packrat

Packrat is a dependency management system for R.

Use packrat to make your R projects more:

• Isolated: Installing a new or updated package for one project won't break your other projects, and vice versa. That's because packrat gives each project its own private package library.
• Portable: Easily transport your projects from one computer to another, even across different platforms. Packrat makes it easy to install the packages your project depends on.
• Reproducible: Packrat records the exact package versions you depend on, and ensures those exact versions are the ones that get installed wherever you go.

See the project page for more information, or join the discussion at packrat-discuss. Read the release notes to learn what's new in Packrat.

Quick-start Guide

Start by installing Packrat:

install.packages("packrat")

Then, start a new R session at the base directory of your project and type:

packrat::init()

This will install Packrat, set up a private library to be used for this project, and then place you in packrat mode. While in packrat mode, calls to functions like install.packages and remove.packages will modify the private project library, rather than the user library.

When you want to manage the state of your private library, you can use the Packrat functions:

• packrat::snapshot(): Save the current state of your library.
• packrat::restore(): Restore the library state saved in the most recent snapshot.
• packrat::clean(): Remove unused packages from your library.

Share a Packrat project with bundle and unbundle:

• packrat::bundle(): Bundle a packrat project, for easy sharing.
• packrat::unbundle(): Unbundle a packrat project, generating a project directory with libraries restored.

Navigate projects and set/get options with:

• packrat::on(), packrat::off(): Toggle packrat mode on and off, for navigating between projects within a single R session.
• packrat::get_opts, packrat::set_opts: Get/set project-specific settings.

Manage ad-hoc local repositories (note that these are a separate entity from CRAN-like repositories):

• packrat::set_opts(local.repos = ...) can be used to specify local repositories; that is, directories containing (unzipped) package sources.
• packrat::install_local() installs packages available in a local repository.

For example, suppose I have the (unzipped) package sources for digest located within the folder~/git/R/digest/. To install this package, you can use:

packrat::set_opts(local.repos = "~/git/R")
packrat::install_local("digest")

There are also utility functions for using and managing packages in the external / user library, and can be useful for leveraging packages in the user library that you might not want as project-specific dependencies, e.g. devtools, knitr, roxygen2:

• packrat::extlib(): Load an external package.
• packrat::with_extlib(): With an external package, evaluate an expression. The external package is loaded only for the duration of the evaluated expression, but note that there may be other side effects associated with the package's .onLoad, .onAttach and .onUnload calls that we may not be able to fully control.

Workflows

Packrat supports a set of common analytic workflows:

1. As-you-go: use packrat::init() to initialize packrat with your project, and use it to manage your project library while you develop your analysis. As you install and remove packages, you can use packrat::snapshot() and packrat::restore() to maintain the R packages in your project. For collaboration, you can either use your favourite version control system, or use packrat::bundle() to generate a bundled version of your project that collaborators can use with packrat::unbundle().

2. When-you're-done: take an existing or complete analysis (preferably collected within one directory), and call packrat::init() to immediately obtain R package sources for all packages used in your project, and snapshot that state so it can hence be preserved across time.

Setting up your own custom, CRAN-like repositories

Please view the set-up guide here for a simple walkthrough in how you might set up your own, local, custom CRAN repository.