Should unnest_wider() default to some separator for unnamed entries?

[hadley]

I can't imagine that this behaviour is ever useful:

library(tidyr)

df <- tribble(
  ~x, ~y,
  1, c(11, 12, 13),
  2, 21
)
df |> unnest_wider(y)
#> New names:
#> New names:
#> • `` -> `...1`
#> • `` -> `...2`
#> • `` -> `...3`
#> # A tibble: 2 × 4
#>       x  ...1  ...2  ...3
#>   <dbl> <dbl> <dbl> <dbl>
#> 1     1    11    12    13
#> 2     2    21    NA    NA

^{Created on 2022-06-06 by the reprex package (v2.0.1)}

What if it was instead:

df |> unnest_wider(y)
#> Warning: 
#> * Using unnest_wider() with column containing unnamed data
#> i Use `names_sep = "_"` to suppress this warning and control the separator.
#> # A tibble: 2 × 4
#>       x   y_1   y_2   y_3
#>   <dbl> <dbl> <dbl> <dbl>
#> 1     1    11    12    13
#> 2     2    21    NA    NA

?

[hadley]

And could it recommend names_sep instead of (as well as?) names_repair here?

library(tidyr)

df <- tibble(x = 1, y = list(list(x = 2)))
df |> unnest_wider(y)
#> Error in `unpack()`:
#> ! Names must be unique.
#> ✖ These names are duplicated:
#>   * "x" at locations 1 and 2.
#> ℹ Use argument `names_repair` to specify repair strategy.

^{Created on 2022-06-06 by the reprex package (v2.0.1)}

[hadley]

I tried to do the first one, and it's a bit more complicated than I thought. You need something like this in unnest_wider()

  internal_names <- map_lgl(data[cols], function(x) all(map_lgl(x, ~ length(.x) == 0 || !is.null(names(.x)))))
  if (is.null(names_sep) && any(!internal_names)) {
    warn("List-column lacks internal names, setting `names_sep = '_'")
    names_sep <- "_"
  }

But that fails to account for df-cols.

[hadley]

Need to do the same thing for separate_wider_*()