R: how to make passing of arguments using ... to multiple functions well-defined?

Question

V1: Suppose functions `f(x, ...)` and `g(x , ...)` can be passed different arguments. If I were to define a new function using both of them, can I make the passing of arguments via the `...` operator well-defined? As a simple example: ``` f1 = function(x, n = 1) x + n g1 = function(x, m = 1) x + m f = function(x, ...) f1(x, ...) g = function(x, ...) g1(x, ...) h = function(x, ...) { fgList = list() fgList[["f"]] = f(x, ...) fgList[["g"]] = g(x, ...) return(fgList) } h(1:4) # $f # [1] 2 3 4 5 # $g # [1] 2 3 4 5 h(1:4, n = 2) # Error in g1(x, ...) : unused argument (n = 2) ``` The argument `n` is being passed down to functions `f` and `g`, but it is only well-defined for function `f`. I want to mitigate against this. V2: If they are functions that I have defined, then Hong Ooi's solution below works perfectly. Can this solution be extended for pre-defined functions which don't have a `...` argument or equivalently, can a `...` argument be 'added' to a predefined function which doesn't h…

Accepted Answer

You can't have multiple versions of `...` in the one environment. What you can do, however, is give each of your called sub-functions a `...` argument of their own. This means they will ignore any parameters passed down that don't match their own formal arguments.

```
f1 = function(x, n = 1, ...) x + n
g1 = function(x, m = 1, ...) x + m

> h(1:4, n = 2)
$f
[1] 3 4 5 6

$g
[1] 2 3 4 5
```

Edit to answer added question: you can make a new version of `median`, which will override the predefined function when you call it in your own code. (Due to how R namespaces work, other predefined functions will still use the existing version.)

```
median <- function(x, na.rm=FALSE, ...)
base::median(x, na.rm)  # function median exported from base package
```

R: how to make passing of arguments using ... to multiple functions well-defined?

Solution

Problem