pmap_dfr() does not work well with nested tibbles of one row

[dllazarov]

When it comes to nested dataframes, pmap_dfr exercises some weird behaviour. I don't know if this is an issue or a desired effect, but worth noting.

tibble(a = 1, b = 2, c = 3) %>% 
    pmap_dfr(
        function(...){
            df <- tibble(...)
            df
        }
    )

# Returns
      a     b     c
  <dbl> <dbl> <dbl>
1     1     2     3

tibble(a = 1, b = 2, c = 3, d = tibble(a = 1, b = 2, c = 3)) %>% 
    pmap_dfr(
        function(...){
            df <- tibble(...)
            df
        }
    )

# returns 
      a     b     c     d
  <dbl> <dbl> <dbl> <dbl>
1     1     2     3     1
2     1     2     3     2
3     1     2     3     3

[romainfrancois]

I believe pmap() and friends should use vec_slice() instead of [[. Moving this to purrr where pmap() lives.

library(tidyverse)

ping <- local({ 
  value <- 0L
  function() {
    value <<- value + 1
    value
  }
})

tibble(a = 1, b = 2, c = 3, d = tibble(a = 5, b = 6, c = 7)) %>% 
  pmap_dfr(
    function(...){
      print(match.call())
      data.frame(x = ping(), d = list(...)$d)
    }
  )
#> .f(a = .l[[1L]][[1L]], b = .l[[2L]][[1L]], c = .l[[3L]][[1L]], 
#>     d = .l[[4L]][[i]])
#> .f(a = .l[[1L]][[1L]], b = .l[[2L]][[1L]], c = .l[[3L]][[1L]], 
#>     d = .l[[4L]][[i]])
#> .f(a = .l[[1L]][[1L]], b = .l[[2L]][[1L]], c = .l[[3L]][[1L]], 
#>     d = .l[[4L]][[i]])
#>   x d
#> 1 1 5
#> 2 2 6
#> 3 3 7

^{Created on 2021-09-17 by the reprex package (v2.0.0)}

[DavisVaughan]

I'm not sure this one is a bug. It uses the same principles as this, i.e. that this data frame is treated as a list with 3 elements

library(purrr)
library(tibble)

map(tibble(a = 1, b = 2, c = 3), identity)
#> $a
#> [1] 1
#> 
#> $b
#> [1] 2
#> 
#> $c
#> [1] 3

Unless we had edition 2 of purrr that used vctrs semantics for data frames, I don't think we can change this