Type stability when a df-col has 0 columns but 1 row

[DavisVaughan]

library(tidyr)

# df-col with 1 row and 0 cols
df <- tibble(x = tibble(.rows = 1))

# should be 1 row and 0 cols
unpack(df, x)
#> # A tibble: 0 × 0

^{Created on 2021-11-03 by the reprex package (v2.0.1)}

[DavisVaughan]

And in expand_grid() due to flatten_nested()

tidyr::expand_grid(tibble::tibble(.rows = 1))
#> # A tibble: 0 × 0

^{Created on 2021-11-03 by the reprex package (v2.0.1)}

Best solution is probably to make flatten_at() return a data frame itself with the right number of rows (that can be promoted to a tibble as needed)