expand_grid with tibbles fails depending on values

[djbirke]

I am trying to create a tibble that contains all combinations of its elements, except for some nesting. In doing so, I encountered the following sensitivity of expand_grid with respect to its contents.

Is this a bug?

tidyr::expand_grid(
  dplyr::tibble(
    fruit = c("Apple", "Banana"),
    fruit_id = c("a", "b")
  ),
  dplyr::tibble(
    status_id = c("c", "d"),
    status = c("cut", "devoured")
  )
)
#> # A tibble: 4 x 4
#>   fruit  fruit_id status_id status  
#>   <chr>  <chr>    <chr>     <chr>   
#> 1 Apple  a        c         cut     
#> 2 Apple  a        d         devoured
#> 3 Banana b        c         cut     
#> 4 Banana b        d         devoured

tidyr::expand_grid(
  dplyr::tibble(
    fruit = c("Apple", "Banana"),
    fruit_id = c("a", "b")
  ),
  dplyr::tibble(
    status_id = c("c", "d"),
    status = c("cut_neatly", "devoured"),
  )
)
#> Error: Column name `dplyr::tibble(...)` must not be duplicated.
#> Use .name_repair to specify repair.

^{Created on 2021-04-27 by the reprex package (v1.0.0)}

[DavisVaughan]

Seems like a tidyr issue. So I'll move it there.

Related to:
#1092
r-lib/rlang#1116

[DavisVaughan]

A quick fix is to assign the inlined tibbles to variables, then call expand_grid() with the variable names.

library(tidyr)

df1 <- dplyr::tibble(
  fruit = c("Apple", "Banana"),
  fruit_id = c("a", "b")
)

df2 <- dplyr::tibble(
  status_id = c("c", "d"),
  status = c("cut_neatly", "devoured"),
)

expand_grid(df1, df2)
#> # A tibble: 4 x 4
#>   fruit  fruit_id status_id status    
#>   <chr>  <chr>    <chr>     <chr>     
#> 1 Apple  a        c         cut_neatly
#> 2 Apple  a        d         devoured  
#> 3 Banana b        c         cut_neatly
#> 4 Banana b        d         devoured

^{Created on 2021-04-27 by the reprex package (v1.0.0)}

Sensitivity to input comes from the fact that quos_auto_name() truncates the expressions inside the tibble() call to ... if they are long enough, and in the first example one is long enough but the other isn't, so they look different and everything works fine. In the second example, both get truncated and look the same.

library(rlang)

quos_auto_name(quos(
  dplyr::tibble(
    fruit = c("Apple", "Banana"),
    fruit_id = c("a", "b")
  ),
  dplyr::tibble(
    status_id = c("c", "d"),
    status = c("cut", "devoured")
  )
))
#> <list_of<quosure>>
#> 
#> $`dplyr::tibble(...)`
#> <quosure>
#> expr: ^dplyr::tibble(fruit = c("Apple", "Banana"), fruit_id = c("a", "b"))
#> env:  global
#> 
#> $`dplyr::tibble(status_id = c("c", "d"), status = c("cut", "devoured"))`
#> <quosure>
#> expr: ^dplyr::tibble(status_id = c("c", "d"), status = c("cut", "devoured"))
#> env:  global

quos_auto_name(quos(
  dplyr::tibble(
    fruit = c("Apple", "Banana"),
    fruit_id = c("a", "b")
  ),
  dplyr::tibble(
    status_id = c("c", "d"),
    status = c("cut_neatly", "devoured"),
  )
))
#> <list_of<quosure>>
#> 
#> $`dplyr::tibble(...)`
#> <quosure>
#> expr: ^dplyr::tibble(fruit = c("Apple", "Banana"), fruit_id = c("a", "b"))
#> env:  global
#> 
#> $`dplyr::tibble(...)`
#> <quosure>
#> expr: ^dplyr::tibble(status_id = c("c", "d"), status = c("cut_neatly",
#>           "devoured"), )
#> env:  global

^{Created on 2021-04-27 by the reprex package (v1.0.0)}