Dirk Eddelbuettel

website: https://dirk.eddelbuettel.com
github: https://github.com/eddelbuettel

Last Notes

2026-05-31 13:05:45 UTC

- reply

I tend to write myself compact little wrappers in the `littler` package (see its `inst/examples/` dir) whenever I need something, and Troy has been busy creating an entire R package toolchain for development that is way more lightweight and nimble than the existing flavors, you could check that out. One blog post is listed below, the packages have likely been updated since.
https://cornball.ai/posts/tinyverse-development-toolchain/

2025-12-31 19:07:40 UTC

- reply

I am also a little non-plus'ed about the accrued time from forks. But the good news is that it should be ficticious money: for open source projects you usually do not get charged.

2025-11-11 22:17:03 UTC

duckdb-mlpack 0.0.4: Added random forest and logistic regression
Bringing mlpack machine learning to duckdb SQL
https://dirk.eddelbuettel.com/blog/2025/11/11#duckdb-mlpack_0.0.4
Adds random forest and (regularized) logistic regression, reworks interface, adds new state variable for verbosity, expands documentation and adds unit tests
https://files.mastodon.social/media_attachments/files/115/533/354/884/819/371/original/da98a37d79931b00.png

2025-11-05 23:15:32 UTC

- reply

I really believe you have to have lived for a while with CRAN in "production" or other "downstream use cases" (and either at scale. ideally) to understand how this "everything at HEAD" monorepo works. The recent post by Julie T. (link below) expressed it well. People with different repo experiences cannot understand how we survive with #rstats. "Oh well" and _tant pis_.
https://jtibs.substack.com/p/if-all-the-world-were-a-monorepo

2025-09-15 19:52:11 UTC

- reply

Congratulations, you just reinvented the use case for Docker, a decade later. 😜

2025-04-18 21:15:10 UTC

- reply

Welcome to seeing the light!
Use code 'at @npub1vl7…epgw', use packages that take stability and backward compatibility seriously, and when you must 'freeze' something build a Docker container (ideally off @npub1vl7…epgw, easily and reliably and quickly with eg #r2u). #rstats

2024-09-12 15:21:00 UTC

- reply

You are extrapolating here and missing the forest for that one tree: you can use `data.table::rbindlist()` _on a list of data.frame objects_ and it will still glue them together faster than the alternatives _and_ it also offers you a data.frame as a return object.
What's not to like? 😉

2024-09-12 14:04:16 UTC

- reply

Take a look at `data.table::rbindlist()` which runs circles around alternatives and excels in all list-based, possibly `lapply((` et al drive iterative builds.