Disclaimer: I have no affiliation with Microsoft Corp. or Revolution Analytics.

For the n-th time in x years, submitting an R package to CRAN ended up like comedy. This time for one anecdotal note (a kind of warning), whereas the previous accepted version of ESGtoolkit has had, for 5 years, 12 warnings and notes combined. No error, nothing’s broken. And I’m not even comparing it to some other packages’ results.

package-results-yeah

Literally no online repo (PyPi, conda forge, submit a Julia package, etc.) works this way these days, with a censor sending emails back and forth to a student receiving an examination. The validation process is automated (sure, you can sometimes contact someone). It’s not as if all of this was about controlling the calculation of EBITDA. It’s just code. And perfection is unattainable. One. Warning. Man. Sighs.

It seems to me like, under the hood and after years and years of observation and curious, repeated subliminal (and not so subliminal) threats, it isn’t just about code on CRAN, and more about a lot of obscure politics and private interests. Even if it was about code, censor, why not submitting a pull request/issue to my GitHub repo directly as everyone does these days, instead of continuously shooting shots for nothing? It also happens that, when you submit something there, your package is kind of confiscated and you can never (ever) take it back. So definitely, and again, think twice (or more) before submitting.

I know exactly what to do to remove the note (xlab and ylab passed to ... in a plotting function, and causing the “not-error” no visible global function definition for ‘xlab’). Which wasn’t suggested by censor. But that’s when I was reminded that in 2020, there are thousands of ways to circumvent reactionaries (you should use GitHub/Gitlab search sometimes, your whole world and certainties will fall apart) on the internet, and of this gem of a package I once heard about: miniCRAN.

How does miniCRAN feels to me? Just like when I want to create a virtual environment in Python with only packages that I’d like to use, isolated from other packages (well, for actually using R in a virtual environment, I don’t know yet). And there’s more to it than that, as shown in this presentation, including the invaluable ability to use a CRAN-like system of R packages behind a firewall, in an entreprise – or wherever – ecosystem. With miniCRAN, I was able to create an online repo just like what I’m used to, with only the set of packages I’d like to use.

And for someone who’s been using R for 15 + years (I wonder why it’s studied and taught in university though), it was delightful to be able to do this for the first time:

# my mirror of packages created with miniCRAN
repo <- "https://techtonique.github.io/r-techtonique-forge/"

# list of packages currently available in the repo for Linux/macOS and Windows
print(miniCRAN::pkgAvail(repos = repo, type = "source"))
print(miniCRAN::pkgAvail(repos = repo, type = "win.binary"))

# installing `pkg_name` from my mirror using the old faithful `install.packages`
install.packages("pkg_name", repos=repo, type="source") 

# using the package as usual 
library(pkg_name)