using a composite to call a variable dynmically in custom build function with dplyr (!!paste0, {{}}, as.name(), eval(parse(text=)

Question

This is an extension of the following questions: (1), (2) and also asked in the comments to (2) by Mario Reutter.

library(dplyr, tidyverse)
string <- c("car", "train", 'bike', 'plain')
speed1 <- runif(4, min = 0, max = 10000)
speed2 <- runif(4, min = 0, max = 10000)
n1  <- sample(1:100, 4)
n1_plus  <- sample(1:100, 4)
n1_minus <- sample(1:100, 4)
n2  <- sample(1:100, 4)
df <- data.frame(string, speed1, speed2, n1, n2, n1_plus, n1_minus)

Thanks to akrun's answer I can build the following function:

my_fun <- function(dataf, V1, V2){
dataf %>%
dplyr::mutate("{{V1}}_{{V2}}" := paste0(format({{V1}}, big.mark   = ",") ,
  '\n(' , format({{V2}}, big.mark   = ",") , ')'))}

df<-df%>%my_fun(speed1, n1)

to create a new variable with a composite name as defined by"{{V1}}_{{V2}}" :=.

However, how were I to call a composite variable name on the right hand side of the equation? E.g. substituting format({{V2}}, big.mark = ",") with something like format('{{V2}}_plus', big.mark = ","). I tried (not working):

my_fun <- function(dataf, V1, V2){
dataf %>%
dplyr::mutate("{{V1}}_{{V2}}_plus" := paste0(format({{V1}}, big.mark   = ",") ,
  '\n(' , format('{{V2}}_plus', big.mark   = ",") , ')'))}

df<-df%>%my_fun(speed1, n1)

Desired output: I would expect a new column speed1_n1_plus that combines the values from speed1and n1_plus:

  string   speed1   speed2 n1 n2 n1_plus n1_minus       speed1_n1_plus
1    car 3958.415 1049.172 70 91      25       53 3,958.415\n(25)
2  train 6203.919 8639.160 52 92      14       91 6,203.919\n(14)
3   bike 2966.391 2997.303 35 55      46       61 2,966.391\n(46)
4  plain 2755.266 1627.379 98 66       8       49 2,755.266\n( 8)

I simply have to do operations on multiple variables with similar names. The variable names are composites of the 'core' name (in this case 'n1', {{V2}}) and suffixes and prefixes. I would like to avoid additional arguments for each variable name as it adds just a suffix to the core name.

I was trying: !!paste0, as.name(), eval(parse(text=), ..., which may work outside a function, but for me not within.

Answer 1

my_fun <- function(dataf, V1, V2){
           dataf %>%
              dplyr::mutate("{{V1}}_{{V2}}_plus" := paste0(format({{V1}}, big.mark   = ","),
                  "\n(", format(!! rlang::sym(paste0(rlang::as_string(ensym(V2)), "_plus")), big.mark  = ","), ")"))}

-testing

df %>%
  my_fun(speed1, n1)
 string   speed1    speed2 n1 n2 n1_plus n1_minus  speed1_n1_plus
1    car 4453.441 3336.7287 92 97      28       56 4,453.441\n(28)
2  train 7718.381  638.5120 82 61       9       13 7,718.381\n( 9)
3   bike 4648.093 4267.8390  7 92      83       29 4,648.093\n(83)
4  plain 3815.145  793.6886 18 56      30       46 3,815.145\n(30)

Answer 2

I agree that it would be helpful to use variable names on the right hand side of an assignment within mutate. The reason this is not implemented, is because you can perform this more efficiently by formating your data_frame appropriately in a longer format.

To me, it seems like speed1 & n1 and speed2 & n2 belong together in pairs. Therefore, you could transform your df from containing 4 rows (one for every vehicle type, i.e., car, train, etc.) to 8 rows (one for every vehicle instance, i.e., car1, car2, etc.).

In your example, it would be easier to construct the data_frame already in this longer format but since you may have to use a database in the format you specified, let's reformat (note: this is very tedious because some of the information is stored in the variable names and needs to be transformed back into individual cells):

df_long = df %>% pivot_longer(-string) %>% #expand on everything but the column "string" (super long format but we need this to grab the information from the column names)
  mutate(number = gsub("\\D+", "", name), name = gsub("\\d+", "", name)) #separate the numbers from the variable names

#separate speed and everything starting with "n" and get them into a wider format
df_n = df_long %>% filter(grepl("^n", name)) %>% pivot_wider(names_from=name)
df_rest = df_long %>% filter(grepl("^n", name)==F) %>% pivot_wider(names_from=name)

df_tidy = full_join(df_rest, df_n) #join the data frames together
View(df_tidy) #take a look how the df looks differently now (including explicit NAs since n2_plus and n2_minus don't exist in your example)

Now you can simply do this to get the result you wanted:

df_tidy = df_tidy %>% mutate(result = paste0(format(speed, big.mark=","), "\n(", format(n_plus, big.mark=","), ")"))

Note: It may make sense to choose an even longer format such that n, n_plus, and n_minus are not different columns but are coded in another column n_kind with factor levels "standard", "plus", and "minus". But I cannot judge from your example.