R help - Sep 2004 - Strange behavior of is.na() on lists

Hello R-users, 


I have observed that is.na() behaves strange on some lists. Here is a
simple example:

> a = list(list('asd'))
> a
[[1]]
[[1]][[1]]
[1] "asd"

> for(i in 1:5)
+ print(is.na(a))
[1] TRUE
[1] FALSE
[1] TRUE
[1] TRUE
[1] TRUE
>
> for(i in 1:10)
+ print(as.integer(is.na(a)))
[1] 0
[1] 10
[1] 1
[1] 0
[1] 140897024
[1] 134567568
[1] 1
[1] 1953720684
[1] 145687400
[1] 1886352499
>

This is a very strange thing. The result is similar when applying
is.na() on more
complex lists: list(list(list(.....)). A concrete example where I
found this behavior is
the following list:

> str(as.list(hgu95av2GO)[1:5])
List of 5
 $ 1114_at  :List of 6
  ..$ GO:0005125:List of 3
  .. ..$ GOID    : chr "GO:0005125"
  .. ..$ Evidence: chr "IEA"
  .. ..$ Ontology: chr "MF"

  ...................................

  ..$ GO:0040007:List of 3
  .. ..$ GOID    : chr "GO:0040007"
  .. ..$ Evidence: chr "IEA"
  .. ..$ Ontology: chr "BP"
 $ 36421_at : logi NA
 $ 329_s_at :List of 5
  ..$ GO:0005198:List of 3
  .. ..$ GOID    : chr "GO:0005198"
  .. ..$ Evidence: chr "TAS"
  .. ..$ Ontology: chr "MF"

  ...................................

  ..$ GO:0005634:List of 3
  .. ..$ GOID    : chr "GO:0005634"
  .. ..$ Evidence: chr "TAS"
  .. ..$ Ontology: chr "CC"
 $ 34687_at :List of 3
  ..$ GO:0016020:List of 3
  .. ..$ GOID    : chr "GO:0016020"
  .. ..$ Evidence: chr "IEA"
  .. ..$ Ontology: chr "CC"
  
  ...................................



Is this behavior normal? I should mention that I try the examples on
different versions of R:

1  R 1.9.1	 Debian Sarge
2  R 2.0.0 beta	 Debian Sarge
3  R 1.9.1	 Windows
4  R 1.8.1	 SunOS


In my view this seems to be a bug. If is not, can somebody explain me what is 
really happening and how can I overcome this behavior?


Many thanks, 

Adrian 


-- 
Adrian Alexa             
Max-Planck-Institut fuer Informatik  
Stuhlsatzenhausweg 85 Room 514 
66123 Saarbruecken, Germany

Adrian Alexa <adrian.alexa at gmail.com> writes:
> 
> I have observed that is.na() behaves strange on some lists. Here is a
> simple example:
> 
> 
> > a = list(list('asd'))
> > a
> [[1]]
> [[1]][[1]]
> [1] "asd"
> 
> 
> > for(i in 1:5)
> + print(is.na(a))
> [1] TRUE
> [1] FALSE
> [1] TRUE
> [1] TRUE
> [1] TRUE
> >
> 
> [...snip...]
> 
> In my view this seems to be a bug. If is not, can somebody explain me what
is
> really happening and how can I overcome this behavior?
This looks like a bug to me.  As far as a workaround goes you
could try this:


is.na <- function(x)
   if (length(x) == 0) {
      mode(x) <- "logical"
      x
   } else if (inherits(x, "list")) {
      lapply(x, function(i) if (is.atomic(x)) base::is.na(x) else FALSE)
   } else 
      base::is.na(x)

On Sat, 25 Sep 2004 18:26:11 +0200, Adrian Alexa
<adrian.alexa at gmail.com> wrote:
>Hello R-users, 
>
>
>I have observed that is.na() behaves strange on some lists. Here is a
>simple example:
>
>
>> a = list(list('asd'))
>> a
>[[1]]
>[[1]][[1]]
>[1] "asd"
>
>
>> for(i in 1:5)
>+ print(is.na(a))
>[1] TRUE
>[1] FALSE
>[1] TRUE
>[1] TRUE
>[1] TRUE
Yes, this is a bug.  The problem is that the code leaves the result
unset in this case (no default on an incomplete switch statement).
I'll fix it.

Thanks for the report!

Duncan Murdoch