Row-wise sort then concatenate across specific columns of data frame

Question

(Related question that does not include sorting. It's easy to just use paste when you don't need to sort.)

I have a less-than-ideally-structured table with character columns that are generic "item1","item2" etc. I would like to create a new character variable that is the alphabetized, comma-separated concatenation of these columns. So for example, in row 5, if item1 = "milk", item2 = "eggs", and item3 = "butter", the new variable in row 5 might be "butter, eggs, milk"

I wrote a function f() below that works on two character variables. However, I am having trouble

• Using mapply or other "vectorization" (I know it's really just a for loop)
• Generalizing the function to an arbitrary number of columns

Any help much appreciated.

df <- data.frame(a =c("foo","bar"), 
                 b= c("baz","qux"))   
paste(df$a,df$b, sep=", ")
# returns [1] "foo, baz" "bar, qux" ... but I want [1] "baz, foo" "bar, qux"

f <- function(a,b) paste(c(a,b)[order(c(a,b))],collapse=", ")
f("foo","baz") 
# returns [1] "baz, foo" ... which is what I want ... how to vectorize?

df$new_var <- mapply(f, df$a, df$b)
df 
#     a   b new_var      <- new_var is not what I want
# 1 foo baz    1, 2
# 2 bar qux    1, 2

# Interestingly, data.table is smart enough to fix my bad mapply
library(data.table)
dt <- data.table(a =c("foo","bar"), 
                 b= c("baz","qux"))  
dt[,new_var:=mapply(f, a, b)]
dt
#     a    b  new_var    <- new var IS what I want
# 1: foo baz baz, foo
# 2: bar qux bar, qux

Answer 1

My first thought would've been to do this:

dt[, new_var := paste(sort(.SD), collapse = ", "), by = 1:nrow(dt)]

But you could make your function work with a couple of simple modifications:

f = function(...) paste(c(...)[order(c(...))],collapse=", ")

dt[, new_var := do.call(function(...) mapply(f, ...), .SD)]