A Note on missing()

(This article was originally published at English Blog on Yihui Xie | 谢益辉, and syndicated at StatsBlogs.)

I had misunderstood the function missing() for several years. Originally I thought it only applies to an argument that does not have a default or user-specified value. For example, this is fairly easy to understand:

f = function(x) {
  missing(x)
}
f()  # should be TRUE

One day I was surprised to find that this also returned TRUE:

f = function(x = 1) {
  missing(x)
}
f()

What?! x does have a default value 1; why is it considered missing? Then I realized missing() really meant “argument/value not passed” (to the function call).

Below is a yet more surprising fact that I discovered:

f = function(x) {
  missing(x)
}
g = function(y) {
  f(y)
}
g()  # still returns TRUE

I was surprised because when g calls f(y), y does not exist, yet it still worked. It looks like we did pass y (whatever it really is) to f(), but f() sees nothing. Sounds like fun of lazy evaluation or something.

Anyway, I don’t recommend using missing(). It is fragile and you may break it unintentionally. Per its help page:

missing(x) is only reliable if x has not been altered since entering the function […]

What I often do is to set the defautl value to NA or NULL, and use is.na() / is.null() to test if the default value was explicitly changed by the user. Of course, this has a different meaning with missing(), but it is more robust. In particular, NA works better when the function is used in a vectorized call, e.g., mapply().



Please comment on the article here: English Blog on Yihui Xie | 谢益辉


Subscribe

Email:

  Subscribe