Hidden NULL-column from tibble::new_tibble() causes error in vctrs::vec_rbind()

[sda030]

Not sure whether to place issue here or at tibble, feel free to move.
I am aware that tibble::new_tibble() is for advanced "only for clean inputs" use, and also that there is perhaps some use for this when creating empty list columns, etc. However, found this a bit surprising. P.S. Context for this is that I wanted to start optimizing my code based on a recent post (https://www.tidyverse.org/blog/2023/04/performant-packages/). Thought I could insert an
list(b = if(condition) "insertme") # else NULL => drop column b.

library(tibble)
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
x<-list(a=1, b=NULL) %>% tibble::new_tibble()
x
#> # A tibble: 1 × 2
#>       a
#>   <dbl>
#> 1     1
str(x)
#> tibble [1 × 2] (S3: tbl_df/tbl/data.frame)
#>  $ a: num 1
#>  $ b: NULL
y <- tibble::tibble(a=1, b=NULL)
str(y)
#> tibble [1 × 1] (S3: tbl_df/tbl/data.frame)
#>  $ a: num 1

## Errors with bind_rows() 
dplyr::bind_rows(x)
#> Error in `vec_rbind()`:
#> ! Unexpected `NULL`.
#> ℹ In file 'slice.c' at line 322.
#> ℹ This is an internal error that was detected in the vctrs package.
#>   Please report it at <https://github.com/r-lib/vctrs/issues> with a reprex (<https://tidyverse.org/help/>) and the full backtrace.
#> Backtrace:
#>     ▆
#>  1. ├─dplyr::bind_rows(x)
#>  2. │ └─vctrs::vec_rbind(!!!dots, .names_to = .id, .error_call = current_env())
#>  3. └─rlang:::stop_internal_c_lib(...)
#>  4.   └─rlang::abort(message, call = call, .internal = TRUE, .frame = frame)

^{Created on 2023-04-22 with reprex v2.0.2}

[DavisVaughan]

I think this is expected behavior everywhere:

• new_tibble() won't do any of the inputs checks
• vec_rbind() detects that something is weird and errors (much better than a crash). It gives an "internal" error but that isn't too far off, because it really does mean something fairly low level and internal has gone wrong

I think you'd be happy with using vctrs::df_list() before new_tibble(). It recycles inputs and drops NULLs automatically, and is still very fast.

vctrs::df_list(x = 1, y = NULL)
#> $x
#> [1] 1