`as_tsibble` generates warning "Using an external vector in selections was deprecated in tidyselect 1.1.0." when passed external vectors as args

[baranovskypd]

library(tsibble)
library(tsibbledata)

# convert to basic dataframe
x <- as.data.frame(aus_retail) 

# directly pass args to as_tsibble - no warning
as_tsibble(x, key = names(x)[1:3], index = names(x)[4])

# pass args as external vectors - warning
period <- names(x)[4]
keys <- names(x)[1:3]

as_tsibble(x, key = keys, index = period)

# # returns: 
# # A tsibble: 64,532 x 5 [1M]
# # Key:       State, Industry, Series ID [152]
# ... 
# Warning message:
#   Using an external vector in selections was deprecated in tidyselect 1.1.0.
# ℹ Please use `all_of()` or `any_of()` instead.
# # Was:
# data %>% select(keys)
# 
# # Now:
# data %>% select(all_of(keys))
# 
# See <https://tidyselect.r-lib.org/reference/faq-external-vector.html>.
# This warning is displayed once every 8 hours.
# Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
# generated. 

[brookslogan]

--- For @baranovskypd: ---

This warning arises due to key and index being tidyselect expressions. In order to allow unquoted variable names in tidyselect expressions, as in some_df %>% as_tsibble(index = YearMonth), that means that there must be more typing to unambiguously refer to vectors of names like period and keys (see below).

Some more context:
This is the "lucky" case where tidyselect can detect what we meant and issue a warning (really, they'd probably want it to be an error but it's a deprecation warning due to historical reasons). But if names(x) happened to include "period" or "keys", then we'd be in the "unlucky" case where tidyselect would do something unintended (and detecting ambiguities without generating errors is more of a pain and potentially impossible without causing performance issues).

Three alternative ways that won't trigger the warning:

library(tsibble)
library(tsibbledata)

# convert to basic dataframe
x <- as.data.frame(aus_retail) 

# directly pass args to as_tsibble - no warning
reference <- as_tsibble(x, key = names(x)[1:3], index = names(x)[4])

period <- names(x)[4]
keys <- names(x)[1:3]

identical(as_tsibble(x, key = !!keys, index = !!period), reference) # TRUE
identical(as_tsibble(x, key = force(keys), index = force(period)), reference) # TRUE
library(dplyr)
identical(as_tsibble(x, key = all_of(keys), index = all_of(period)), reference) # TRUE

See the tidyselect syntax and the FAQ linked in the error message for more information.

--- For tsibble devs: ---

This seems like a bit of a documentation issue: maybe these should be more clearly marked as tidyselect + explained (although perhaps it could also be seen as an interface issue: maybe these should not be tidyselect but something else). The rlang reference says to mark tidyselect args in a special way:

@param bar <[`tidy-select`][dplyr::dplyr_tidy_select]> What `bar` does.

Here's a documentation snippet from {epiprocess} (fixed up a bit) where we've tried to make this friendlier:

#' @param .col_names <[`tidy-select`][dplyr_tidy_select]> An unquoted column
#'   name (e.g., `cases`), multiple column names (e.g., `c(cases, deaths)`),
#'   [other tidy-select expression][tidyselect::language], such as `c("cases", "deaths")` or
#'   `force(character_vector_variable_containing_column_names)`.
#'   Variable names can be used as if
#'   they were positions in the data frame, so expressions like `x:y` can be
#'   used to select a range of variables.
#'
#'   The tidy-selection renaming interface is not supported, and cannot be used
#'   to provide output column names; if you want to customize the output column
#'   names, use [`dplyr::rename`] after the slide.

[mitchelloharawild]

Thanks @brookslogan for the thorough investigation and clear description of the problem.
Would you be interested in submitting a PR for the improved documentation?