across(all_of()) doesn't see variables outside it's own environment

[vovoch]

I faced the strange behavior in mutate(across(all_of())) combination of functions. Objects defined previously cannot be used inside of this construction. Two examples:

reduce(mget(ls(pattern = "^mcmc_stat_")), full_join, by = "ParameterMCMC") %>%
    select(ParameterMCMC, Median, LCL95, HCL95, Mean, SD) %>%
    nest(data = !all_of("ParameterMCMC")) %>%
    mutate(retaining_cols = map(ParameterMCMC, ~global_settings$mcmc_param_presets %>%
                                  filter(INPUT == .x) %>% pull(COLUMNS) %>% unlist()),
           data = map2(retaining_cols, data, ~.y %>% mutate(across(!all_of(.x), ~NA))))

#Error: object '.x' is not found

reduce(mget(ls(pattern = "^mcmc_stat_")), full_join, by = "ParameterMCMC") %>%
    select(ParameterMCMC, Median, LCL95, HCL95, Mean, SD) %>%
    nest(data = !all_of("ParameterMCMC")) %>%
    mutate(data = map2(ParameterMCMC, data, ~{
      retaining_cols <- global_settings$mcmc_param_presets %>%
        filter(INPUT == .x) %>% pull(COLUMNS) %>% unlist()
      .y %>% mutate(across(!all_of(retaining_cols), ~NA))
    }))

#Error: object 'retaining_cols' is not found

But objects from the R_GlobalEnv can be used. For example, the "<<-" operator in the following code resolves this problem and works in the expected way:

reduce(mget(ls(pattern = "^mcmc_stat_")), full_join, by = "ParameterMCMC") %>%
    select(ParameterMCMC, Median, LCL95, HCL95, Mean, SD) %>%
    nest(data = !all_of("ParameterMCMC")) %>%
    mutate(data = map2(ParameterMCMC, data, ~{
      retaining_cols <<- global_settings$mcmc_param_presets %>%
        filter(INPUT == .x) %>% pull(COLUMNS) %>% unlist()
      .y %>% mutate(across(!all_of(retaining_cols), ~NA))
    }))

[lionel-]

It would really help if you could reduce the example to a bare bones reprex with:

• the least amount of function calls
• the least amount of control flow (e.g. is reduce() essential to your example?)
• the least amount of symbols and texts (but still informative names)
• some typographic air around the essential constructs to help the reader focus on the flow of code

Also please make it reproducible with https://github.com/tidyverse/reprex

[hadley]

Can you please provide a minimal reprex (reproducible example)? The goal of a reprex is to make it as easy as possible for me to recreate your problem so that I can fix it: please help me help you!

If you've never heard of a reprex before, start by reading "What is a reprex", and follow the advice further down the page. Please make sure your reprex is created with the reprex package as it gives nicely formatted output and avoids a number of common pitfalls.

[vovoch]

I have the full set of statistics for all parameters after an MCMC simulation (data_tbl).
My intention is to keep only the essential statistics for each parameter based on the special tibble (preset_tbl).
It is possible only using the global assignment (res_ok_tbl), but doesn't work without it (res_error_tbl).

library(tidyverse)

data_tbl <- tibble(
  PARAMETER = c("Rate",   "mu"),
  MEDIAN    = c("28.01%", "-0.94"),
  LCL95     = c("8.15%",  "-2.42"),
  HCL95     = c("62.97%", "0.53"),
  MEAN      = c("29.93%", "-0.95"),
  SD        = c("14.02%", "0.74")
)

preset_tbl <- tribble(
  ~PARAMETER, ~STAT_COLUMNS,
  "Rate",    c("MEDIAN", "LCL95", "HCL95"),
  "mu",      c("MEAN", "SD")
)

join_tbl <- left_join(data_tbl, preset_tbl, by = "PARAMETER")

res_error_tbl <- join_tbl %>%
  nest(data = !all_of(c("PARAMETER", "STAT_COLUMNS"))) %>%
  mutate(data = map2(data, STAT_COLUMNS, ~{
    .x %>% mutate(across(!all_of(.y), ~NA_character_))
  })) %>% unnest(cols = data) %>% select(-STAT_COLUMNS)
#> Error: Problem with `mutate()` input `data`.
#> x Problem with `mutate()` input `..1`.
#> x object '.y' not found
#> i Input `..1` is `across(!all_of(.y), ~NA_character_)`.
#> i Input `data` is `map2(...)`.

res_ok_tbl <- join_tbl %>%
  nest(data = !all_of(c("PARAMETER", "STAT_COLUMNS"))) %>%
  mutate(data = map2(data, STAT_COLUMNS, ~{
    y_global <<- .y
    .x %>% mutate(across(!all_of(y_global), ~NA_character_))
  })) %>% unnest(cols = data) %>% select(-STAT_COLUMNS)

# Input and desirable output of the target code

join_tbl
#> # A tibble: 2 x 7
#>   PARAMETER MEDIAN LCL95 HCL95  MEAN   SD     STAT_COLUMNS
#>   <chr>     <chr>  <chr> <chr>  <chr>  <chr>  <list>      
#> 1 Rate      28.01% 8.15% 62.97% 29.93% 14.02% <chr [3]>   
#> 2 mu        -0.94  -2.42 0.53   -0.95  0.74   <chr [2]>

res_ok_tbl
#> # A tibble: 2 x 6
#>   PARAMETER MEDIAN LCL95 HCL95  MEAN  SD   
#>   <chr>     <chr>  <chr> <chr>  <chr> <chr>
#> 1 Rate      28.01% 8.15% 62.97% <NA>  <NA> 
#> 2 mu        <NA>   <NA>  <NA>   -0.95 0.74

^{Created on 2020-08-30 by the reprex package (v0.3.0)}

[lionel-]

Thank you for the reproducible example. Here is a minimal version:

library(tidyverse)

df <- tibble(
  vars = list("foo"),
  data = list(data.frame(foo = 1))
)

map2(df$data, df$vars, ~ mutate(.x, across(all_of(.y), ~ NA)))
#> [[1]]
#>   foo
#> 1  NA

df %>% mutate(data = map2(data, vars, ~ {
  .x %>% mutate(across(all_of(.y), ~ NA))
}))
#> Error: Problem with `mutate()` input `data`.
#> ✖ Problem with `mutate()` input `..1`.
#> ✖ object '.y' not found
#> ℹ Input `..1` is `across(all_of(.y), ~NA)`.
#> ℹ Input `data` is `map2(...)`.

[lionel-]

From #5475