[Rd] table(factor(x), exclude=NULL) (PR#11494)

[Peter Dalgaard]

David.Duffy at qimr.edu.au wrote:
> Hi. I don't know if this a bug or just annoying to me:
>
>   
>> x <- c(1,2,3,NA)
>>     
>
>   
>> table(x, exclude=NULL)
>>     
> x
>     1    2    3 <NA>
>     1    1    1    1
>
>   
>> table(factor(x), exclude=NULL)
>>     
>
> 1 2 3
> 1 1 1
>
> I don't think many people use factor(x, exclude=NULL): it is not the 
> default handling of character data by read.table().
>
> Cheers, David Duffy.
>
>   
I've moved this to "wishlist" in the  bug repository. It's a documented 
annoyance, but we might be able to do better.

The underlying issue is the following:

 > f <- factor(c(1:3,NA),labels=letters[3:1])
 > f
[1] c    b    a    <NA>
Levels: c b a
 > factor(f)
[1] c    b    a    <NA>
Levels: c b a
 > factor(f,exclude=NULL)
[1] c    b    a    <NA>
Levels: c b a <NA>
 > factor(f,levels=levels(f),exclude=NULL)
[1] c    b    a    <NA>
Levels: c b a
 > factor(f,levels=c(levels(f),NA),exclude=NULL)
[1] c    b    a    <NA>
Levels: c b a <NA>

and this code in table() suggests that the latter is what was intended 
(since we actually try to pass exclude=NULL):

        cat <- if (is.factor(a)) {
            if (!missing(exclude)) {
                ll <- levels(a)
                factor(a, levels = ll[!(ll %in% exclude)], exclude = if 
(is.null(exclude))
                  NULL
                else NA)
            }
            else a

(levels = if(is.null(exclude)) c(ll,NA) else ll[!(ll %in% exclude)] 
should do. This will change behaviour of the case where you have mixed 
factors and non-factors, so I'm not just implementing it right away.

-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark      Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)              FAX: (+45) 35327907