CODEWITHSUNDEEP
rdplyrtidyeval
MoltenLight
MoltenLight

Reputation: 72

Using multiple defused function arguments in dplyr::mutate_at

I have a simple function that maps NAs of a column of interest (here var_to_set_1) to something more specific when the variable var_x has the value "Unassigned".

The function in its current state.

library(dplyr)
dummy_data <- tribble(
  ~var_x,       ~var_to_set_1,  ~var_to_set_2,
  "A",          "Type_A",       "Subtype_A",
  "Unassigned", NA_character_,  NA_character_
)

test_fun <- function(data_in, var_to_set){

  var_to_set <- enquo(var_to_set)
  
  data_out <- data_in %>%
    mutate(!!var_to_set := if_else((var_x == "Unassigned" & is.na(!!var_to_set)),
                                    true = "Type_Unassigned",
                                    false = !!var_to_set))
  return(data_out)
}

dummy_data %>% test_fun(var_to_set_1)

So far this works well, but now I want to extend this function so that it can map NA values from two or more variables of interest. Again, the assignment should depend on the variable var_x as explained above. Obviously I could call the function twice for the different var_to_set, but I want to take advantage of tidy evaluation.

Following a similar logic, I have now tried the plural version of enquo in combination with elipsis and mutate_at to change the columns var_to_set_1 and var_to_set_2 in one step. However, I quickly realised that the seemingly simple extension to several variables was much more difficult than I thought. I began to venture into uncharted territory and some questions emerged that I was unable to answer myself.

Here is a snippet of what I tried to do.

test_fun_2 <- function(data_in, ...){
  
  var_to_set <- enquos(...)
  
  data_out <- data_in %>%
    mutate_at(.vars = !!!var_to_set,
              .funs = ~{
                if_else((var_x == "Unassigned" & is.na(.)),
                        true = "Type_Unassigned",
                        false = .)}
              )

  return(data_out)
}

dummy_data %>% test_fun_2(var_to_set_1, var_to_set_2)

The expected output should look like this:

expected_data <- tribble(
  ~var_x,       ~var_to_set_1,     ~var_to_set_2,
  "A",          "Type_A",          "Subtype_A",
  "Unassigned", "Type_Unassigned", "Type_Unassigned"
)

Btw, I use version 1.0.9 of dplyr.

Upvotes: 2

Views: 88

Answers (1)

thothal
thothal

Reputation: 20329

You want across:

test_fun <- function(data_in, ...){
  data_in %>%
    mutate(across(c(...), 
                  .fns = ~ if_else((var_x == "Unassigned" & is.na(.x)),
                                    true = "Type_Unassigned",
                                    false = .x)
                 )
           )
}

dummy_data %>% test_fun(var_to_set_1, var_to_set_2)
# # A tibble: 2 x 3
#   var_x      var_to_set_1    var_to_set_2   
#   <chr>      <chr>           <chr>          
# 1 A          Type_A          Subtype_A      
# 2 Unassigned Type_Unassigned Type_Unassigned

Update. In an earlier version I used enquos but as @Lionel was mentioning in the comments, this is not really necessary.

Upvotes: 1

Related Questions