recode_shadow() special missings are not accounted for by summary functions

Hi! Thanks so much for the package; it's such an important tool that the R ecosystem really needed!

I've been using recode_shadow() to handle some special missings, and while that works to change the shadow columns / update the factor levels, when I try to use functions to summarize characteristics of missingness in the dataframe, eg, add_any_miss() or miss_var_table(), it doesn't recognize these recoded special missings AS missing; it marks the first row of the dataframe as complete despite the -99 value for wind being a special missing. It might be nice if there were an option to choose whether NA aggregations distinguish between "true" / plain NA and special NAs, but if not, I think this omission could easily mislead someone about the completeness of their data.

library(naniar)
df <- tibble::tribble(
  ~wind, ~temp,
  -99,    45,
  68,    NA,
  72,    25
)

df
#> # A tibble: 3 × 2
#>    wind  temp
#>   <dbl> <dbl>
#> 1   -99    45
#> 2    68    NA
#> 3    72    25
df_recode <- df |> bind_shadow() |>
  recode_shadow(wind = .where(wind == -99 ~ "broken_machine"))

df_recode |> add_any_miss()
#> # A tibble: 3 × 5
#>    wind  temp wind_NA           temp_NA any_miss_all
#>   <dbl> <dbl> <fct>             <fct>   <chr>       
#> 1   -99    45 NA_broken_machine !NA     complete    
#> 2    68    NA !NA               NA      missing     
#> 3    72    25 !NA               !NA     complete
df_recode |> miss_var_table()
#> # A tibble: 2 × 3
#>   n_miss_in_var n_vars pct_vars
#>           <int>  <int>    <dbl>
#> 1             0      3       75
#> 2             1      1       25

^{Created on 2024-02-02 with reprex v2.1.0}```

njtierney / naniar

recode_shadow() special missings are not accounted for by summary functions #340