Test if an argument of a function is set or not in R

Question

I have a function f that takes two parameters (p1 and p2):

If for the parameter p2 no value was passed to the function, the value of p1^2 should be used instead. But how can I find out within the function, if a value is given or not. The problem is that the variable p2 is not initialized if there was no value. Thus I can't test for p2 being NULL.

f <- function(p1, p2) {
    if(is.null(p2)) {
        p2=p1^2
    }
    p1-p2
}

Is it somehow possible to check if a value for p2 was passed to the function or not? (I could not find an isset() - function or similar things.)

Solution 1

You use the function missing() for that.

f <- function(p1, p2) {
    if(missing(p2)) {
        p2=p1^2
    }
    p1-p2
}

Alternatively, you can set the value of p2 to NULL by default. I sometimes prefer that solution, as it allows for passing arguments to nested functions.

f <- function(p1, p2=NULL) {
    if(is.null(p2)) {
        p2=p1^2
    }
    p1-p2
}

f.wrapper <-function(p1,p2=NULL){
    p1 <- 2*p1
    f(p1,p2)
}
> f.wrapper(1)
[1] -2
> f.wrapper(1,3)
[1] -1

EDIT: you could do this technically with missing() as well, but then you would have to include a missing() statement in f.wrapper as well.

Solution 2

I think '?missing' should do it.

Solution 3

In a case like this you can also use something like this:

f <- function(p1, p2 = p1 ^ 2) {
    p1-p2
}

See the part on Lazy evaluation at http://adv-r.had.co.nz/Functions.html

Comments

[Josh O'Brien]

+1 for the NULL default alternative. Re: the EDIT you added, how would you pass missing'ness down through the wrapper function?

[Tommy]

@JoshO'Brien Passing missing-ness on is complicated AFAIK - but possible. That's why NULL is so much easier in this case.

[Tommy]

... So a separate question might be warranted for passing missing-ness on.

[Joris Meys]

@JoshO'Brien That would have to be if(missing(p2)) f(p1) else f(p1,p2). Or you could pass the arguments using the dots argument.

[Josh O'Brien]

Thanks. FWIW, I can't see how to do this with dots, if both functions have formal args p1 and p2. Another alternative uses the following in the body of f.wrapper(): {cl <- match.call(); cl$p1 <- 2*p1; cl[[1]] <- f; eval(cl)}. All this goes to show just how much cleaner your NULL alternative really is!

[Joris Meys]

@JoshO'Brien : you only give f the formal args p1 and p2. f.wrapper <- function(p1, ...){ blabla f(p1, ...)}. That would work and is how it's often done in R itself

[Diego]

I usually write it like this, so when would missing be most appropriate?

[Johan]

@Diego, if you look at the ?missing documentation you'll find an example that would be difficult to implement using this technique.