Adding dplyr::select.list() to override utils::select.list()

[ColinFay]

Following the Twitter conversation from here.

As utils::select.list() is registered as an S3, calling dplyr::select() on a list calls select.list() without informing the user, who is faced with an interactive prompt asking them to do a selection. As far as I have understood this behavior is not intended.

I think this misuse can arise if you intend to write map_df() %>% select() and you end up writing map() %>% select().

Here's a suggested solution:

• having a message to the user that they are redirected to utils::select.list()

#' @export
select.list <- function(.data, ...) {
  message("Calling `select()` on a list, redirecting to `utils::select.list()`")
  utils::select.list(choices = .data)
}

> list(a = 12) %>% dplyr::select()
Calling `select()` on a list, redirecting to `utils::select.list()`

1: 12

Selection: 1
$a
[1] 12

> list(a = 12) %>% utils::select.list()

1: 12

Selection: 1
$a
[1] 12

Note: related to #2817

[romainfrancois]

🤔 I'd rather have it abort. select.list() is not a method in the first place, signatures don't match, ...

if the user wants to call select.list() they they should call select.list() not select().

[ColinFay]

That was also my first idea, but

select.list <- function(.data, ...) {
  abort("`select()` doesn't handle lists.")
}

makes select.list() fail 😕

> list(a = 12) %>% dplyr::select()
Error: `dplyr::select()` doesn't handle lists.
Call `rlang::last_error()` to see a backtrace

> list(a = 12) %>% select.list()
Error: `dplyr::select()` doesn't handle lists.
Call `rlang::last_error()` to see a backtrace

> list(a = 12) %>% utils::select.list()

1: 12

Selection: 1
$a
[1] 12

(But maybe there's a workaround?)

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/