Lorenzo Gaborini

Without leaving the tidyverse, just extract the numbers with readr::parse_number(), then use order() on the string:

Treatment <- c("Control", "Asprin 10mg/day", "Asprin 50mg/day", "Asprin 100mg/day", "Acetomycin 100mg/day" "Acetomycin 1000mg/day")

Treatment[order(readr::parse_number(Treatment))]

ggplot2 and now plumber2! What a year! 🙌

It is like saying that I know how to fix a faucet because I can call a plumber

Another instance of the Norway problem!

Most companies don’t need deep learning. They need clean joins, honest metrics, and interpretable results.
Startups often chase GPU clusters before fixing their CSVs.

#DataScience #MachineLearning #AI #RStats

How thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart Contracted AI raters describe grueling deadlines, poor pay and opacity around work to make chatbots intelligent

How thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart

www.theguardian.com/technology/2...

TIL! I've always used {tibble} enframe()/deframe()

Microsoft Powertoys: calculator, fuzzy file search, OCR text extractor, advanced paste and SO MUCH MORE with a single keystroke from anywhere! 👌

And then it's up to you to reflow anything since it's too difficult for them to open a text file. Not to mention when they require Track Changes.

Pairs up nicely with the Norway problem!

In princilple you can play around with the `cue` to avoid some of the invalidation causes. But I'm guilty too, I like to dump intermediate/cleaned results to csv and rds, even if I completely trust my pipeline. "You never know".

One solution: fix the function environment of \() by making one on the spot. Modify the loop contents with:

list_vec[[i]] <- local({
i <- i
\() paste("vec is:", i)
})

Or manually::

e <- new.env()
e$i <- i
environment(list_vec[[i]]) <- e
list_vec[[i]] <- \() paste("vec is:", i)

Proof:

rlang::fn_env(list_vec[[1]]) # <environment: R_GlobalEnv>
rlang::fn_env(list_vec[[2]]) # <environment: R_GlobalEnv>

list_vec[[1]]() # "bar"
list_vec[[2]]() # "bar"

i <- "other"
list_vec[[1]]() # "other"

The functional wants to access the name `i` when it is called later, so it looks up in the function environment. Since each iteration is executed inside R_GlobalEnv, the value of `i` is not fixed.

No it's because of the function environment of the \() functional (R_GlobalEnv), which is associated during its creation (each iteration of the loop).

More than 10 years of R, and this still caught me off guard... #rstats

vec <- c("foo", "bar")
list_vec <- list()

for (i in vec) {
list_vec[[i]] <- \() paste("vec is:", i)
}

list_vec[[1]]() # "vec is: bar"
list_vec[[2]]() # "vec is: bar"

For VSCode I liked vscode-data-preview when it worked!

Amazing! I always end up doing something that could benefit from this (a big qmd with wrangling and literate exploration — for me — and a qmd with the selected results)

Pasta is lunch/dinner soup

And some are still against NSE!

VSCode is a much more pleasant coding experience (better multi-cursors, multi-file regex search, minimap, goto anything), and the current Copilot works pretty well for oneliners! I just miss R-specific workflows (debugging, Rmd/Qmd auto-loaded params, tests) but they're slowly coming

Why the slots and the quoted variables?

Why not go full {lubridate}?
parse_date_time("May-23", "bd")

Blink twice for help

It's much easier if you can start from the rocker images! They already have the whole tidyverse preinstalled (until you need a specific version: renv+PPM+pak helps a lot but it's still lots of dependencies)

I discovered that too by reading {readr}'s code!

They are also never completely true or false — they work well inside their own (i.e. the observer's) limits. Geocentrism is "true" if you do not care about astronomy!

I hope no one is working on {err} (a spell-checker?) 🤣

Why use air.toml instead of the .editorconfig?

I routinely copy things from dplyr: there are pages for specific concepts (non-functions) with inline references across the whole package. Super useful to document bigger packages (with data dictionaries, etc.)! Also, reactable and gt for the huge list of examples!