Use {palmerpenguins} dataset in arrange() examples

[KittJonathan]

I added examples using the penguins dataset in the arrange() documentation

• added {palmerpenguins} to suggests in the dplyr DESCRIPTION file

[DavisVaughan]

We really appreciate the PR! I do not currently see any reason to switch away from mtcars to penguins in this example. The story is a little complicated since R >=4.5 already has penguins built in, meaning we shadow it by loading this package and users would get a possibly confusing message

I don't think switching to penguins right now is worth the mental overhead of having to remember that there are now 2 datasets with the same name. Also, the column names of both are unfortunately different, making it even more confusing.

I do however like your additional inline comments, and how you added an extra example of arranging by just 1 column. I would accept a PR that kept mtcars but added those inline comments and the extra example, if you'd like to do that.

[yutannihilation]

I do not currently see any reason to switch away from mtcars to penguins in this example

A bit drifted topic: On the other hand, I believe there are reasons for dplyr (and other tidyverse packages) to switch from iris to penguins because I believe that's the very reason penguins was added.

ggplot2 removed iris example code 5 years ago. I was hoping other packages would follow, but it had not happen so far. I think it's a good time to consider migrating to penguins.

That said, I have no idea how the replacement should be done. basepenguins might help, but I'm not fully sure if it won't make another confusion.

[KittJonathan]

I get your point @DavisVaughan. I can indeed work on using mtcars. But I personally do think it would be nice having examples using palmerpenguins across as many tidyverse functions as possible, given the popularity of the dataset in tutorials and courses.
@hadley clarified by email the tidyverse version policy, supporting the current and previous four versions of R.
Would it be worth while considering switching to palmerpenguins (or basepenguins) examples when four versions of R have been released?

[hadley]

I think this change is worth it, but you need to (a) put palmerpenguins in imports, not suggests (since otherwise you need to protect all the examples) and (b) load the data using data() instead of by attaching the package. I'd also suggest keeping the existing pipe based examples primary, since I don't think we want to get into that discussion now.

I can't believe that R core changed the variable names, making the whole substitution more complicated.

[hadley]

Thanks for continuing to work on this! It's really important to get the details 100% correct in this prototype, so I appreciate you bearing with us as we deliver very detailed feedback.

[KittJonathan]

Thanks for continuing to work on this! It's really important to get the details 100% correct in this prototype, so I appreciate you bearing with us as we deliver very detailed feedback.

Thanks for your help, I appreciate you taking the time to give very detailed feedback :-)