CODEWITHSUNDEEP
rdplyrtidyversenon-standard-evaluation
MatSchu
MatSchu

Reputation: 419

Pass a named list or character value to dplyr’s mutate to create multiple columns in a dataframe

My goal is to create multiple dataframe columns as in the following example.

# Goal
mtcars %>% 
  rownames_to_column("model") %>% 
  mutate(MAZDA = case_when(grepl('^Mazda', model) ~ 1, TRUE ~ 0),
         MERC  = case_when(grepl('^Merc',  model) ~ 1, TRUE ~ 0),
         VOLVO = case_when(grepl('^Volvo', model) ~ 1, TRUE ~ 0) )

However, in my real-world application, I cannot be sure, how many columns are to be created. The number of case-when conditions may vary. This is why I want to either create a named list or a character value that contains the conditions based on the input data. However, one piece is missing: How can I pass a named list or a character value to dplyr’s mutate verb? Both of the following examples do not work.

# Named list
condition1 <- list("case_when(grepl('^Mazda', model) ~ 1, TRUE ~ 0)",
                   "case_when(grepl('^Merc',  model) ~ 1, TRUE ~ 0)",
                   "case_when(grepl('^Volvo', model) ~ 1, TRUE ~ 0)")

names(condition1) <- c("MAZDA", "MERC", "VOLVO")

result1 <- mtcars %>% mutate(!!!condition1)

# Character string
condition2 <- 
"MAZDA = case_when(grepl('^Mazda', model) ~ 1, TRUE ~ 0),
 MERC  = case_when(grepl('^Merc',  model) ~ 1, TRUE ~ 0),
 VOLVO = case_when(grepl('^Volvo', model) ~ 1, TRUE ~ 0)"

result2 <- mtcars %>% mutate(eval(parse(text = condition2)))

How can I make them work? Is there an alternative base R approach to solve the problem?

Upvotes: 0

Views: 74

Answers (1)

G. Grothendieck
G. Grothendieck

Reputation: 269586

1) Assuming that the conditions are of the form shown we can simplify the input to just the Names and then create the columns using map_dfr and bind that to the original data frame.

library(dplyr)
library(purrr)

Names <- c("Mazda", "Merc", "Volvo")  # input names

mtcars %>% 
  bind_cols(
    rownames(.) %>%
    { map_dfr(set_names(Names, toupper), \(x) +startsWith(., x)) }
  )

2) If the forms can differ creat4 a list of functions as shown and then apply them:

L2 <- list(MAZDA = \(x) +startsWith(x, "Mazda"), 
           MERC = \(x) +startsWith(x, "Merc"), 
           VOLVO = \(x) +startsWith(x, "Volvo"))

mtcars %>%
  bind_cols(rownames(.) %>% { map_dfr(L2, \(f) f(.)) })

3) or if for some reason a character vector is needed then try the following (although to me (1) or (2) is preferable):

L3 <- list(MAZDA = '\\(x) +startsWith(x, "Mazda")', 
           MERC = '\\(x) +startsWith(x, "Merc")', 
           VOLVO = '\\(x) +startsWith(x, "Volvo")')

mtcars %>%
  bind_cols(rownames(.) %>% { map_dfr(L3, \(f) eval(str2lang(f))(.)) })

Upvotes: 2

Related Questions