rix, my package that leverages #Nix to provide reproducible data science environments for #RStats (and #python and #julia) is in the top 10 starred ropensci packages!
that’s pretty cool
for my R rixpress package, I was discussing with one of the julia on nix maintainers, and there were some issues with proper dependency detection that are now resolved. I don’t know about plots specifically, but what I did test, worked flawlessly! docs.ropensci.org/rixpress/art...
In T, pipelines are first class, and mixing #Python and #Rstats nodes is easy! Passing objects to and from R or Python is handled seamlessly with built-in serializers, with the option for users to provide their own!
github.com/b-rodrigues/...
actually no, I just let LLMs do it
I'm begging AI companies to use Nix, how useful it would be if agents simply used the flake's repository! only works reliably with local agents
and Italian complain about pineapple on pizza
I think this is the GOTY
open source contributions will stop until further notice
tstats-project.org/index.html
guess what, data frames in T use arrow under the hood :) I'll have to see how to be able to do 0 copy between the languages, but not sure it's going to work, as each node runs in its own isolated nix build sandbox
the language github.com/b-rodrigues/...
a package: github.com/b-rodrigues/...
a pipeline: github.com/b-rodrigues/...
So I wanted pipelines to be first class since the beginning, but I realised that the added value would really come from being able to easily run R or Python as well, in a controlled manner, not just T code. So I'm going to focus on this and integration of the R and Python serialisers into T.
Polyglot pipelines are the future :D
In T, pipelines are first-class, and it's possible to run R or Python code. Right the pipeline code, with some nodes running #RStats code, and other #Python code.
T will have the opposite of surprises: errors as first class objects
#rstats 4.5.3 "Reassured Reassurer" scheduled for March 11. Full schedule on developer.r-project.org (or the svn if you're impatient.) This should be the wrap-up release for the 4.5 series.
the underlying issue is not using tools like targets, I agree. Functions vs pipes is not the right way to look at it
A nicer print.data.frame method showing column types, as well as a subset of rows. Inspired by data.table's print method.
I think the main issue is that many people, quite reasonably tbf, don't like the default base data.frame print method...
But this is easy to override! gist.github.com/grantmcdermo...
The language I'm working on, T, a reproducibility- and pipeline-first DSL for Data Science, has now a basic packaging system. Say hello to `hello_t`, the very first package for T!
github.com/b-rodrigues/...
In the large: Mortality in France, 1816-2016.
github.com/b-rodrigues/...
It's coming :)
join me to create the next programming language for data science: powered by #nix, #ocaml and #arrow, heavily inspired by #RStats tidyverse and designed to facilitate collaboration with LLMs!
no idea I just thought it would be useful, afaik T is the first language with this
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
there might be no point in developing yet another programming language in this day and age, but it sure is fun!
github.com/b-rodrigues/...
