mixing constant and variable column names in dplyr::mutate/transmute

Question

I am trying to change the class of a dataframe column using dplyr. The name of the target column is contained in a variable

my_df<-data.frame(colour=c("red","blue","green"),
                  val1=as.character(c(1,12,13)),
                  val2=c(21,22,23))

target_var="val1"

After some fiddling I managed to reach my goal using standard R subsetting:

my_df %>% transmute(colour = colour,
                 !!myval := as.numeric(.[,myval]))

But I suspect that there are less complex ways of referring to the target column, which is more consistent with other 'dplyr' expressions. I have tried solving this question with the information from the "Programming with dplyr" vignette, but no luck. Could anyone point me in the right direction?

If you found that a possible answer provided you with the solution you were looking for, please consider accepting it as the answer! — David, Dec 01 '17 at 10:15

score 4 · Answer 1 · answered Dec 01 '17 at 09:35

You are looking for the mutate_at function. Your code would look like this then:

library(tidyverse) # to load dplyr and tibble

# I took the liberty to add val3 to show how you can do it with multiple variables
my_df <- data_frame(colour = c("red", "blue", "green"),
                    val1 = as.character(c(1, 12, 13)),
                    val2 = c(21, 22, 23),
                    val3 = as.character(c(1, 12, 13)))

# same here...
target_var <- c("val1", "val3")


my_df %>% 
  mutate_at(.funs = as.numeric, .vars = target_var)
#> # A tibble: 3 x 4
#>   colour  val1  val2  val3
#>    <chr> <dbl> <dbl> <dbl>
#> 1    red     1    21     1
#> 2   blue    12    22    12
#> 3  green    13    23    13

The only drawback to this method is that you end up with the full dataset (mutate) and not with the selected variables (as you would with transmutate). You could use transmutate_at, but that applies the supplied function to all selected variables.

thanks! Although it doesn't directly answer my question it is certainly very useful for my current script! — Joost Keuskamp, Dec 02 '17 at 08:21

score 3 · Accepted Answer · edited Jun 20 '20 at 09:12

3

We could use the sym to convert to symbol and then with !!

my_df %>%
    transmute(colour = colour, 
        !!target_var := as.numeric(as.character(!!rlang::sym(target_var))))
#    colour val1
#1    red    1
#2   blue   12
#3  green   13

NOTE: There the 'val1' is factor because by default stringsAsFactors = TRUE. So, we need to convert it to character and then to numeric

data

my_df<-data.frame(colour=c("red","blue","green"),
              val1=as.character(c(1,12,13)),
              val2=c(21,22,23))

target_var <- "val1"

edited Jun 20 '20 at 09:12

Community

1
1

answered Dec 01 '17 at 09:35

akrun

874,273
37
540
662

1

Thanks akrun -- this is what i've been looking for. I've posted a slightly simpler version of your solution, that does not require rlang. – Joost Keuskamp Dec 02 '17 at 08:21

Joost Keuskamp · Answer 3 · 2017-12-02T09:05:05.123

1

I've found how to do this with a minimum of conversions:

my_df<-data.frame(colour=c("red","blue","green"),
              val1=as.character(c(1,12,13)),
              val2=c(21,22,23))

target_var="val1"

my_df %>% transmute(colour = colour,
                 !!my_val := as.numeric(!!as.symbol(target_var)))

edited Dec 02 '17 at 09:05

answered Dec 02 '17 at 08:16

Joost Keuskamp

125
1
9

Be careful, you still use dplyr in this solution! (%>% and transmute, as well as !!) – David Dec 02 '17 at 08:24
You are right, I wrongly assumed that 'rlang' had to be loaded explicity. – Joost Keuskamp Dec 02 '17 at 09:02

mixing constant and variable column names in dplyr::mutate/transmute

3 Answers3

data