In R and dplyr, replace multiple calls of "mutate" with a single call using "mutate&q-CodePudding

We have the following dataframe in R

# Create example dataframe
df <- data.frame(gp = c(0, 1, 0, 1), 
                 col1A = c(1, 2, 3, 4), 
                 col1B = c(5, 6, 7, 8), 
                 col2A = c(11, 12, 13, 14), 
                 col2B = c(15, 16, 17, 18),
                 col3A = c(11, 12, 13, 14), 
                 col3B = c(15, 16, 17, 18))

We are looking to apply the following logic:

df %>%
  dplyr::mutate(col1A = ifelse(gp == 0, col1B, col1A)) %>%
  dplyr::mutate(col2A = ifelse(gp == 0, col2B, col2A)) %>%
  dplyr::mutate(col3A = ifelse(gp == 0, col3B, col3A))

However we are looking to replace the 3 mutate calls with 1 call that combines mutate and across (or some other approach). Assume we have these variables in strings, so aVars = c('col1A', 'col2A', 'col3A') and bVars = c('col1B', 'col2B', 'col3B').

Is this type of consolidation possible to do? We've used mutate and across together before, but it seems more difficult to do so when using two sets of variables like we are doing with the A and the B variables here...

CodePudding user response：

We can use cur_column() and modify it (for the trailing "B") to reference both versions of each variable.

df %>%
  mutate(
    across(ends_with("A"),
           ~ if_else(gp == 0, cur_data()[[sub("A$", "B", cur_column())]], .))
  )
#   gp col1A col1B col2A col2B col3A col3B
# 1  0     5     5    15    15    15    15
# 2  1     2     6    12    16    12    16
# 3  0     7     7    17    17    17    17
# 4  1     4     8    14    18    14    18

CodePudding user response：

With dplyover::across2:

library(dplyr)
library(dplyover)
df %>% 
  mutate(across2(ends_with("A"), ends_with("B"), ~ ifelse(gp == 0, .y, .x), 
                 .names = "{xcol}"))

  gp col1A col1B col2A col2B col3A col3B
1  0     5     5    15    15    15    15
2  1     2     6    12    16    12    16
3  0     7     7    17    17    17    17
4  1     4     8    14    18    14    18

CodePudding user response：

Maybe a more reliable answer, using glue and rlang. Specify the columns in cols, the function in exprs and the names of the new columns in names(exprs):

library(glue)
library(dplyr)

cols <- paste0("col", 1:3)
exprs <- glue("ifelse(gp == 0, {cols}B, {cols}A)")
names(exprs) <- glue("{cols}A")
df %>% 
  mutate(!!!rlang::parse_exprs(exprs))

  gp col1A col1B col2A col2B col3A col3B
1  0     5     5    15    15    15    15
2  1     2     6    12    16    12    16
3  0     7     7    17    17    17    17
4  1     4     8    14    18    14    18