R: paste string as code or function argument within dplyr

Question

How do i paste a string within a dplyr function i.e. summarise( ) and run it as a code?

library('tidyverse')
df <- tibble(ID = c('a','a','b','c','c','e','e','f','g','g'),
              var1 = floor(runif(10, min=0, max=100)),
              var2 = floor(runif(10, min=0, max=100)),
              var3 = floor(runif(10, min=0, max=100)),
              var4 = floor(runif(10, min=0, max=100))
              )

sample data

> df
# A tibble: 10 x 5
   ID     var1  var2  var3  var4
   <chr> <dbl> <dbl> <dbl> <dbl>
 1 a        82     4    21    32
 2 a        90    34    12    51
 3 b        67    77    69    32
 4 c        56     3    96    76
 5 c        38     2    46    79
 6 e        34    91    12    12
 7 e        49    16    38    31
 8 f        34     1    76    82
 9 g        95    84    54    70
10 g        13    53    65    79

Replace this

df %>% 
  group_by(ID) %>% 
  summarise(var1 = sum(var1),
            var2 = sum(var2),
            var3 = sum(var3))

With this

#Define character string vector to replace command line
sum_var <- select(df,starts_with('var')) %>% names()
sum_var_str <- paste0(sum_var," = sum(",sum_var,")")
sum_var_str <- str_c(sum_var_str, collapse = ", ")
> sum_var
[1] "var1" "var2" "var3" "var4"
> sum_var_str
[1] "var1 = sum(var1), var2 = sum(var2), var3 = sum(var3), var4 = sum(var4)"

#run code with character string
df %>% 
  group_by(ID) %>% 
  summarise(sum_var_str) #this line doesn't work

I have tried

summarise(!!parse_quosure(sum_var_str))
summarise(parse(text =sum_var_str))

What am i missing?

thanks,

#--------------- In case you question why am i doing this? ---------

I want to use multidplyr, and it has yet to have anything for summarise_at. i have hundreds if not thousands, so the summarise_at is necessary, but unfortunately, not available in multidplyr.

looking for an alternative to work around it.

library('multidplyr')
cluster <- new_cluster(5)

#works
df %>% 
  group_by(ID) %>% 
  #partition(cluster) %>% 
  summarise_at(.vars = vars(starts_with('var')),sum) 
  #collect()

#works
df %>% 
  group_by(ID) %>% 
  partition(cluster) %>% 
  summarise(var1 = sum(var1),
            var2 = sum(var2),
            var3 = sum(var3)) %>% 
  collect()

#doesnt works
df %>% 
  group_by(ID) %>% 
  partition(cluster) %>%
  summarise_at(.vars = vars(starts_with('var')),sum)  %>% 
  collect()

Error in UseMethod("group_vars") : 
  no applicable method for 'group_vars' applied to an object of class "multidplyr_party_df"

#I want to see if this works
df %>% 
  group_by(ID) %>% 
  partition(cluster) %>%
  summarise(parse(text =sum_var_str)) %>% #incorrect line of code
  collect()

Why do you want to parse strings and not do `df %>% group_by(ID) %>% summarise(across(starts_with('var'), sum))` ? — Ronak Shah, Jul 25 '20 at 12:12
Is this just for academic interest? It seems like a really bad idea. Whatever problem it is supposed to solve could almost certainly be done in a better way. — Allan Cameron, Jul 25 '20 at 12:13
I've added additional information to why i'm doing this. @RonakShah — Choc_waffles, Jul 25 '20 at 12:33
I can't test this but does reshaping work? `df %>% pivot_longer(cols = -ID) %>% group_by(ID, name) %>% summarise(value = sum(value)) %>% pivot_wider()` — Ronak Shah, Jul 25 '20 at 12:43

G. Grothendieck · Answer 1 · 2020-07-25T19:49:40.710

Parse and eval the entire statement, not just the arguments:

df %>% 
  group_by(ID) %>% 
  { eval(parse(text = sprintf("summarise(., %s, .groups = 'drop')", sum_var_str))) }

giving:

# A tibble: 6 x 5
  ID     var1  var2  var3  var4
  <chr> <dbl> <dbl> <dbl> <dbl>
1 a       106   140   157   186
2 b        40    67    64    69
3 c       182    67   164    81
4 e        56   113   124   122
5 f        89     4    59    21
6 g       100   127    42    54

Note

set.seed is needed to make this reproducible. Also, we have reduced the dependencies to just dplyr.

library(dplyr)

set.seed(123)

df <- tibble(ID = c('a','a','b','c','c','e','e','f','g','g'),
              var1 = floor(runif(10, min=0, max=100)),
              var2 = floor(runif(10, min=0, max=100)),
              var3 = floor(runif(10, min=0, max=100)),
              var4 = floor(runif(10, min=0, max=100))
              )
sum_var_str <- toString(sprintf("%s = sum(%s)", names(df)[-1], names(df)[-1]))

MrPmosh · Accepted Answer · 2020-08-12T23:21:37.660

1

I encountered the same issue and noticed my dplyr version was updated to dplyr 1.0.0.

You can resolve/workaround this issue by going back to dplyr 0.8.5 then your summarise_at should work fine with a party_df object.

dplyr 0.8.5's group_vars methods work with objects of class multidplyr_party_df.

All the _at,_all,_if verbs have been superseded by the across() function in dplyr 1.0.0.

You can read more about how to use the new methods here: dplyr 1.0.0 Changelog

edited Aug 12 '20 at 23:21

answered Aug 12 '20 at 16:29

MrPmosh

100
12

Thanks to your comment @MrPmosh, i have posted a solution using multidplyr and summarise from dplyr version >= 1. https://stackoverflow.com/questions/63088754/r-multidplyr-for-summarise-at-work-around – Choc_waffles Aug 26 '20 at 11:59

R: paste string as code or function argument within dplyr

2 Answers2

Note

Linked