Supplying multiple groups of variables to a function for dplyr arguments in the body

Question

Here is the data:

library(tidyverse)

data <- tibble::tribble(
  ~var1, ~var2, ~var3,  ~var4,    ~var5,
    "a",   "d",   "g",  "hello",    1L,
    "a",   "d",   "h",  "hello",    2L,
    "b",   "e",   "h",  "k",        4L,
    "b",   "e",   "h",  "k",        7L,
    "c",   "f",   "i",  "hello",    3L,
    "c",   "f",   "i",  "hello",    4L
  )

and the vectors, I want to use:

filter_var <- c("hello")
groupby_vars1 <- c("var1", "var2", "var3")
groupby_vars2 <- c("var1", "var2")
joinby_vars1 <- c("var1", "var2")
joinby_vars2 <- c("var1", "var2", "var3")

2nd & 5th, and 3rd & 4th vectors are same, but please assume they are different and retain them as different vectors.

Now I want to create a generic function where I can take data and these vectors to get the results.

my_fun <- function(data, filter_var, groupby_vars1,groupby_vars2, joinby_vars1, joinby_vars2) {

  data2 <- data %>% filter(var4 == filter_var) 

  data3 <- data2 %>%
    group_by(groupby_vars1) %>% 
    summarise(var6 = sum(var5))

  data4 <- data3 %>%
    ungroup() %>%
    group_by(groupby_vars2) %>% 
    summarise(avg = mean(var6,na.rm = T))

  data5 <- data3 %>% left_join(data4, by = joinby_vars1)

  data6 <- data %>% left_join(data5, by = joinby_vars2)
}

The problem is of supplying multiple vectors of multiple variables to a function to be used as dplyr arguments in the body. I tried looking into the http://dplyr.tidyverse.org/articles/programming.html, but could not solve the above problem.

Tung · Accepted Answer · 2018-09-28T22:27:28.867

group_by cannot take groupby_vars... strings as input. You need to use rlang::syms() to turn string vector into variables then use !!! to unquote them so that they can be evaluated inside group_by

library(tidyverse)
library(rlang)

data <- tibble::tribble(
  ~var1, ~var2, ~var3,  ~var4,    ~var5,
  "a",   "d",   "g",  "hello",    1L,
  "a",   "d",   "h",  "hello",    2L,
  "b",   "e",   "h",  "k",        4L,
  "b",   "e",   "h",  "k",        7L,
  "c",   "f",   "i",  "hello",    3L,
  "c",   "f",   "i",  "hello",    4L
)

filter_var <- c("hello")
groupby_vars1 <- c("var1", "var2", "var3")
groupby_vars2 <- c("var1", "var2")
joinby_vars1  <- c("var1", "var2")
joinby_vars2  <- c("var1", "var2", "var3")

my_fun <- function(data, filter_var, 
                   groupby_vars1, groupby_vars2, 
                   joinby_vars1,  joinby_vars2) {

  groupby_vars1 <- syms(groupby_vars1)
  groupby_vars2 <- syms(groupby_vars2)

  data2 <- data %>% 
    filter(var4 == filter_var) 

  data3 <- data2 %>%
    group_by(!!! groupby_vars1) %>% 
    summarise(var6 = sum(var5))

  data4 <- data3 %>%
    ungroup() %>%
    group_by(!!! groupby_vars2) %>% 
    summarise(avg = mean(var6, na.rm = TRUE))

  data5 <- data3 %>% 
    left_join(data4, by = joinby_vars1)

  data6 <- data %>% 
    left_join(data5, by = joinby_vars2)

  return(data6)
}

my_fun(data, filter_var, 
       groupby_vars1, groupby_vars2, 
       joinby_vars1,  joinby_vars2)

#> # A tibble: 6 x 7
#>   var1  var2  var3  var4   var5  var6   avg
#>   <chr> <chr> <chr> <chr> <int> <int> <dbl>
#> 1 a     d     g     hello     1     1   1.5
#> 2 a     d     h     hello     2     2   1.5
#> 3 b     e     h     k         4    NA  NA  
#> 4 b     e     h     k         7    NA  NA  
#> 5 c     f     i     hello     3     7   7  
#> 6 c     f     i     hello     4     7   7

Another way to do it: parse the string vector using parse_exprs outside then unquote them inside the function. See also this

my_fun2 <- function(data, filter_var, 
                   groupby_vars1, groupby_vars2, 
                   joinby_vars1,  joinby_vars2) {

  data2 <- data %>% 
    filter(var4 == filter_var) 

  data3 <- data2 %>%
    group_by(!!! groupby_vars1) %>% 
    summarise(var6 = sum(var5))

  data4 <- data3 %>%
    ungroup() %>%
    group_by(!!! groupby_vars2) %>% 
    summarise(avg = mean(var6, na.rm = TRUE))

  data5 <- data3 %>% 
    left_join(data4, by = joinby_vars1)

  data6 <- data %>% 
    left_join(data5, by = joinby_vars2)

  return(data6)
}

my_fun2(data, filter_var, 
        parse_exprs(groupby_vars1), parse_exprs(groupby_vars2), 
        joinby_vars1,  joinby_vars2) 

identical(my_fun(data, filter_var, 
                 groupby_vars1, groupby_vars2, 
                 joinby_vars1,  joinby_vars2),
          my_fun2(data, filter_var, 
                  parse_exprs(groupby_vars1), parse_exprs(groupby_vars2), 
                  joinby_vars1,  joinby_vars2))

[1] TRUE

Created on 2018-04-24 by the reprex package (v0.2.0).

Good to know an alternative solution. I guess, the first solution would prove to be more user friendly compared to the second as users won’t have to specify parse_exprs explicitly during the function call, thanks to syms unless I am missing something. — Geet, Apr 26 '18 at 01:02
The 2nd case might be helpful in cases like this https://stackoverflow.com/questions/49469982/r-using-a-string-as-an-argument-to-mutate-verb-in-dplyr/49470372 — Tung, Apr 26 '18 at 01:28
Very informative! Can you recommend a nice source or tutorial to learn these rlang functions, preferably in a simpler to understand examples? — Geet, Apr 26 '18 at 05:04
These are my favorites: [tidyeval](http://colinfay.me/tidyeval-1/) > [5min video by Hadley](https://youtu.be/nERXS3ssntw) > [multiple lags with tidy evaluation](https://purrple.cat/blog/2018/03/02/multiple-lags-with-tidy-evaluation/) — Tung, Apr 26 '18 at 06:20

Supplying multiple groups of variables to a function for dplyr arguments in the body

1 Answers1

Linked