R dplyr mutate multiple columns using custom function to create new column

Question

I'd like to create new column in data.frame using dplyr::mutate using custom function which argument is vector of data.frame's column names but I got following output:

customFun <- function(col.vec) {
  paste0(gsub("\\s", "_", col.vec), collapse = "-")
}

df <- data.frame(A = c("x 1", "x", "x w"), B = c("E", "D", "2 w"), stringsAsFactors = FALSE)

df %>%
   mutate(C = customFun(c(A, B)))
    A   B                 C
1 x 1   E x_1-x-x_w-E-D-2_w
2   x   D x_1-x-x_w-E-D-2_w
3 x w 2 w x_1-x-x_w-E-D-2_w

instead of:

data.table::data.table(df)[, C := customFun(c(A, B)), by = .(A, B)]
     A   B       C
1: x 1   E   x_1-E
2:   x   D     x-D
3: x w 2 w x_w-2_w

It can be achieved in many ways, but I'm interested in dplyr solution only. Thank you for your help.

Answer 1

We can use map and lift_dl . We first map over each col.vec (notice I have used a list instead of a vector as input, since c flattens any vector elements, while list doesn't) and apply gsub . Then the list output is fed into paste . Since paste takes ... , we can use purrr::lift_dl to lift it's input domain from ... to list type:

library(dplyr)
library(purrr)

customFun <- function(col.vec) {
  map(col.vec, ~gsub("\\s", "_", .x)) %>%
    lift_dl(paste, sep = "-")()
}

df %>%
  mutate(C = customFun(list(A, B)))

or with ... as input:

customFun <- function(...) {
  col.vec <- list(...)
  map(col.vec, ~gsub("\\s", "_", .x)) %>%
    lift_dl(paste, sep = "-")()
}

df %>%
  mutate(C = customFun(A, B))

Output:

    A   B       C
1 x 1   E   x_1-E
2   x   D     x-D
3 x w 2 w x_w-2_w

Answer 2

Why use by=.(..) in your data.table solution? If at all you have two rows with exactly similar values, then these will be collapsed into one. You need to modify your customFun . It is not correct the way it is:

library(tidyverse)
customFun = function(data) invoke(paste, data.frame(gsub('\\s+', '_', as.matrix(data))), sep='-')

df %>% 
    mutate(c = customFun(.))

    A   B       C
1 x 1   E   x_1-E
2   x   D     x-D
3 x w 2 w x_w-2_w

You can replace the invoke with do.call or even lift etc.

Your function is not doing exactly what you want. Read the comment above

Answer 3

Just add rowwise before your mutate so only each row's A & B values are used in paste , rather than the vectors of all rows.

library(dplyr)

df %>%
  rowwise() %>%
  mutate(C = customFun(c(A, B)))
#> Source: local data frame [3 x 3]
#> Groups: <by row>
#> 
#> # A tibble: 3 x 3
#>   A     B     C      
#>   <chr> <chr> <chr>  
#> 1 x 1   E     x_1-E  
#> 2 x     D     x-D    
#> 3 x w   2 w   x_w-2_w

^{Created on 2019-02-05 by the reprex package (v0.2.1)}