new package alert! {tidychain} is a #rstats packaged inspired by the below authors experience in showing how an excel file was changed / manipulated by looking at the underlying xml files to prove fraud in research
datacolada.org/109package
usrbinr.codeberg.page/tidychain/
bracketeer is now in CRAN!!
#rstats Can the people complaining about R pipes not being "functional" enough please go hang out in the SQL forums? If any language needs help with ease of debugging and interpreting it's SQL.
Thank you and please.
I'm not sure where the threshold is for roundtripping the data for a single operation. But there is a substantial speedup even on smaller datasets when the data is kept on gpu and repeated operations are run, like hypothesis testing.
Agree completely. This is why I implemented glimpse style printing as the default for tbl_gpus in cuplyr
Thanks man! Really appreciate it
Trying to stay neutral in the #rstats flame wars but my side projects are:
* model tournament structures using |>
* running tibbles on gpus
Packages pipe-first API where you chain stages and control who advances. Build complex tournament dynamics with zero dependencies beyond base R — lightweight enough for Shiny internals.
Try the World Cup 2026 simulation:
colab.research.google.com/github/bbthe...
Started building {bracketeer} because I needed to run a pétanque tournament for friends and couldn't believe this wasn't solved already.
Turns out tournament modeling is harder than it first looks.
bbtheo.github.io/bracketeer/
#rstats
Want to try it without installing anything? This notebook simulates a full 48-team FIFA World Cup 2026 in your browser. 12 groups, 104 matches, custom qualification rules, all the way to a champion.
colab.research.google.com/github/bbtheo/bracketeer/blob/main/notebooks/world_cup_2026.ipynb
Pub quiz. Office ping pong. 10,000 simulated World Cups. A Champions League bracket for your fantasy league.
tournament(teams) |>
swiss("open", rounds = 5) |>
single_elim("playoffs", take = top_n(8))
bracketeer 0.1.0 tournament modeling in base #Rstats
github.com/bbtheo/bracketeer
Not really as it is completely reliant on cuda and the cudf library which are sadly locked to Nvidia gpus.
DuckDB is an amazing tool that I also daily drive! I just wanted to see if I could put my GPU to work and it's been a fun experiment. Works better than I expected.
cuplyr version 0.1.0 is now out!
A GPU-accelerated dplyr backend for R, powered by RAPIDS cuDF.
Write familiar tidyverse code, execute on GPU. Lazy eval with AST optimization.
In my benchmarks 60x faster than dplyr on 50M rows.
github.com/bbtheo/cuplyr
#rstats #cuda #DataScience
tstats-project.org
New post: the design behind secretbase, a 40KB streaming hash package I maintain.
The entire source package is smaller than most hex sticker PNGs.
Hash large objects with constant memory. Powers targets pipelines. Plus CBOR encoding comes to R.
shikokuchuo.net/posts/28-int...
#RStats #OpenSource
The first data science book that has a chapter on monads reproducible-data-science.dev
Learn how to build robust #DataScience pipelines with #RStats, #Python , #Julia and #Nix !
As alway it is quite iteresting when working with C++ dependecies. But I use Pixi as a package manager which has made it relatively painless.
What if your dplyr pipelines ran on GPU?
That's what I built with cuplyr! A CUDA-powered backend for #rstats data manipulation. Looking for testers and feedback!
github.com/bbtheo/cuplyr
