a few
reasons:
Your system will need to have a compatible compiler toolchain available. In some cases, R packages may depend on C / C++ features that aren’t available in an older system toolchain, especially in some older Linux enterprise environments.
Your system will need requisite system libraries, as many R packages contain compiled C / C++ code that depend on and link to these packages.
Downloads
By default, renv uses curl for file downloads when
available. This allows renv to support a number of download
features across multiple versions of R, including:
- Custom headers (used especially for authentication),
- Connection timeouts,
- Download retries on transient errors.
If curl is not available on your machine, it is highly
recommended that you install it. Newer versions of Windows 10 come with
a bundled version of curl.exe; other users on Windows can
use renv::equip() to download and install a recent copy of
curl. Newer versions of macOS come with a bundled version
of curl that is adequate for usage with renv,
and most Linux package managers have a modern version of
curl available in their package repositories.
curl downloads can be configured through
renv’s configuration settings – see
?renv::config for more details.
If you’ve already configured R’s downloader and would like to bypass
renv’s attempts to use curl, you can use the R
option renv.download.override. For example, executing:
options(renv.download.override = utils::download.file)would instruct renv to use R’s own download machinery
when attempting to download files from the internet (respecting the R
options download.file.method and
download.file.extra as appropriate). Advanced users can
also provide their own download function, provided its signature matches
that of utils::download.file().
You can also instruct renv to use a different download
method by setting the RENV_DOWNLOAD_METHOD environment
variable. For example:
# use Windows' internal download machinery
Sys.setenv(RENV_DOWNLOAD_METHOD = "wininet")
# use R's bundled libcurl implementation
Sys.setenv(RENV_DOWNLOAD_METHOD = "libcurl")
Note that other features (e.g. authentication) may not be supported
when using an alternative download file method – you will have to
configure the downloader yourself if that is required. See
?download.file for more details.
Proxies
If your downloads need to go through a proxy server, then there are a variety of approaches you can take to make this work:
Set the
http_proxyand / orhttps_proxyenvironment variables. These environment variables can contain the full URL to your proxy server, including a username + password if necessary.You can use a
.curlrc(_curlrcon Windows) to provide information about the proxy server to be used. This file should be placed in your home folder (seeSys.getenv("HOME"), orSys.getenv("R_USER")on Windows); alternatively, you can set theCURL_HOMEenvironment variable to point to a custom ‘home’ folder to be used bycurlwhen resolving the runtime configuration file. On Windows, you can also place your_curlrcin the same directory where thecurl.exebinary is located.
See the curl documentation on proxies and config files for more details.
As an example, the
following _curlrc works when using authentication with NTLM
and SSPI on Windows:
--proxy "your.proxy.dns:port"
--proxy-ntlm
--proxy-user ":"
--insecure
The curl R package also has a helper:
curl::ie_get_proxy_for_url()
which may be useful when attempting to discover this proxy address.
Authentication
Your project may make use of packages which are available from remote
sources requiring some form of authentication to access – for example, a
GitHub enterprise server. Usually, either a personal access token (PAT)
or username + password combination is required for authentication.
renv is able to authenticate when downloading from such
sources, using the same system as the remotes package.
In particular, environment variables are used to record and transfer the
required authentication information.
| Remote Source | Authentication |
|---|---|
| GitHub | GITHUB_PAT |
| GitLab | GITLAB_PAT |
| Bitbucket |
BITBUCKET_USER + BITBUCKET_PASSWORD
|
| Git Remotes |
GIT_PAT / GIT_USER +
GIT_PASSWORD
|
These credentials can be stored in e.g. .Renviron, or
can be set in your R session through other means as appropriate.
If you require custom authentication for different packages (for
example, your project makes use of packages available on different
GitHub enterprise servers), you can use the renv.auth R
option to provide package-specific authentication settings.
renv.auth can either be a a named list associating package
names with environment variables, or a function accepting a package name
+ record, and returning a list of environment variables. For
example:
# define a function providing authentication
options(renv.auth = function(package, record) {
if (package == "MyPackage")
return(list(GITHUB_PAT = "<pat>"))
})
# use a named list directly
options(renv.auth = list(
MyPackage = list(GITHUB_PAT = "<pat>")
))
# alternatively, set package-specific option
options(renv.auth.MyPackage = list(GITHUB_PAT = "<pat>"))For packages installed from Git remotes, renv will
attempt to use git from the command line to download and
restore the associated package. Hence, it is recommended that
authentication is done through SSH keys when possible.
Authentication with Custom Headers
If you want to set arbitrary headers when downloading files using
renv, you can do so using the
renv.download.headers R option. It should be a function
that accepts a URL, and returns a named character vector indicating the
headers which should be supplied when accessing that URL.
For example, suppose you have a package repository hosted at
https://my/repository, and the credentials required to
access that repository are stored in the AUTH_HEADER
environment variable. You could define
renv.download.headers like so:
options(renv.download.headers = function(url) {
if (grepl("^https://my/repository", url))
return(c(Authorization = Sys.getenv("AUTH_HEADER")))
})With the above, renv will set the
Authorization header whenever it attempts to download files
from the repository at URL https://my/repository.
Shims
To help you take advantage of the package cache, renv
places a couple of shims on the search path:
| Function | Shim |
|---|---|
install.packages() |
renv::install() |
remove.packages() |
renv::remove() |
update.packages() |
renv::update() |
In effect, calling install.packages() within an
renv project will call renv::install()
instead. This can be useful when installing packages which have already
been cached. For example, if you use
renv::install("dplyr"), and renv detects that
the latest version on CRAN has already been cached, then
renv will just install using the copy available in the
cache – thereby skipping some of the installation overhead.
If you’d like to bypass these shims within an session, you can
explicitly call the version of these functions from the
utils package, e.g. with
utils::install.packages(<...>).
If you’d prefer not to use the renv shims at all, they
can be disabled by setting the R option
options(renv.config.shims.enabled = FALSE), or by setting
the environment variable RENV_CONFIG_SHIMS_ENABLED = FALSE.
See ?config for more details.
History
If you’re using a version control system with your project, then as
you call renv::snapshot() and later commit new lockfiles to
your repository, you may find it necessary later to recover older
versions of your lockfiles. renv provides the functions
renv::history() to list previous revisions of your
lockfile, and renv::revert() to recover these older
lockfiles.
Currently, only Git repositories are supported by
renv::history() and renv::revert().
Comparison with Packrat
renv differs from Packrat in the following ways:
The
renvlockfilerenv.lockis formatted as JSON. This should make the lockfile easier to use and consume with other tools.renvno longer attempts to explicitly download and track R package source tarballs within your project. This was a frustrating default that operated under the assumption that you might later want to be able to restore a project’s private library without access to a CRAN repository. In practice, this is almost never the case, and the time spent downloading + storing the package sources seemed to outweigh the potential reproducibility benefits.Packrat tried to maintain the distinction between so-called ‘stale’ packages; that is, R packages which were installed by Packrat but were not recorded in the lockfile for some reason. This distinction was (1) overall not useful, and (2) confusing.
renvno longer makes this distinction:snapshot()saves the state of your project library torenv.lock,restore()loads the state of your project library fromrenv.lock, and that’s all.In
renv, the global package cache is enabled by default. This should reduce overall disk-space usage as packages can effectively be shared across each project usingrenv.renv’s dependency discovery machinery is more configurable. The functionrenv::dependencies()is exported, and users can create.renvignorefiles to instructrenvto ignore specific files and folders in their projects. (See?renv::dependenciesfor more information.)
Migrating from Packrat
The renv::migrate() function makes it possible to
migrate projects from Packrat to renv. See the
?migrate documentation for more details. In essence,
calling renv::migrate("<project path>") will be
enough to migrate the Packrat library and lockfile such that they can
then be used by renv.
Uninstalling renv
If you find renv isn’t the right fit for your project,
deactivating and uninstalling it is easy.
To deactivate
renvin a project, userenv::deactivate(). This removes therenvauto-loader from the project.Rprofile, but doesn’t touch any otherrenvfiles used in the project. If you’d like to later re-activaterenv, you can do so withrenv::activate().To remove
renvfrom a project, userenv::deactivate()to first remove therenvauto-loader from the project.Rprofile, then delete the project’srenvfolder andrenv.locklockfile as desired.
If you want to completely remove any installed renv
infrastructure components from your entire system, you can do so with
the following R code:
root <- renv::paths$root()
unlink(root, recursive = TRUE)
The renv package can then also be uninstalled via:
utils::remove.packages("renv")
Note that if you’ve customized any of renv’s
infrastructure paths as described in ?renv::paths, then
you’ll need to find and remove those customized folders as well.
Future Work
renv, like Packrat, is designed to work standalone
without the need to depend on any non-base R packages. However, the
following (future) integrations are planned:
Use pak for parallel package installation,
Use sysreqsdb to validate and install system dependencies as required before attempting to install the associated packages.
These integrations will be optional (so that renv can
always work standalone) but we hope that they will further improve the
speed and reliability of renv.