Davis Vaughan

- Retaining names of `x`, which is different than case_when() github.com/tidyverse/ti...

- Retaining intent / obviousness of the code via argument ordering and a more expressive name for the action (important for beginners) github.com/tidyverse/ti...

I think there are 4 reasons to have replace_when(), even if it feels like a wrapper at surface level. Taken together, I think they are convincing enough:

- Type stability, as discussed github.com/tidyverse/ti...

- Native pipe support, which I think is an underrated issue github.com/tidyverse/ti...

I also don't think the named-argument-based API is so bad in forcats, since the only type involved is character, so it's possible that could continue to exist in whatever new utils we'd create there (although I do think that has some weird edge cases with `NA` as a name, so maybe not).

We are also exposing these utilities in a lower level package, vctrs, which has very few dependencies. My hope is that forcats can import this and use these new vctrs tools to build factor specific versions - unfortunately I think the tooling has to be there, this case is very specific to factors.

Is that not something like `case_when(grepl(...) ~ x, grepl(...) ~ y)`?

Honestly if there’s one example that convinces me we need replace_values(), it’s watching Hadley struggle with this without it 😂

I think of _values slightly differently (but close!). replace/recode_values() have dual APIs (see linked post). _values indicates you supply values on the LHS of the formula (or from/to), _when (case_when, replace_when) indicates you provide logical vectors on the LHS bsky.app/profile/davi...

Oh man, right_join(), rows_update(), multiple select()s! That is SCREAMING to all be replaced by a single call to replace_values()

We talk about this exact case! I think you nailed it:
- It's a large mental leap for beginners
- It requires extra steps for a full column replacement
- Getting `unmatched` right is tricky when you want extra safety
github.com/tidyverse/ti...

Using the `unmatched` argument of `recode_values()` to error on unmatched values rather than falling through to `default`

If you've found yourself in the middle of a `case_when()` going, "oh shit, did I cover all of those cases??", then I think you'll REALLY like the new `unmatched` argument of `recode_values()`!

Here we catch a `0` that wasn't matched by any Likert score code in our lookup table.

New `replace_values()` being used to collapse multiple groups into a single group within a `name` vector

If you find yourself doing `x = case_when(..., .default = x)` to "partially replace" a few values within a column, I think you'll like `replace_values()`!

For the life of me I can’t ever remember how na_if() works. Does it turn a value into NA or NA into a value? Why can’t I do na_if(x, c(-99, -98)) when I have multiple problem values?? Easily solved with:

replace_values(x, c(-99, -98) ~ NA)

Yes, replace_values() has a dual API. A from/to API and a formula API (like case_when).

So you can do either

replace_values(x, “one” ~ “1”, “two” ~ “2”)

Or

replace_values(x, from = from, to = to)

I’m convinced we need both!

The other huge limitation of named vectors is that it only works with a character lookup table - the names have to be character “1” rather than numeric 1, which is what you actually want here. `score` is numeric, not character!

I even planned to ping you directly since `replace_values()` draws a bit of inspiration from naniar, and in theory is a swiss army knife that can do all of what `tidyr::replace_na()`, `naniar::replace_na_with()`, `naniar::replace_with_na()`, `na_if()`, and `coalesce()` do (see tidyup examples)!

Recoding a `score` column using a Likert scale lookup table and a new `recode_values()` function in dplyr

If you find yourself still reaching for `dplyr::recode(x, !!!values)` or `plyr::mapvalues()`, then I think you'll like `recode_values()`!

We are looking for some community feedback on 3 new dplyr functions!

- replace_when()
- recode_values()
- replace_values()

New tools for recoding (think, lookup tables!) and replacing (think, replace `-99` with `NA` in `col`) in the tidyverse - I'm pretty excited about these!

Remember this #rstats post? I wasn't the only one talking about it & the tidyverse team was listening 😎 #databs

New #dplyr functions? They're looking for feedback!!
🤔 replace_when, recode_values, replace_values

👀 Read this:
github.com/tidyverse/ti...

🗣️ Comment on PR:
github.com/tidyverse/ti...

YouTube video by useR! Conference A first look at Positron - Julia Silge

It's the virtual day for useR! Some of the talks are already available, including this one: I always enjoy @juliasilge.com explaining things so if you are curious about the Positron IDE (it's out of beta now!), this is for you:
www.youtube.com/watch?v=aKSr... #rstats

YouTube video by useR! Conference A first look at Positron - Julia Silge

My @user-conf.bsky.social virtual talk, a first look at Positron for #rstats folks, posted while I was asleep 😴 but you can catch it here:
youtu.be/aKSrptGegeo?...

Air, an extremely fast R formatter We are thrilled to announce Air, a new R formatter.

Okay, I saw @libbyheeren.bsky.social mention Air recently but did not realize what it was until today's Data Science Hangout. Looks very cool for formatting code!

www.tidyverse.org/blog/2025/02...

The virtual useR! 2025 event is happening tomorrow (August 1). There's still time to register. Claim your spot, for free, before time runs out!

The program: user2025.r-project.org/program/virt...
Registration: user2025.r-project.org/register

#useR2025 #rstats

Today at 12pm ET! In just under an hour!

Making libcurl work in webassembly We explain how to make libcurl based applications work in webassembly without changes by tunneling all traffic over a websocket proxy.

Making libcurl work in webassembly

https://jeroen.github.io/notes/webassembly-curl/

We found a pretty neat way to make curl/httr2 work in WebR:

Oh I'm fairly certain it's a bug that RStudio shows `.fn = \(x) { body` in the outline but doesn't pick up on `.fn = \(x) body` without the braces. They should both show up, there is no semantic difference between the two functions. So I think you're relying on a bug 😆

You can read about when we auto brace anonymous functions here. Short one liners are allowed, everything else is auto braced. posit-dev.github.io/air/formatte...

What is the bookmarks issue? Can you show a screenshot or video?

air/docs/editors.qmd at main · posit-dev/air R formatter and language server. Contribute to posit-dev/air development by creating an account on GitHub.

I know nothing about Emacs but if anyone wants to do a PR to add the setup workflow to our Air website, we'd be open to it! github.com/posit-dev/ai...

For now you could copy that binary to a path that you own (not one Positron "owns"). Or work with IT to make the official installer script work to get the CLI binary the right way.

Oh I wouldn't really encourage using the bundled one in RStudio. As soon as the Air extension auto-updates itself in Positron, you'll get a new binary in a new path (like `-0.16.0/`).