Best Practices

Rapp 0.3.0

Rapp is an R front-end (like Rscript) that turns simple scripts into polished CLIs, with automatic argument parsing, generated help, and support for commands and installable launchers

Tomasz Kalinowski

mirai 2.6.0

mirai 2.6.0 brings an HTTP launcher for cloud and Posit Workbench, a C-level dispatcher for near-zero overhead task dispatch, and new tools for racing and debugging async tasks

Charlie Gao

nanonext 1.8.0

nanonext 1.8.0 adds a low-level streaming HTTP/WebSocket server to R's web infrastructure, with TLS support, new async primitives, and redesigned documentation

Charlie Gao

yaml12: YAML 1.2 for R and Python

We’re pleased to announce two new YAML 1.2 packages: `yaml12` for R and `py-yaml12` for Python. Both are implemented in Rust and designed for fast, predictable YAML 1.2 parsing, with safe opt-in tag handling and document stream support

Tomasz Kalinowski

testthat 3.3.0

testthat 3.3.0 brings improved expectations with better error messages, new expectations for common testing patterns, and lifecycle changes including the removal of `local_mock()` and `with_mock()`. It also includes a write-up of my experience doing package development with Claude Code

Hadley Wickham

pkgdown 2.2.0

The latest version of pkgdown automatically builds markdown files that make it easy for LLMs to use your website

Hadley Wickham

mirai 2.5.0

mirai - minimalist async evaluation framework for R - brings production-grade parallel and distributed computing to the ecosystem

Charlie Gao

nanonext 1.7.0

nanonext is a communications and concurrency toolbox that facilitates fast, powerful and reliable data exchange in polyglot data science workflows

Charlie Gao

Air 0.7.0

Read all about Air 0.7.0, including: even better Positron support, a new feature we call autobracing, and an official GitHub Action!

Davis Vaughan, Lionel Henry

Fonts in R

Taking control of fonts and text rendering in R can be challenging. This deep-dive teaches you everything (and then some) you need to know to keep your sanity

Thomas Lin Pedersen

S7 0.2.0

S7 is a new package that simplifies object-oriented programming (OOP) in R. It combines the simplicity of S3 with the structure of S4 to create a clearer system that's accessible to everyone

Tomasz Kalinowski, Hadley Wickham

pkgdown 2.1.0

pkgdown 2.1.0 includes two major new features: support for quarto vignettes and a "light switch" that lets the reader switch between light and dark mode. It also contains a bunch of other improvements to both the user and the developer experience

Hadley Wickham

withr 3.0.0

withr is the tidyverse solution for automatically cleaning up after yourselves (temporary files, options, etc). This milestone makes withr much faster

Lionel Henry

roxygen2 7.3.0

This release automatically warns if you forget to export an S3 method, regenerates the `NAMESPACE` before documenting the rest of the package, and does a better job generating aliases for the package documentation topic

Hadley Wickham

Three ways errors are about to get better in tidymodels

The tidymodels team's biannual spring cleaning gave us a chance to revisit the way we raise some error messages

Simon Couch

testthat 3.2.0

Catch up on the last two years of testthat development which includes improved documentation, new expectations, a new style for error snapshots, support for mocking, a new way to detect if a test has changed global state, and a handful of UI improvements

Hadley Wickham

pak 0.6.0

pak installs R packages from various sources. pak is fast, safe and convenient. This post shows the improvements in system requirements support on Linux, in the just released pak 0.6.0 version

Gábor Csárdi

webR 0.2.0 has been released

webR 0.2.0 has been released. Updates and improvements to the webR REPL app, HTML canvas graphics device, internationalisation, Wasm R packages, Shiny support, and the developer API

George Stagg

Understanding LoRA with a minimal example

LoRA (Low Rank Adaptation) is a new technique for fine-tuning deep learning models that works by reducing the number of trainable parameters and enables efficient task switching. In this blog post we will talk about the key ideas behind LoRA in a very minimal torch example

Daniel Falbel

What are Large Language Models? What are they not?

This is a high-level, introductory article about Large Language Models (LLMs), the core technology that enables the much-en-vogue chatbots as well as other Natural Language Processing (NLP) applications. It is directed at a general audience, possibly with some technical and/or scientific background, but no knowledge is assumed of either deep learning or NLP. Having looked at major model ingredients, training workflow, and mechanics of output generation, we also talk about what these models are not

Package spring cleaning

When Spring comes around, it's time to embark on some Spring Cleaning to take care of the package maintenance tasks that you never seem to get around to. This post outlines the processes and tools that the tidyverse team uses to make this job fun and efficient

Tidyteam code review principles

We've written a collection of tidyteam code review principles that act as a resource for new contributors and as a source of truth when there are questions about our code review process. We hope you find them useful!

Davis Vaughan

Group-equivariant neural networks with escnn

Escnn, built on PyTorch, is a library that, in the spirit of Geometric Deep Learning, provides a high-level interface to designing and training group-equivariant neural networks. This post introduces important mathematical concepts, the library's key actors, and essential library use

Differences between the base R and magrittr pipes

A discussion of the (relatively minor) differences between the native R pipe, `|>`, and the magrittr pipe, `%>%`

Hadley Wickham

Writing performant code with tidy tools

When performance becomes an issue for code using tidy interfaces, switching to the backend tools used by tidy developers can offer substantial speedups

Simon Couch

Deep Learning and Scientific Computing with R torch: the book

Please allow us to introduce Deep Learning and Scientific Computing with R torch. Released in e-book format today, and available freely online, this book starts out by introducing torch basics. From there, it moves on to various deep-learning use cases. Finally, it shows how to use torch for more general topics, such as matrix computations and the Fourier Transform

New CRAN requirements for packages with C and C++

A few recent changes in CRAN requirements for packages containing C or C++ code have caused package developers some headaches. This post outlines the issues and provides solutions the tidyverse team has used to address them

webR 0.1.0 has been released

webR 0.1.0 has been released! Using the magic of WebAssembly, webR allows you to run R code directly within a web browser

George Stagg

Upside down, a cat's still a cat: Evolving image recognition with Geometric Deep Learning

In this first in a series of posts on group-equivariant convolutional neural networks (GCNNs), meet the main actors — groups — and concepts (equivariance). With GCNNs, we finally revisit the topic of Geometric Deep Learning, a principled, math-driven approach to neural networks that has consistently been rising in scope and impact

Wavelet Transform - with torch

torch does not have built-in functionality to do wavelet analysis. But we can efficiently implement what we need, making use of the Fast Fourier Transform (FFT). This post is a very first introduction to wavelets, suitable for readers that have not encountered it before. At the same time, it provides useful starter code, showing an (extensible) way to perform wavelet analysis in torch. It is an excerpt from the corresponding chapter in the forthcoming book, Deep Learning and Scientific Computing with R torch, to be published by CRC Press

Discrete Fourier Transform - with torch

About the Fourier Transform, it has been said that it is one of the greatest wonders of the universe. At the same time, it can be realized in a mere half-dozen lines of code. Even if in the end, you're just going to call torch's built-in functions directly, it helps to understand, and be able to reproduce in code, the ideas that underlie the magic. This post is an excerpt from the forthcoming book, Deep Learning and Scientific Computing with R torch, to be published by CRC Press

Five ways to do least squares (with torch)

Get to know torch's linalg module, all while learning about different ways to do least-squares regression from scratch. This post is a condensed version of the corresponding chapter in the forthcoming book, Deep Learning and Scientific Computing with R torch, to be published by CRC Press

Make your ggplot2 extension package understand the new linewidth aesthetic

The next release of ggplot2 will contain a number of internal improvements and fixes long-time inconsistencies. One of these are the conflation of point size and linewidth into the same aesthetic. This post will go into detail with how you can make your extension package work well with the new linewidth aesthetic

Thomas Lin Pedersen

lintr 3.0.0

lintr 3.0.0 is a major release featuring a more consistent API for using linter and dozens of new linters included

GitHub Actions for R developers, v2

We have updated our GitHub Actions at `r-lib/actions`. Consider upgrading to the new `v2` version, for faster and more reliable GHA jobs

Gábor Csárdi

roxygen2 7.2.0

roxygen2 7.2.0 brings improvements to `NAMESPACE` generation, better multiparameter argument inheritance, and improved warnings

Hadley Wickham

Using case weights with tidymodels

Support for case weights is now available across many tidymodels packages

Max Kuhn

ragg, svglite, and the new graphics features

The graphics engine in R is continually improving, and the ragg and svglite graphic devices follow along in supporting them. This blog post outlines what is now possible, and what might come in the future

Thomas Lin Pedersen

New error style coming up in rlang 1.0.0

rlang 1.0.0 is near and introduces a new style of error display. Feedback welcome!

Lionel Henry

pkgdown 2.0.0

pkgdown 2.0.0 includes a major refresh of the default template (now using bootstrap 5), many new ways to customise your site, improvements to code styling, and much, much, more

Hadley Wickham

cli 3.1.0

cli 3.1.0 introduces customizable color palettes plus it comes with a number of other smaller improvements

Gábor Csárdi

Survival Analysis in tidymodels

We are working on extending support for survival analysis in tidymodels. We are looking for early adopters to try out the new package called censored and give feedback

Hannah Frick

testthat 3.1.0

testthat 3.1.0 brings improvements to snapshot testing and one breaking change

Hadley Wickham

Beyond alchemy: A first look at geometric deep learning

Geometric deep learning is a "program" that aspires to situate deep learning architectures and techniques in a framework of mathematical priors. The priors, such as various types of invariance, first arise in some physical domain. A neural network that well matches the domain will preserve as many invariances as possible. In this post, we present a very conceptual, high-level overview, and highlight a few applications

waldo 0.3.0

waldo 0.3.0 improves the display of data frame differences, and gives the objects being compared the ability to control the detail of their comparisons

Hadley Wickham

Starting to think about AI Fairness

The topic of AI fairness metrics is as important to society as it is confusing. Confusing it is due to a number of reasons: terminological proliferation, abundance of formulae, and last not least the impression that everyone else seems to know what they're talking about. This text hopes to counteract some of that confusion by starting from a common-sense approach of contrasting two basic positions: On the one hand, the assumption that dataset features may be taken as reflecting the underlying concepts ML practitioners are interested in; on the other, that there inevitably is a gap between concept and measurement, a gap that may be bigger or smaller depending on what is being measured. In contrasting these fundamental views, we bring together concepts from ML, legal science, and political philosophy

Off-label uses in ggplot2

How can ggplot2 break with no breaking changes? This short post goes into detail with how and why the latest ggplot2 release disrupted some users workflow

Thomas Lin Pedersen

vdiffr 1.0.0

This major release of vdiffr includes an updated SVG engine and integrates with the snapshot management mechanism of testthat 3

Lionel Henry

reprex 2.0.0

reprex is now much easier to use on RStudio Server and RStudio Cloud

Jenny Bryan

Online Educational Content Development: Pain Points & the Future

Frustrations and possible solutions for online course development

lifecycle 1.0.0

The lifecycle package documentation received a major overhaul based on what I learned preparing for my rstudio::global keynote

Hadley Wickham

Modern Text Features in R

ragg has taken a major leap forward in text-rendering capabilities with the latest releases of systemfonts, textshaping, and ragg itself. This post will go into detail with what is now possible and how it compares to the build in devices

Thomas Lin Pedersen

reprex 1.0.0

We've never blogged about reprex before, so the release of v1.0.0 seems like a good occasion for it

Jenny Bryan

usethis 2.0.0

This is a big release aimed at improving usability, especially around Git and GitHub functionality

Jenny Bryan

testthat 3.0.0

testhat 3.0.0 brings a raft of major improvements including snapshot testing and parallel testing. It also includes a new "edition" that allows you opt-in to a set of substantial improvements that are not backward compatible

Hadley Wickham

waldo

waldo is a new package that makes it easier to see the differences between a pair of complex R objects

Hadley Wickham

pkgdown 1.6.0

This release mostly contains bug fixes and minor improvements, but it now uses the downlit and ragg packages for syntax highlighting and graphical output, respectively

Hadley Wickham

How to Deliver Maximum Value Using R & Python

For data science teams to be successful, they need to embrace both R and Python. The ease of interoperability gives the user the flexibility to fill in any tool gaps for their own needs

Parallelized sampling using exponential variates

How can the seemingly iterative process of weighted sampling without replacement be transformed into something highly parallelizable? Turns out a well-known technique based on exponential variates accomplishes exactly that

Finding Errors in Data Analyses

A preview of a research project studying how and how well data scientists find errors

dplyr 1.0.0 for package developers

dplyr 1.0.0 is scheduled for release on May 15. This blog post talks about what package maintainers can do to prepare

Hadley Wickham

Self-cleaning test fixtures

A wild romp through environments -- namely, the environments associated with functions and tests. How to adopt a low-impact lifestyle

Jenny Bryan

usethis 1.6.0

usethis 1.6.0: GitHub Actions for CI, smoother package creation, and improved PR workflows

Hadley Wickham

tibble 3.0.0

tibble 3.0.0 is on CRAN now! Tibbles are a modern reimagining of the data frame, keeping what time has shown to be effective, and throwing out what is not, with nicer default output too! This article describes the latest major release and provides an outlook on further developments

pkgdown 1.5.0

pkgdown 1.5.0: article descriptions, better navbar control, and richer Open Graph metadata

Hadley Wickham

roxygen2 7.1.0

A minor roxygen2 release with some new features

Gábor Csárdi

Infinite surprise - the iridescent personality of Kullback-Leibler divergence

Kullback-Leibler divergence is not just used to train variational autoencoders or Bayesian networks (and not just a hard-to-pronounce thing). It is a fundamental concept in information theory, put to use in a vast range of applications. Most interestingly, it's not always about constraint, regularization or compression. Quite on the contrary, sometimes it is about novelty, discovery and surprise

Tidy eval now supports glue strings

Use {{ and { with glue strings in tidy eval to create dynamic column names in tidyverse verbs

Lionel Henry

NumPy-style broadcasting for R TensorFlow users

Broadcasting, as done by Python's scientific computing library NumPy, involves dynamically extending shapes so that arrays of different sizes may be passed to operations that expect conformity - such as adding or multiplying elementwise. In NumPy, the way broadcasting works is specified exactly; the same rules apply to TensorFlow operations. For anyone who finds herself, occasionally, consulting Python code, this post strives to explain

reticulate 1.14

reticulate 1.14: automatic Python environment configuration for R packages that use Python

Ten Quick Tips for Making Your Stuff Findable

Ten ways to ensure users can actually find your work product

roxygen2 7.0.0

A massive update to roxygen2 now on CRAN

Hadley Wickham

testthat 2.3.0

testthat 2.3.0: verify_output() for testing printed output and improved parallel testing

Hadley Wickham

renv: Project Environments for R

Introducing renv: isolated, portable, reproducible R project environments to replace Packrat

styler 1.2.0

A major update of the styler package is available on CRAN now

On leapfrogs, crashing satellites, and going nuts: A very first conceptual introduction to Hamiltonian Monte Carlo

TensorFlow Probability, and its R wrapper tfprobability, provide Markov Chain Monte Carlo (MCMC) methods that were used in a number of recent posts on this blog. These posts were directed to users already comfortable with the method, and terminology, per se, which readers mainly interested in deep learning won't necessarily be. Here we try to make up leeway, introducing Hamitonian Monte Carlo (HMC) as well as a few often-heard "buzzwords" accompanying it, always striving to keep in mind what it is all "for"

devtools 2.2.1

Make package development easier by providing R functions that simplify and expedite common tasks

Multi Process Task Queue in 100 Lines of R Code

We use `callr::r_session` to implement a worker pool and task queue in 100 lines of R code

Gábor Csárdi

pkgdown 1.4.0

pkgdown 1.4.0: deployment vs development modes for building package websites

Hadley Wickham

So, how come we can use TensorFlow from R?

Have you ever wondered why you can call TensorFlow - mostly known as a Python framework - from R? If not - that's how it should be, as the R packages keras and tensorflow aim to make this process as transparent as possible to the user. But for them to be those helpful genies, someone else first has to tame the Python

callr 3.3.0

callr 3.3.0 is now on CRAN

Gábor Csárdi

rlang 0.4.0

rlang 0.4.0 is now on CRAN!

Lionel Henry

Adding uncertainty estimates to Keras models with tfprobability

As of today, there is no mainstream road to obtaining uncertainty estimates from neural networks. All that can be said is that, normally, approaches tend to be Bayesian in spirit, involving some way of putting a prior over model weights. This holds true as well for the method presented in this post: We show how to use tfprobability, the R interface to TensorFlow Probability, to add uncertainty estimates to a Keras model in an elegant and conceptually plausible way

It depends - A dialog about dependencies

It depends - A dialog about dependencies

Resource Cleanup in C and the R API

How to avoid resource leaks in C code in R packages? Introducing the cleancall package

Gábor Csárdi, Lionel Henry

testthat 2.1.0

Highlights include that `context()` is now optional, and new `expect_invisible()` and `expect_mapequal()` expectations

Hadley Wickham

usethis 1.5.0

usethis 1.5.0 is now on CRAN

Jenny Bryan

Getting into the flow: Bijectors in TensorFlow Probability

Normalizing flows are one of the lesser known, yet fascinating and successful architectures in unsupervised deep learning. In this post we provide a basic introduction to flows using tfprobability, an R wrapper to TensorFlow Probability. Upcoming posts will build on this, using more complex flows on more complex data

R version support

Which versions of R do tidyverse packages support?

Math, code, concepts: A third road to deep learning

Not everybody who wants to get into deep learning has a strong background in math or programming. This post elaborates on a concepts-driven, abstraction-based way to learn what it's all about

Audio classification with Keras: Looking closer at the non-deep learning parts

Sometimes, deep learning is seen - and welcomed - as a way to avoid laborious preprocessing of data. However, there are cases where preprocessing of sorts does not only help improve prediction, but constitutes a fascinating topic in itself. One such case is audio classification. In this post, we build on a previous post on this blog, this time focusing on explaining some of the non-deep learning background. We then link the concepts explained to updated for near-future releases TensorFlow code

rlang 0.3.1

rlang 0.3.1 is now on CRAN

Lionel Henry

vdiffr 0.3.0

vdiffr 0.3.0 is now on CRAN

Lionel Henry

lobstr 1.0.0

lobstr is now on CRAN. lobstr provides tools that allow you to dig into the detail of an object

pkgdown 1.3.0

pkgdown 1.3.0 is now on CRAN!

generics 0.0.1

Reduce package dependencies by using common lightweight S3 generics

Max Kuhn, Davis Vaughan

devtools 2.0.0

Make package development easier by providing R functions that simplify and expedite common tasks

rlang 0.3.0

API for working with tidy evaluation, base R types, and errors

Lionel Henry

Winner takes all: A look at activations and cost functions

Why do we use the activations we use, and how do they relate to the cost functions they tend to co-appear with? In this post we provide a conceptual introduction

processx 3.2.0

Run external processes with processx

Gábor Csárdi

roxygen2 6.1.0

roxygen2 6.1.0 is now on CRAN!

Tidy evaluation in ggplot2

Using tidy evaluation in ggplot2 3.0.0

bench 1.0.1

bench 1.0.1 is now available on CRAN. bench allows you to benchmark code, by tracking execution time, memory allocations and garbage collections

The tidyverse is for EDA, not packages

Why the tidyverse package shouldn't be in your dependencies

pkgdown 1.1.0

pkgdown 1.1.0 is now on CRAN

conflicted: a new approach to resolving ambiguity

Hadley Wickham

pkgdown 1.0.0

Hadley Wickham

rlang 0.2.0

rlang 0.2.0 is on CRAN! This version brings many improvements to tidy evaluation: improved !! syntax and printing of quosures in colours

Lionel Henry

usethis 1.3.0

usethis 1.3.0 is now on CRAN

Jenny Bryan

fs 1.0.0

fs 1.0.0 is now available on CRAN. fs provides a cross-platform, uniform interface to file system operations

styler 1.0.0

styler, a package that can format your code according to the tidyverse style guide, is now on CRAN

testthat 2.0.0

Hadley Wickham

Project-oriented workflow

Advice on workflows for developing R scripts. How to think about whether an action belongs in the script or elsewhere

Jenny Bryan

usethis 1.0.0 (and 1.1.0)

Hadley Wickham

withr 2.1.0

withr 2.1.0 is now available on CRAN

glue 1.2.0

glue 1.2.0 is now available on CRAN. glue is designed to make it easy to interpolate ("glue") your data into strings

Contributing code to the tidyverse

Contributing code to open source projects can be intimidating. These projects are often widely used and have well known maintainers. Contributing code and having it accepted seems an almost insurmountable task. However if you follow a few simple strategies you can have your code accepted into even the most popular projects in the tidyverse

roxygen2 6.0.0

roxygen2 6.0.0 introduces markdown syntax for documentation and improved inheritance with @inherit tags

Hadley Wickham

testthat 1.0.0

testthat 1.0.0 adds pipe support, new expectations, consistent side-effect testing, and C++ unit testing with Catch

Hadley Wickham

R on Travis-CI

Improved R support on Travis-CI with container-based infrastructure, package caching, and multiple R versions

memoise 1.0.0

memoise 1.0.0 caches function results with better printing, argument forwarding, and time-based invalidation

Devtools 1.10.0

devtools 1.10.0 improves Windows RTools detection, adds use_news_md() and use_mit_license(), with lazy GitHub installs

Hadley Wickham

RcppParallel: Getting R and C++ to work (some more) in parallel

RcppParallel provides portable parallel C++ algorithms using Intel TBB or TinyThread, achieving 200-300x speedups

RStudio Team

roxygen2 5.0.0

roxygen2 5.0.0 adds RoxygenNote for version tracking, easier package documentation, and support for documenting re-exported functions

Hadley Wickham

testthat 0.11.0

testthat 0.11.0 adds expect_silent() for checking no output/warnings, skip_on_os() and skip_on_appveyor(), plus random encouragement

Hadley Wickham

devtools 1.9.1

devtools 1.9.1 adds remote dependencies via DESCRIPTION's Remotes field, improved GitHub integration, and better revdep_check()

Hadley Wickham

testthat 0.10.0

testthat 0.10.0 improves R CMD check output, adds skip_if_not_installed(), and soft-deprecates expect_that() style

Hadley Wickham

devtools 1.8.0

devtools 1.8 adds use_git(), use_test(), dr_devtools(), and better dependency management with package_deps()

Hadley Wickham

devtools 1.6

devtools 1.6 improves install_github(), adds session_info(), and introduces use_data(), use_testthat(), use_travis() helpers

Hadley Wickham

testthat 0.9

testthat 0.9 adds skip() and skip_on_cran(), describe() for BDD-style tests, with_mock() for mocking, and reference comparisons

Hadley Wickham

Packrat on CRAN

Packrat 0.4.1 hits CRAN with local repository support, experimental package caching, and Windows improvements

Announcing Packrat v0.4

Packrat 0.4 adds automatic snapshots, bundle/unbundle for sharing, on/off mode, and RStudio IDE integration

roxygen2 4.0.1

roxygen2 4.0.1 adds strict parsing with line numbers in errors, new vignettes, and safe file management

Hadley Wickham

devtools 1.5

devtools 1.5 adds add_test_infrastructure(), add_travis(), add_rstudio_project() and safer install_github()

Hadley Wickham

testthat 0.8 (and 0.8.1)

testthat 0.8 moves tests to tests/testthat/, adds line numbers in errors, and four new expectations

Hadley Wickham

roxygen2 3.1.0

roxygen2 3.1.0 automatically documents reference class methods from docstrings

Hadley Wickham

roxygen2 3.0.0

roxygen2 3.0.0 adds seamless S4 and RC class documentation, smarter S3 handling, and safer roxygenise()

Hadley Wickham

devtools 1.4 now available

devtools 1.4 adds automated vignette building and a simpler install_github() syntax

Hadley Wickham

Announcing Packrat

Packrat: a new tool for reproducible package management that keeps dependencies private to each project

Version 1.2 of devtools released

devtools 1.2 speeds up installation, supports R 3.0 vignettes, and adds sha verification for source_url()

Hadley Wickham

Version 1.0 of devtools released!

devtools 1.0 adds S4 and Rcpp support plus much better Rtools detection on Windows

Hadley Wickham

New version of devtools: 0.8

devtools 0.8 rewrites load_all() for better namespace simulation and adds support for C/C++ code and GitHub PRs

Hadley Wickham