James Balamuta

Screenshot of the details extension quickstart documentation showing three methods to create collapsible blocks in Quarto: using the summary attribute, using a heading, and using a summary div. Each method shows the code syntax and its rendered output as expandable disclosure widgets.

Hexagonal sticker logo for the details Quarto extension. Orange background with a stylized illustration of three accordion-style collapsible content panels in cream and brown, with the word "details" in white text at the bottom.

Released a #Quarto extension for collapsible content blocks.

Why? The {details} are inside.

📚 quarto.thecoatlessprofessor.com/details/
💻 github.com/coatless-qua...

Turning another year older tomorrow and celebrating by releasing Quarto extensions into the wild. Gift to me, gift to you. And yes, if you've been peeking at my GitHub, you already know what's coming. Act surprised!

Ripper extension hex logo with Halloween-inspired design. Black hexagonal badge featuring a white .qmd document on the left with red blood splatter effect, and multiple file extension labels on the right (.R, .py, .jl, .js, .lua, .sh, .ts, .md) resembling torn paper. The word 'RIPPER' appears in large red distressed lettering at the bottom, with additional blood drip effects throughout.

Screenshot demonstrating the ripper Quarto extension workflow. Shows the ripper logo (a dark house silhouette with red text), documentation page for Default Configuration, example R and Python code blocks in the rendered document, and two code editor windows displaying the generated script files (qripper-default-config.R and qripper-default-config.py) with extracted code. A red arrow connects the 'Script files' section to the generated output files.

{ripper} dissects your #Quarto documents, extracting code blocks by language and leaving executable scripts behind.

Supports #rstats, #python, #julia, and 13 other victims.

For those who never stopped celebrating Halloween.

📚 quarto.thecoatlessprofessor.com/ripper/
💻 github.com/coatless-qua...

One does not simply stay away from social media. I return with: R packages, portable R, Shiny/shinylive apps, Electron integrations, Python packages, Quarto extensions, LLMs, and blog posts. The code has been reforged. The roadmap is full.

A screenshot of a statistics tutorial made with R, quarto and webR

Still fascinated with #webR and its potential for teaching stats

Here students can play (even from their phones) and find out themselves how the p-value depends on sample size

pakillo.github.io/LM-GLM-GLMM-... #rstats

GitHub - coatless-quarto/toggle: Toggle Code Output in Quarto Code Cells Toggle Code Output in Quarto Code Cells. Contribute to coatless-quarto/toggle development by creating an account on GitHub.

@coatless.bsky.social also did that few months ago using another approach based on JavaScript.

"Pick your poison" 🙃

github.com/coatless-qua...

Quarto has a way to collapse code, but what if you want to collapse code OUTPUT?

The amazing @mickael.canouil.fr created an extension that does just that!!!

Check it out! github.com/mcanouil/qua...

I did; but at the end of the day it's just eye candy for a README that I touch once every 3 years. The license section at the end of the README with "GPL (>=2)" is more than sufficient. In the end, I'm trying to pay it forward to my future self on simplifying CRAN maintenance tasks pings.

Aye, I noticed the timeout persisting despite the changes. So, to get ~10 packages update; I'm just dropping all badges in the README sin the R CMD check one.

Screenshot of a code diff showing line 10 being updated. The removed line (red) shows a GPL 2.0 license URL pointing to gnu.org/licenses/gpl-2.0.html, while the added line (green) shows the same URL updated to point to gnu.org/licenses/old-licenses/gpl-2.0.html, indicating the license documentation has been moved to an 'old-licenses' directory.

Nothing makes you feel vintage like your license getting moved to the retirement home directory. GNU put GPL 2.0 in /old-licenses/ causing #rstats packages to throw URL warnings. Somewhere a CRAN maintainer is rubbing their hands together, ready to trigger 847 email requests...

Flying to Istanbul to meet my SO's family felt like nervous excitement. Flying to California felt like possibility. Flying back home for a funeral feels like gravity remembering how to work. Now Illinois feels like living in a house where all the furniture is made of memories.

A screenshot showing R 4.5.1 working under RStudio 2025.9.0.387 on macOS Tahoe (26.0.0) with a call to compiled code requiring gfortran through RcppArmadillo.

A screenshot showing R 4.5.1 working under Positron 2025.9.0 (Universal) on macOS Tahoe (26.0.0) with a call to compiled code requiring gfortran through RcppArmadillo.

macOS 26 Tahoe + R 4.5.1: Keeps #rstats moving along on #macOS. Upgrade confidently, just remember to update Xcode CLI afterwards if using stan & compiled code.

Terminal session showing the download and execution of an R AppImage from GitHub releases. The commands show wget downloading the ARM64 AppImage built on Ubuntu 24.04, making it executable with chmod +x, and then launching it to run R statistical analysis including a linear regression model with income vs age data, demonstrating the workflow from GitHub release to running statistical computations.

VS Code showing the successful completion of an R AppImage build process. The terminal displays the build summary indicating a 76MB minimal R 4.5.1 AppImage for ARM64 architecture with package installation disabled (immutable). The output shows usage examples and confirms the build completed successfully, with R actually running at the bottom showing the standard R startup message and executing system environment commands, demonstrating the fully functional portable R installation.

VS Code terminal showing R 4.5.1 running from an AppImage on ARM64 architecture. The file explorer shows the typical AppImage directory structure with usr/bin, usr/lib, and usr/share folders, demonstrating the self-contained portable nature of the R installation.

R that travels light on #Linux: Portable R AppImages.

Now working everywhere: your Ubuntu, friend's Fedora, cousin's Arch setup (btw)

No sudo, no tears, just base R science ✨

(package support coming soon!)

#RStats #AppImage #DataScience

RcppEnsmallen 0.2.22.1.2 on CRAN: Upstream Armadillo Changes

blog.thecoatlessprofessor.com/software-rel...

#rstats #rcpp #rcppensmallen #ensmallen

AppImage is Problematic under Dev Mode with AppArmor on Ubuntu 24.04 / 23.10 · Issue #6 · coatless-rpkg/shinyelectron After building the electron app on Ubuntu 24.04 ARM, we're getting: Running npm run electron > My-App-Title@1.0.0 electron > electron . [77996:0908/150845.938417:FATAL:sandbox/linux/suid/client/set...

Ubuntu 24.04 dev preview has some AppArmor issues during development (github.com/coatless-rpk...) but the built AppImage works perfectly.

Linux support 🐧

Same #rshiny → desktop workflow on #Fedora 42 aarch64:

shinyelectron::export() → #rshinylive → #AppImage → portable #Linux app

Plot twist: no more "go buy yourself a real computer" moments - you get zero #rstats dependencies too! The condescending Unix users have won this round.

Core Shiny Application (Shinylive Layer): 61.9 MB

This represents the base Shiny application code and dependencies

Platform-Specific Packaging Overhead (ARM64):

Windows: 300.1 MB
macOS: 264.1 MB
Linux: 304.9 MB

Total Application Size by Platform:

Windows: 362 MB
macOS: 326.4 MB
Linux: 366.8 MB

{flexdashboard} will not work as that requires a genuine R shiny backend. The rest depend on capabilities of webR and what assets are included in r-shinylive.

For persistence, electron does allow writing to disk in the user directory. This isn't supported under the VFS for shinylive variant though

{golem} should work with some creativeness, e.g. place app.R in the package-level directory:

pkgload::load_all(export_all = FALSE, helpers = FALSE, attach_testthat = FALSE)
options( "golem.app.prod" = TRUE)
mygolemapp::run_app()

This side steps the custom WASM compile of the golem app.

Eh, I've tended to stay away from VMware post-broadcom acquisition. Appreciate the note though...

Correct, we're not embedding genuine R (like prior attempts). We're limited by webR/Shinylive capabilities. If the package compiles under webR, you're set!

As of webR v0.5.5, making HTTP requests was addressed by George. But, genuine R embedding is on the feature list.

Agreed. More coming soon...

Should work out of the box on Windows 10. I only have a Windows 11 license to test with.

By request: Same #rshiny app → native #electron desktop app workflow now on #Windows11

shinyelectron::export() → #rshinylive → installer → app.exe

[Worth the Parallels license to record this from my Mac]

Still no #rstats dependencies for end users. Living the cross-platform life (virtually).

It’s straightforward under shinylive variants at the cost of needing to be WebAssembly compliant + lowers the security risk considerably (VFS instead of FS access).

Still working on genuine embedded R. Likely possible with outsourcing installer packaging on different OS to GH actions.

JJ Allaire & Charles Teague (of Posit fame) launched Meridian Labs, a nonprofit dedicated to frontier AI research and evaluation tools.

Early focus: LLM evals + visualization & SWE Agent tooling for Inspect AI.

meridianlabs.ai | github.com/meridianlabs-ai

#AI #Research

Yes, the goal is to create a standalone Shiny application that mirrors a traditional application. For now, the experimental build supports R shiny apps via `r-shinylive`. This will likely be available for PyShiny via `py-shinylive`; but, that's a ways off and maybe a Python package later ;)

Prototyping #rshiny apps to native #electron desktop apps:

shinyelectron::export() → #rshinylive conversion → .dmg → Native Mac app

Zero #rstats dependencies for end users! Early days but promising 👀

Screenshot of JupyterLab launcher interface showing the main dashboard with two sections: "Notebook" and "Console". Each section displays kernel options for Python 3.9 (py39) and R 4.2.2 with their respective icons. The left sidebar contains a file browser showing "Storage" and "Temporary" folders, both modified "a minute ago", with a search filter box at the top. The interface includes typical JupyterLab navigation elements and is running on sciserver.org as shown in the browser tab.

Screenshot of RStudio Server interface showing an "About RStudio Server" dialog box. The dialog displays version 2022.07.2 Build 576 with copyright information from 2009-2022 RStudio, PBC. Behind the dialog, the RStudio interface is visible with the console panel showing R version 4.2.1 startup information, a terminal tab, and various menu options (File, Edit, Code, View, Plots, Session, Build, Debug, Profile, Tools, Help). The right side shows a Tutorial panel and file browser. The browser tab indicates this is running on sciserver.org.

I think that covers SciServer Essentials v1.0. There are 4 different versions. The latest has software at #rstats 4.2.2 and #Python 3.9 inside of #JupyterLab.

The R+RStudio image has R 4.2.1 and RStudio Build 2022.07.2.

Maybe raise a ticket to update the underlying suites?

Screenshot of SciServer homepage showing a blue header with the SciServer logo and "Collaborative data-driven science" tagline. The main section features a colorful nebula background with white text reading "The Science Platform" and "A collaborative environment for server-side analysis with extremely large datasets." Below are six feature boxes with icons for About, Hosted Datasets, Compute Images, Science Domains, Education, and Help, each containing brief descriptions. The page includes a "Login to SciServer" button and NSF funding acknowledgment at the bottom.

Screenshot of SciServer's Education page with the same blue header design. The main content shows "Education" as the page title, followed by explanatory text about using SciServer to teach data science. The page includes sections on "How to Teach with SciServer" with detailed instructions, and an "Examples" section featuring astronomy course examples from University of St Andrews and University of North Carolina Asheville, with links to GitHub repositories containing sample notebooks.

TIL about SciServer.org: Free scientific computing platform with persistent accounts (#NSF-funded). All the power of #JupyterHub without the setup hassle. Focus on building great teaching + research notebooks, not infrastructure! #OpenScience