How to use R's ellipsis feature when writing your own function?

Ryan C. Thompson picture Ryan C. Thompson · Jun 16, 2010 · Viewed 56.4k times · Source

The R language has a nifty feature for defining functions that can take a variable number of arguments. For example, the function data.frame takes any number of arguments, and each argument becomes the data for a column in the resulting data table. Example usage:

> data.frame(letters=c("a", "b", "c"), numbers=c(1,2,3), notes=c("do", "re", "mi"))
  letters numbers notes
1       a       1    do
2       b       2    re
3       c       3    mi

The function's signature includes an ellipsis, like this:

function (..., row.names = NULL, check.rows = FALSE, check.names = TRUE, 
    stringsAsFactors = default.stringsAsFactors()) 
{
    [FUNCTION DEFINITION HERE]
}

I would like to write a function that does something similar, taking multiple values and consolidating them into a single return value (as well as doing some other processing). In order to do this, I need to figure out how to "unpack" the ... from the function's arguments within the function. I don't know how to do this. The relevant line in the function definition of data.frame is object <- as.list(substitute(list(...)))[-1L], which I can't make any sense of.

So how can I convert the ellipsis from the function's signature into, for example, a list?

To be more specific, how can I write get_list_from_ellipsis in the code below?

my_ellipsis_function(...) {
    input_list <- get_list_from_ellipsis(...)
    output_list <- lapply(X=input_list, FUN=do_something_interesting)
    return(output_list)
}

my_ellipsis_function(a=1:10,b=11:20,c=21:30)

Edit

It seems there are two possible ways to do this. They are as.list(substitute(list(...)))[-1L] and list(...). However, these two do not do exactly the same thing. (For differences, see examples in the answers.) Can anyone tell me what the practical difference between them is, and which one I should use?

Answer

Marek picture Marek · Jun 21, 2010

I read answers and comments and I see that few things weren't mentioned:

  1. data.frame uses list(...) version. Fragment of the code:

    object <- as.list(substitute(list(...)))[-1L]
    mrn <- is.null(row.names)
    x <- list(...)
    

    object is used to do some magic with column names, but x is used to create final data.frame.
    For use of unevaluated ... argument look at write.csv code where match.call is used.

  2. As you write in comment result in Dirk answer is not a list of lists. Is a list of length 4, which elements are language type. First object is a symbol - list, second is expression 1:10 and so on. That explain why [-1L] is needed: it removes expected symbol from provided arguments in ... (cause it is always a list).
    As Dirk states substitute returns "parse tree the unevaluated expression".
    When you call my_ellipsis_function(a=1:10,b=11:20,c=21:30) then ... "creates" a list of arguments: list(a=1:10,b=11:20,c=21:30) and substitute make it a list of four elements:

    List of 4
    $  : symbol list
    $ a: language 1:10
    $ b: language 11:20
    $ c: language 21:30
    

    First element doesn't have a name and this is [[1]] in Dirk answer. I achieve this results using:

    my_ellipsis_function <- function(...) {
      input_list <- as.list(substitute(list(...)))
      str(input_list)
      NULL
    }
    my_ellipsis_function(a=1:10,b=11:20,c=21:30)
    
  3. As above we can use str to check what objects are in a function.

    my_ellipsis_function <- function(...) {
        input_list <- list(...)
        output_list <- lapply(X=input_list, function(x) {str(x);summary(x)})
        return(output_list)
    }
    my_ellipsis_function(a=1:10,b=11:20,c=21:30)
     int [1:10] 1 2 3 4 5 6 7 8 9 10
     int [1:10] 11 12 13 14 15 16 17 18 19 20
     int [1:10] 21 22 23 24 25 26 27 28 29 30
    $a
       Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
       1.00    3.25    5.50    5.50    7.75   10.00 
    $b
       Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
       11.0    13.2    15.5    15.5    17.8    20.0 
    $c
       Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
       21.0    23.2    25.5    25.5    27.8    30.0 
    

    It's ok. Lets see substitute version:

       my_ellipsis_function <- function(...) {
           input_list <- as.list(substitute(list(...)))
           output_list <- lapply(X=input_list, function(x) {str(x);summary(x)})
           return(output_list)
       }
       my_ellipsis_function(a=1:10,b=11:20,c=21:30)
        symbol list
        language 1:10
        language 11:20
        language 21:30
       [[1]]
       Length  Class   Mode 
            1   name   name 
       $a
       Length  Class   Mode 
            3   call   call 
       $b
       Length  Class   Mode 
            3   call   call 
       $c
       Length  Class   Mode 
            3   call   call 
    

    Isn't what we needed. You will need additional tricks to deal with these kind of objects (as in write.csv).

If you want use ... then you should use it as in Shane answer, by list(...).