`across(.names =)` evaluates `"{col}"` as if it was `"{.col}"`

[arnaudgallou]

It doesn't seem to be the expected behavior according to the documentation.

col <- "foo"

tibble::tibble(
  x = 1:3,
  y = 4:6
) |> 
  dplyr::mutate(
    dplyr::across(c(x, y), \(.x) .x * 2, .names = "{col}_{.col}")
  )
#> # A tibble: 3 × 4
#>       x     y   x_x   y_y
#>   <int> <int> <dbl> <dbl>
#> 1     1     4     2     8
#> 2     2     5     4    10
#> 3     3     6     6    12

packageVersion("dplyr")
#> [1] '1.1.4'

^{Created on 2025-06-09 with reprex v2.1.1}

[DavisVaughan]

I think it is undocumented but "technically expected" on our end

dplyr/R/across.R

Lines 414 to 419 in be3e3a0

	# TODO: we can make these bindings louder later
	env_bind_active(
	glue_mask,
	col = function() glue_mask$.col,
	fn = function() glue_mask$.fn
	)

We switched to the dot prefix but kept the non-dot versions in this PR
#5448

It looks like dplyr 1.0.0 originally used {col}, and the idea here was that we switched to {.col} and updated all documentation to use {.col}, but kept the old way to avoid breaking anything. At this point we should:

• Document {col} and {fn} as technically allowed but discouraged
• Consider starting to deprecate {col} and {fn} usage, as the TODO suggests