Should `one_call_pipe_linter()` apply only to `%>%`?

[IndrajeetPatil]

I can see how as.character(x) can be preferable over x %>% as.character(), but I don't think the alternatives for other pipes work as well.

Consider %<>% (e.g.)

This example lints:

library(lintr)
lint(text = "data %<>% mutate(n.obs = n)", linters = one_call_pipe_linter())
#> <text>:1:2: warning: [one_call_pipe_linter] Expressions with only a single call shouldn't use pipe %<>%.
#>  data %<>% mutate(n.obs = n)
#>  ^~~~~~~~~~~~~~~~~~~~~~~~~~~

^{Created on 2023-11-21 with reprex v2.0.2}

To avoid this, I instead need to do data <- mutate(data, n.obs = n), and avoiding repetition of data is the entire raison d'etre of %<>%. So when {lintr} says one shouldn't use this pipe only with a single call, it's basically saying don't use this pipe for the very thing for which it was built. That's not the case for %>%/|>. I hope I am being cogent here.

I could imagine a similar argument for other pipes.

---

So I am wondering if the following should be changed only to cover %>%/|>:

lintr/R/one_call_pipe_linter.R

Line 48 in 209e250

(//SPECIAL[{pipes_cond}] | //PIPE)[

[MichaelChirico]

Hmm I think it makes sense to drop %<>%, not sure about the other pipes. E.g. %$%

mtcars %$% cyl

Should really be mtcars$cyl.

And %T>%, there's really no need to use it unless it's part of a longer pipeline AIUI.

%!>% is identical to %>% for the purposes here.

So it's really only %<>% to worry about.

(Of course, the style guide also prohibits %<>%, but your point about %<>% + one_call_pipe_linter() applies regardless of whether a user adheres to the style guide)

[IndrajeetPatil]

Okay, fair enough. We can just carve out an exception for %<>% then.