Problem with Nest leading to r error 4 (R code execution error)

[jestarr]

I've been getting an error message when using nest. Let's use the example from the documentation.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(tidyr)

as_tibble(iris) %>% nest(-Species)
#> Warning: All elements of `...` must be named.
#> Did you want `data = c(Sepal.Length, Sepal.Width, Petal.Length, Petal.Width)`?
#> # A tibble: 3 x 2
#>   Species              data
#>   <fct>      <list<df[,4]>>
#> 1 setosa                [4]
#> 2 versicolor            [4]
#> 3 virginica             [4]
Created on 2019-06-26 by the reprex package (v0.3.0)

If I assign this to an object and go to view, I get the following error message "r error 4 (R code execution error)".

[batpigandme]

Adding the relevant warning from View() (note: that part not run through reprex, because it's IDE-dependent)

library(dplyr)
library(tidyr)

iris2 <- as_tibble(iris) %>% nest(-Species)
#> Warning: All elements of `...` must be named.
#> Did you want `data = c(Sepal.Length, Sepal.Width, Petal.Length, Petal.Width)`?

View(iris2)
#> Error: Can't cast `x` <list_of<
#>   tbl_df<
#>     Sepal.Length: double
#>     Sepal.Width : double
#>     Petal.Length: double
#>     Petal.Width : double
#>   >
#> >> to `to` <character>.

backtrace

<error>
message: Can't cast `x` <list_of<
  tbl_df<
    Sepal.Length: double
    Sepal.Width : double
    Petal.Length: double
    Petal.Width : double
  >
>> to `to` <character>.
class:   `vctrs_error_incompatible_cast`
backtrace:
  1. (function (x, start, len, ...) ...
  7. vctrs:::vec_cast.character.default(...)
  8. vctrs::vec_default_cast(x, to, x_arg = x_arg, to_arg = to_arg)
  9. vctrs::stop_incompatible_cast(x, to, x_arg = x_arg, to_arg = to_arg)
 10. vctrs:::stop_incompatible(...)
 11. vctrs:::stop_vctrs(...)

[lionel-]

There doesn't seem to be a reasonable way of implementing as.character() for arbitrary objects contained in list_of() vectors, short of resorting to rlang::as_label(). Perhaps the IDE should take another approach for representing complex column types, e.g. using pillar if available?

[hadley]

IDE source code at https://github.com/rstudio/rstudio/blob/34c39a3321e1273ede3fe40f7c5683269e03f3a5/src/cpp/session/modules/SessionDataViewer.R#L24-L51.

[kevinushey]

Right now, we just print the deparsed representation for nested lists, but I think we'd be open to something better. It's tough because that representation is the most useful in many cases though, e.g.

library(tibble)
tbl <- tibble(x = list(list(a = 1), list(a = 2), list(a = 3)))
View(tbl)

[lionel-]

Thinking more about it, I think this default method for list_of is reasonable:

#' @export
as.character.vctrs_list_of <- function(x, ...) {
  map_chr(x, as_label)
}

This gives:

WDYT @hadley?

[hadley]

Would it be better to return vec_ptype_full()?

[lionel-]

I think it's generally better to return a single-line string per row from as.character()? It could be vec_ptype_abbr() given that list_of() cannot contain scalars.

[hadley]

We could use vec_ptype_abbr() and then which to vec_ptype_line() when it exists?

[lionel-]

Sounds good!