R devel - Jan 2013 - Arguments passing through dot-dot-dot lose ability to check for missing()?

Hi R-devel. Is the following behavior in g1() and h1() expected? It seems
to make "..." arguments work slightly differently from named
arguments.

#missing() has the property that it looks "up the chain"
#for example, "z" can be missing in f3 even if
#that argument did have a name ("y") in f2
f1 <- function(x, ...) {
  cat("In f1, missing(x) is ", missing(x), "\n")
  f2(x, ...)
}

f2 <- function(y, ...) {
  cat("In f2, missing(y) is ", missing(y), "\n")
  f3(y, ...)
}

f3 <- function(z, ...) {
  cat("in f3, missing(z) is ", missing(z), "\n")
}

f1( , 2) #all TRUE

#does this also work with ... arguments? It seems to lose track.
g1 <- function(...) {
  cat("In g1, missing(..1) is ", missing(..1), "\n")
  g2(...)
}

g2 <- function(...) {
  cat("In g2, missing(..1) is ", missing(..1), "\n")
  g3(...)
}

g3 <- function(...) {
  cat("in g3, missing(..1) is ", missing(..1), "\n")
}

g1( , 2) #TRUE, TRUE, FALSE ?!

# we can also elicit this lossiness without looking at the virtual ..n
# symbols:

h1 <- function(x, ...) {
  cat("in h1, missing(x) is ", missing(x), "\n")
  h2(x, ...)
}

hh1 <- function(...)
  h2(...)

h2 <- function(...)
  h3(...)

h3 <- function(z, ...)
  cat("in h3, missing(z) is ", missing(z), "\n")

h3( , 1) #missing in h3
h2( , 1) #missing in h3
h1( , 1) #missing in h1 but NOT h3
hh1( , 1) #also not missing in h3

	[[alternative HTML version deleted]]

This seems related also. Using match.call to capture ... arguments returns
odd results for missing arguments. Compare against another method of
capturing ... expressions:

capture_subst <- function(...) eval(substitute(alist(...)))
capture_match <- function(...) match.call(expand.dots=FALSE)$...

f1 <- function(..., capture) {
  capture(...)
}

f1(x=1, q=, capture=capture_subst)
f1(x=1, q=, capture=capture_match)


On Wed, Jan 23, 2013 at 3:08 PM, Peter Meilstrup
<peter.meilstrup@gmail.com>wrote:
> Hi R-devel. Is the following behavior in g1() and h1() expected? It seems
> to make "..." arguments work slightly differently from named
arguments.
>
> #missing() has the property that it looks "up the chain"
> #for example, "z" can be missing in f3 even if
> #that argument did have a name ("y") in f2
> f1 <- function(x, ...) {
>   cat("In f1, missing(x) is ", missing(x), "\n")
>   f2(x, ...)
> }
>
> f2 <- function(y, ...) {
>   cat("In f2, missing(y) is ", missing(y), "\n")
>   f3(y, ...)
> }
>
> f3 <- function(z, ...) {
>   cat("in f3, missing(z) is ", missing(z), "\n")
> }
>
> f1( , 2) #all TRUE
>
> #does this also work with ... arguments? It seems to lose track.
> g1 <- function(...) {
>   cat("In g1, missing(..1) is ", missing(..1), "\n")
>   g2(...)
> }
>
> g2 <- function(...) {
>   cat("In g2, missing(..1) is ", missing(..1), "\n")
>   g3(...)
> }
>
> g3 <- function(...) {
>   cat("in g3, missing(..1) is ", missing(..1), "\n")
> }
>
> g1( , 2) #TRUE, TRUE, FALSE ?!
>
> # we can also elicit this lossiness without looking at the virtual ..n
> # symbols:
>
> h1 <- function(x, ...) {
>   cat("in h1, missing(x) is ", missing(x), "\n")
>   h2(x, ...)
> }
>
> hh1 <- function(...)
>   h2(...)
>
> h2 <- function(...)
>   h3(...)
>
> h3 <- function(z, ...)
>   cat("in h3, missing(z) is ", missing(z), "\n")
>
> h3( , 1) #missing in h3
> h2( , 1) #missing in h3
> h1( , 1) #missing in h1 but NOT h3
> hh1( , 1) #also not missing in h3
>
	[[alternative HTML version deleted]]