R lazy evaluation paradox (R bug?)

Question

I have multiple functions handing around arguments that may be missing.

e.g. i have

mainfunction <- function(somearg) {
    mytest(somearg)
    fun <- function() { subfunction(somearg) }
    fun()
}

with the interesting aspect that the only interaction of mytest(somearg) with the arg is that it tests if the argument isn’t missing:

mytest = function(somearg) {
    print(missing(somearg))
}

subfunction then again tests if it’s missing and treats it accordingly:

subfunction = function(somearg) {
    if (missing(somearg))
        somearg = NULL
    else
        somearg = matrix(somearg, cols = 2)
    # somearg is used here…
}

the kicker is that, with somearg missing, this doesn’t work: matrix(somearg, cols = 2) throws

argument "somearg" is missing, with no default

during debugging, i found the following:

1. at the start of mainfunction, missing(somearg) returns TRUE
2. in mytest, missing(somearg) returns TRUE
3. insubfunction, missing(somearg) returns FALSE (!!!!)

therefore the matrix branch is hit, but in reality, somearg is missing, so it fails…

wat.

Answer 1

the @BenBolker way:

mainfunction <- function(somearg = NULL) {
    mytest(somearg)
    fun <- function() { subfunction(somearg) }
    fun()
}

mytest = function(somearg) {
    print(is.null(somearg))
}

subfunction = function(somearg) {
    if (is.null(somearg))
        somearg = 1:10
    else
        somearg = matrix(somearg, ncol = 2)
   somearg
}

Another way, using explicit missing argument

mainfunction <- function(somearg) {
    is_missing <- missing(somearg)
    mytest(is_missing)
    fun <- function() { subfunction(somearg, is_missing) }
    fun()
}
mytest = function(x) {  print(x) }
subfunction = function(somearg, is_arg_missing) {
    if (is_arg_missing)
        somearg = 1:10
    else
        somearg = matrix(somearg, ncol = 2)
   somearg
}

A third way, using plain missing arg passing:

    mainfunction <- function(somearg) {
        is_missing <- missing(somearg)
        mytest(somearg)
        fun <- function() { 
          if (is_missing) subfunction() else 
              subfunction(somearg) 
        }
        fun()
    }

    mytest = function(somearg) {
        print(missing(somearg))
    }

    subfunction = function(somearg) {
        if (missing(somearg))
            somearg = 1:10
        else
            somearg = matrix(somearg, ncol = 2)
       somearg
    }