vec_rbind(<table with numeric values>)

[romainfrancois]

library(vctrs)

vec_rbind(
  table(c(1, 2)) / 2, 
  table(c(1, 2)) / 2
)
#> Error: Internal error in `vec_proxy_assign_opts()`: `proxy` of type `integer` incompatible with `value` proxy of type `double`.
#> Backtrace:
#>     █
#>  1. └─vctrs::vec_rbind(table(c(1, 2))/2, table(c(1, 2))/2)
#>  2.   └─(function () ...

^{Created on 2020-07-16 by the reprex package (v0.3.0.9001)}

This is the root cause of keyATM/keyATM#145, mentioned in tidyverse/dplyr#5376.

In keyATM/keyATM#145, I propose a solution that works around it by essentially cancelling the "tableness" of the object as I presume it's not useful. I'm not sure we would want to cover objects like:

str(table(c(1, 2)) / 2)
#>  'table' num [1:2(1d)] 0.5 0.5
#>  - attr(*, "dimnames")=List of 1
#>   ..$ : chr [1:2] "1" "2"

^{Created on 2020-07-16 by the reprex package (v0.3.0.9001)}

but the error message was not helpful in keyATM/keyATM#145

[DavisVaughan]

I just learned that summary() creates subclassed "table"s, which can be numeric, so it definitely isn't just used for integer contingency tables by base R, even though the docs state:

table() returns a contingency table, an object of class "table", an array of integer values.

x <- summary(mtcars$mpg)

class(x)
#> [1] "summaryDefault" "table"

x
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>   10.40   15.43   19.20   20.09   22.80   33.90

^{Created on 2020-07-20 by the reprex package (v0.3.0)}

It seems to be more like a wrapper type around integer / numeric arrays