[R] missing() fails to detect missing

[Peter Dalgaard]

I'm not really sure how verboten it is to have methods that omit "...". The code seems to be instrumented to handle that case. 

The catch is that it does so by inserting a call to .local where .local is the method function, e.g. if you insert a browser() call at the start of the method, you'll see

> foo("a")
Called from: .local(x, y, ...)
Browse[1]> sys.status()
$sys.calls
$sys.calls[[1]]
foo("a")

$sys.calls[[2]]
foo("a")

$sys.calls[[3]]
.local(x, y, ...)

$sys.calls[[4]]
browser()

and the 3rd call is absent for foo(0). But y is not missing in the .local call, because, hey, it has a default in the parent. 

This also happens in plain function calls, e.g.

> f <- function(a=0)(function(a)missing(a))(a)
> f()
[1] FALSE

Also notice that it is not really "..." that is the trigger of this effect, but the fact that the method has different arguments from the generic in any way: You get the same situation with

> setMethod("foo", signature="numeric", function(x, y=0, z=1, ...){
+ if (missing(y))
+    return("must give y for numeric")
+  y
+ })
> foo(0)
[1] 0

 -pd 



> On 18 Apr 2024, at 06:32 , Jeff Newmiller via R-help <r-help using r-project.org> wrote:
> 
> Fascinating. Or, well, not.
> 
> Failing to use a method signature that is compatible with the generic is a no-no. So your bug seems to me to be outside the bounds of how R is supposed to be used. So don't do that.
> 
> On April 17, 2024 4:25:38 PM PDT, "Boylan, Ross via R-help" <r-help using r-project.org> wrote:
>> When a generic (S4) has an argument with a default followed by ..., missing() doesn't seem to work if the method omits the ...
>> --------------------Sample-------------------------------
>> foo <- function(x, y=0, ...){
>> "you are very generic"
>> }
>> 
>> # no ... in function arguments
>> setMethod("foo", signature="character", function(x, y=0){
>> if (missing(y))
>>   return("must give y for character")
>> y
>> })
>> 
>> setMethod("foo", signature="numeric", function(x, y=0, ...){
>> if (missing(y))
>>   return("must give y for numeric")
>> y
>> })
>> 
>> print(foo("a"))  #[1] 0
>> print(foo(0))    #[1] "must give y for numeric"
>> ------------------------------------------------------------------
>> It's the result for foo("a") I'm puzzled by, since missing(y) does not evaluate to TRUE.
>> 
>> Background
>> ==========
>> The methods documentation has 2 points on which the above definitions may fail.
>> 
>> 1. The generic has regular arguments and ... arguments.  But dotsMethods docs say
>>> either the signature of the generic function is "..." only, or it
>>> does not contain "..."
>> Since the arguments in ... are not part of the signatures I think I'm OK, but another reading is that
>> one just shouldn't mix ... and other arguments.
>> 
>> 2. setMethod docs say
>>> The definition must be a function with the same formal arguments as the generic; however, setMethod() will handle methods that add arguments, if ... is a formal argument to the generic.
>> Since the initial definition has arguments x, y, ... and the first method definition has only x, y, the arguments don't match.  So maybe that's the problem.
>> 
>> I don't know if the fact that y has a default value matters.
>> 
>> The real code has a function f that ordinarily requires an additional piece of information, y, to compute a result.  But for one class, the result doesn't depend on y and so that argument may be omitted.
>> 
>> Context
>> ======
>> R 4.3.3 on MS-Windows under RStudio 2023.12.1 build 402.
>> 
>> Thanks for any insights.
>> Ross
>> 
>> ______________________________________________
>> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
> 
> -- 
> Sent from my phone. Please excuse my brevity.
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes using cbs.dk  Priv: PDalgd using gmail.com