Pass multiple calling arguments to a formal argument in dplyr custom function without using “…”

问题

To make a custom function flexible to receiving one or more calling arguments per formal argument I currently rely on "...":

library(dplyr)

foo <- function(data, ..., dv){
  groups <- enquos(...)
  dv <- enquo(dv)
  data %>% 
    group_by(!!!groups) %>% 
    summarise(group_mean = mean(!!dv))
}

mtcars %>% foo(am, dv = mpg)
mtcars %>% foo(vs, am, dv = mpg)

But "..." obscures the logic of the function, and it could not be used in a custom function with 2 or more formal arguments requiring multiple calling arguments.

Is there a way to write the above function to utilize a formal argument (e.g., "groups") rather than "..." that can receive a single vector name or a vector of vector names as its argument(s)? Something like:

foo <- function(data, groups, dv){
  groups <- enquos(groups)
  dv <- enquo(dv)

  data %>% 
    group_by(!!!groups) %>% 
    summarise(group_mean = mean(!!dv))
}

# Failing code
mtcars %>% foo(groups = c(vs, am), dv = mpg)

Note that this code would work, but require user to remember to use quos() in function body:

foo <- function(data, groups, dv){
  dv <- enquo(dv)

  data %>% 
    group_by(!!!groups) %>% 
    summarise(group_mean = mean(!!dv))
}

mtcars %>% foo(groups = quos(vs, am), dv = mpg)

I'd love to rely on enquos() in body of function instead.

回答1:

We could place the ... at the end

foo <- function(data,  dv, ...){
   groups <- enquos(...)
   dv <- enquo(dv)
   data %>% 
     group_by(!!!groups) %>% 
     summarise(group_mean = mean(!!dv))
  }

If we want to pass a vector of 'group', then one option is group_by_at

foo <- function(data, groups, dv){
  dv <- enquo(dv)

  data %>% 
     group_by_at(vars(groups)) %>% 
     summarise(group_mean = mean(!!dv))
  }

mtcars %>% 
    foo(groups = c("vs", "am"), dv = mpg)
# A tibble: 4 x 3
# Groups:   vs [?]
#     vs    am group_mean
#  <dbl> <dbl>      <dbl>
#1     0     0       15.0
#2     0     1       19.8
#3     1     0       20.7
#4     1     1       28.4

One option if we want to pass unquoted expression with c would be to convert it to expression and then evaluate it

foo <- function(data, groups, dv){

 groups <- as.list(rlang::enexpr(groups))[-1]
 dv <- enquo(dv)
   data %>% 
      group_by(!!! groups) %>% 
      summarise(group_mean = mean(!!dv))
 }

mtcars %>% 
      foo(groups = c(vs, am), dv = mpg)
# A tibble: 4 x 3
# Groups:   vs [?]
#     vs    am group_mean
#  <dbl> <dbl>      <dbl>
#1     0     0       15.0
#2     0     1       19.8
#3     1     0       20.7
#4     1     1       28.4

Or as @Joe mentioned in the comments, enquo should also work with group_by_at

foo <- function(data, groups, dv){
   dv <- enquo(dv) 
   groups <- enquos(groups) 
   data %>% 
        group_by_at(vars(!!!groups)) %>% 
        summarise(group_mean = mean(!!dv))
   } 

mtcars %>% 
     foo(groups = c(vs, am), dv = mpg)
# A tibble: 4 x 3
# Groups:   vs [?]
#     vs    am group_mean
#  <dbl> <dbl>      <dbl>
#1     0     0       15.0
#2     0     1       19.8
#3     1     0       20.7
#4     1     1       28.4

来源：https://stackoverflow.com/questions/53160709/pass-multiple-calling-arguments-to-a-formal-argument-in-dplyr-custom-function-wi

标签

dplyr

quasiquotes