Programming with dplyr: Renaming a column with variable using glue syntax

Question

I've read through Programming with dplyr and understand that rename() and select() use tidy selection. I'm trying to combine this with the glue syntax to create a custom function using the new double curly syntax (rlang v0.4.0), however I'm getting extra quotation marks:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

sel_var = "homeworld"

# Attempt at using (newer) double curly syntax:
starwars %>% 
  select("{{sel_var}}_old" := {{ sel_var }})
#> # A tibble: 87 x 1
#>    `"homeworld"_old`
#>    <chr>            
#>  1 Tatooine               
#> # ... with 77 more rows

# Working, but uglier (and older) bang bang syntax:
starwars %>% 
  select(!!sym(paste0(sel_var, "_old")) := {{ sel_var }})
#> # A tibble: 87 x 1
#>    homeworld_old
#>    <chr>        
#>  1 Tatooine          
#> # ... with 77 more rows

^{Created on 2021-02-16 by the reprex package (v0.3.0)}

How can I avoid the extra quotations marks in `"homeworld"_old` using the double curly {{ }} and glue := syntax? This is shown to work for summarise("mean_{{expr}}" := mean({{ expr }}), ...) in a function here.

Artem Sokolov · Accepted Answer · 2021-02-16T16:17:07.087

The {{ operator inside the glue mechanism works at the level of expressions, not strings. When an expression contains a string, the quotes (") are also a part of that same expression, which is why you see them in the output. If you convert your string to a variable name, everything should work as expected:

sel_var <- as.name("homeworld")

starwars %>% 
  select("{{sel_var}}_old" := {{ sel_var }})
# # A tibble: 87 x 1
#    homeworld_old
#    <chr>        
#  1 Tatooine     
#  2 Tatooine     
# ...

NOTE that the summarise("mean_{{expr}}" := mean({{ expr }}), ...) example you linked has the same property. For example, here's one of the functions defined in that vignette:

my_summarise5 <- function(data, mean_var, sd_var) {
  data %>% 
    summarise(
      "mean_{{mean_var}}" := mean({{ mean_var }}), 
      "sd_{{sd_var}}" := mean({{ sd_var }})
    )
}

Everything works as expected when you pass variable names to the function:

my_summarise5( mtcars, mpg, mpg )
#   mean_mpg   sd_mpg
# 1 20.09062 20.09062

However, passing strings will include " in the output, as in your case:

my_summarise5( mtcars, "mpg", "mpg" )
#   mean_"mpg" sd_"mpg"
# 1         NA       NA
# Warning messages:
# 1: In mean.default(~"mpg") :
#   argument is not numeric or logical: returning NA
# 2: In mean.default(~"mpg") :
#   argument is not numeric or logical: returning NA

Ah that makes a lot more sense now, thanks! – Alwin Feb 17 '21 at 02:50 — Alwin, Feb 17 '21 at 02:50

score 2 · Answer 2 · answered Feb 16 '21 at 10:00

2

The value inside {{}} should be unquoted to be evaluated so create the column name before using them.

Here are two ways :

Using curly-curly ({{}}).

library(dplyr)
library(rlang)

sel_var = 'homeworld'
new_col <- paste0(sel_var, '_old')

starwars %>% 
  select({{ sel_var }}) %>% 
  rename({{new_col}} := {{ sel_var }})

# A tibble: 87 x 1
#   homeworld_old
#   <chr>        
# 1 Tatooine     
# 2 Tatooine     
# 3 Naboo        
# 4 Tatooine     
# 5 Alderaan     
# 6 Tatooine     
# 7 Tatooine     
# 8 Tatooine     
# 9 Tatooine     
#10 Stewjon      
# … with 77 more rows

Using bang-bang (!!) will return the same output.

starwars %>% 
  select({{ sel_var }}) %>% 
  rename(!!new_col := {{ sel_var }})

answered Feb 16 '21 at 10:00

Ronak Shah

377,200
20
156
213

Thanks for the response, but I'm trying to avoid `paste0` and use the glue syntax. In the other dplyr examples, they show `data %>% summarise("mean_{{expr}}" := mean({{ expr }}),...` as a standard way of using the two together. I'll update my question to make it clearer – Alwin Feb 16 '21 at 10:12
@Alwin in the link that you shared they have used glue syntax when the variable is passed as unquoted and not as string. Let's take `my_summarise4` function for example from that page. `my_summarise4(mtcars, cyl)` works but `my_summarise4(mtcars, "cyl")` does not. `sel_var` that you have is a string value. – Ronak Shah Feb 16 '21 at 10:40
I would steer away from the term "(un)quoted", because there's an unfortunate ambiguity between the string quotes `"` and the expression quoting via `quote()`. When somebody says "x must be (un)quoted", it's not always clear if the are referring to `"x"` or `quote(x)`. – Artem Sokolov Feb 16 '21 at 16:33

Programming with dplyr: Renaming a column with variable using glue syntax

2 Answers2