Using case_when with multiple vectors

Question

I am trying to use case_when to modify/mutate a column based on two separate inputs. One that used to create the LHS logical and the respective input value on the RHS. An example is provided below.

library(dplyr)
library(purrr)
library(tibble)

df <- tibble(var = paste0(rep("var", 10), 1:10),
                 label = c("label1", "label2", rep(NA, 7), "label10"))

match_var <- paste0(rep("var", 7), 3:9)
new_labels <- paste0(rep("add_this_label", 7), 3:9)

df %>% 
  mutate(test = map2(match_var , new_labels,
                     ~case_when(
                       var == .x ~ .y,
                       TRUE ~ label
                     )
  ))

I think the issue is that within case_when everything is evaluated as expression but I'm not completely sure. One can manually type out all 7 lines within case_when but my application requires me to accomplish this when the vectors match_vars and new_labels are very long - making manual typing of case_when infeasible.

df %>% 
  mutate(label = case_when(
    var == match_var[1] ~ new_labels[1],
    var == match_var[2] ~ new_labels[2],
    var == match_var[3] ~ new_labels[3],
    var == match_var[4] ~ new_labels[4],
    var == match_var[5] ~ new_labels[5],
    var == match_var[6] ~ new_labels[6],
    var == match_var[7] ~ new_labels[7],
    TRUE ~ label
  ))

EDIT: the desired result can be accomplished using a for loop but now I want to know if is this possible using case_when and map2_* function?

for (i in seq_along(match_var)) {
  df$label <- ifelse(df$var == match_var[i], new_labels[i], df$label)
}

akrun · Accepted Answer · 2019-03-28T21:28:33.820

We create a named vector and use that to match the the values in 'var' so as to change the NA elements to 'new_labels'

library(tibble)
library(dplyr)
df %>%
    mutate(label = case_when(is.na(label) ~ 
                       deframe(tibble(match_var, new_labels))[var], 
         TRUE ~ label))
# A tibble: 10 x 2
#   var   label          
#   <chr> <chr>          
# 1 var1  label1         
# 2 var2  label2         
# 3 var3  add_this_label3
# 4 var4  add_this_label4
# 5 var5  add_this_label5
# 6 var6  add_this_label6
# 7 var7  add_this_label7
# 8 var8  add_this_label8
# 9 var9  add_this_label9
#10 var10 label10

NOTE: Instead of using deframe, the named vector can be created with setNames as well

this is great! if you don’t mind me asking, could you explain how named vectors work within ‘case_when’? Is it related to NSE? — EJJ, Mar 28 '19 at 21:56

score 1 · Answer 2 · answered Mar 28 '19 at 21:31

You can join the new labels to the data frame and fill in with the old label as necessary.

library("tidyverse")

df <- tibble(var = paste0(rep("var", 10), 1:10),
             label = c("label1", "label2", rep(NA, 7), "label10"))

match_var <- paste0(rep("var", 7), 3:9)
new_label <- paste0(rep("add_this_label", 7), 3:9)

new_labels <-  tibble(match_var, new_label)

df %>%
  left_join(new_labels, by = c("var" = "match_var")) %>%
  mutate(new_label = if_else(is.na(new_label), label, new_label))
#> # A tibble: 10 x 3
#>    var   label   new_label      
#>    <chr> <chr>   <chr>          
#>  1 var1  label1  label1         
#>  2 var2  label2  label2         
#>  3 var3  <NA>    add_this_label3
#>  4 var4  <NA>    add_this_label4
#>  5 var5  <NA>    add_this_label5
#>  6 var6  <NA>    add_this_label6
#>  7 var7  <NA>    add_this_label7
#>  8 var8  <NA>    add_this_label8
#>  9 var9  <NA>    add_this_label9
#> 10 var10 label10 label10

^{Created on 2019-03-28 by the reprex package (v0.2.1)}

This is a creative approach using joins I didn’t consider. I might opt for @akrun solution just because I want to avoid joins due to performance but I will test that tomorrow — EJJ, Mar 28 '19 at 21:58

score 1 · Answer 3 · answered Sep 30 '19 at 18:00

Since you're comparing the == condition, this can also be done with dplyr::recode using a named vector (note the need for unquote splicing !!!):

df %>%
    mutate(
        label = recode(
            var,
            !!!setNames(new_labels, match_var),
            .default = label
        )
    )

Using case_when with multiple vectors

3 Answers3

Linked