Tidyeval functions with character inputs

[davidhodge931]

Hi All,

I have a tidyeval question that is puzzling me, which I'd be grateful for any help or thoughts.

Is it possible to use this grouped_mean function with character variable inputs?

library(tidyverse)

grouped_mean <- function(data, group_var, summary_var) {
  group_var <- enquo(group_var)
  summary_var <- enquo(summary_var)
  
  data %>%
    group_by(!!group_var) %>%
    summarise(mean = mean(!!summary_var))
}

grouped_mean(mtcars, cyl, mpg) # this works
grouped_mean(mtcars, "cyl", "mpg") # but this doesn't work, and I would like it to

Ideally I want to not modify the grouped_mean function, but instead modify how the character variable input comes into the function

The reason I ask is that I have made some similar tidyeval functions for graphing and mapping that I would like to use shiny inputs

Thanks!
Dave

[raytong]

Hi @davidhodge931. You may try to use use !!as.name.

library(tidyverse)

grouped_mean <- function(data, group_var, summary_var) {
  group_var <- enquo(group_var)
  summary_var <- enquo(summary_var)
  
  data %>%
    group_by(!!group_var) %>%
    summarise(mean = mean(!!summary_var))
}

grouped_mean(mtcars, cyl, mpg) # this works
#> # A tibble: 3 x 2
#>     cyl  mean
#>   <dbl> <dbl>
#> 1     4  26.7
#> 2     6  19.7
#> 3     8  15.1
grouped_mean(mtcars, !!as.name("cyl"), !!as.name("mpg"))
#> # A tibble: 3 x 2
#>     cyl  mean
#>   <dbl> <dbl>
#> 1     4  26.7
#> 2     6  19.7
#> 3     8  15.1

^{Created on 2019-09-17 by the reprex package (v0.3.0)}

[cderv]

@davidhodge931

Ideally I want to not modify the grouped_mean function, but instead modify how the character variable input comes into the function

Not modifying your function would require to recreate symbol to pass to the function.
using !!sym() from tidyeval would work I guess

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

grouped_mean <- function(data, group_var, summary_var) {
  group_var <- enquo(group_var)
  summary_var <- enquo(summary_var)
  
  data %>%
    group_by(!!group_var) %>%
    summarise(mean = mean(!!summary_var))
}

grouped_mean(mtcars, cyl, mpg)
#> # A tibble: 3 x 2
#>     cyl  mean
#>   <dbl> <dbl>
#> 1     4  26.7
#> 2     6  19.7
#> 3     8  15.1
grouped_mean(mtcars, !!sym("cyl"), !!sym("mpg"))
#> # A tibble: 3 x 2
#>     cyl  mean
#>   <dbl> <dbl>
#> 1     4  26.7
#> 2     6  19.7
#> 3     8  15.1

^{Created on 2019-09-17 by the reprex package (v0.3.0)}

If you accept to modify the function, you could try to accept both type of input.

[davidhodge931]

@cderv

summary_var <- enquo(summary_var)

Thanks @raytong and @cderv that is really helpful, and solves out my problem!

Out of curiousity @cderv how would I modify the function to accept quoted or non-quoted variables?

[aosmith]

I'm hoping someone comes along with a recommended way for a function to take both bare symbols and strings as inputs since I haven't found anything that seems particularly straightforward.

One way, which seems pretty roundabout, is to convert everything (symbols or strings) to strings with rlang::as_name(). Then sym() can be used to convert the strings to symbols (as I said, roundabout ).

grouped_mean = function(data, group_var, summary_var) {
    group_var = sym( rlang::as_name( enquo(group_var) ) )
    summary_var = sym( rlang::as_name( enquo(summary_var) ) )

    data %>%
        group_by(!!group_var) %>%
        summarise(mean = mean(!!summary_var))
}

Then using either strings or symbols as inputs returns the same output.

grouped_mean(mtcars, "cyl", "mpg")
# A tibble: 3 x 2
#    cyl  mean
#  <dbl> <dbl>
#1     4  26.7
#2     6  19.7
#3     8  15.1
> grouped_mean(mtcars, cyl, mpg)
# A tibble: 3 x 2
#    cyl  mean
#  <dbl> <dbl>
#1     4  26.7
#2     6  19.7
#3     8  15.1

Edited to add another approach
I thought there should be some way to use a conditional statement within the function. This would be based on checking whether the input was a symbol or not and then use enquo() or sym() based on that. It looks like rlang::quo_is_symbol() is the predicate symbol that can do the checking.

grouped_mean = function(data, group_var, summary_var) {
    if(rlang::quo_is_symbol( enquo(group_var) ) ) {
        group_var = enquo(group_var)
    } else { 
        group_var = sym(group_var) 
    }
    
    if(rlang::quo_is_symbol( enquo(summary_var) ) ) {
        summary_var = enquo(summary_var)
    } else {
        summary_var = sym(summary_var)
    }
    
    data %>%
        group_by(!!group_var) %>%
        summarise(mean = mean(!!summary_var))
}