R help - Aug 2010 - as.logical(factor) behaviour

Hello,

According to ?as.logical:

"as.logical attempts to coerce its argument to be of logical type. For 
factors, this uses the levels (labels)."

However,

 > as.logical(factor(c("FALSE", "TRUE")))
[1] TRUE TRUE

Shouldn't it be the same as:

 > as.logical(levels(factor(c("FALSE", "TRUE"))))
[1] FALSE  TRUE

according to the documentation? Did I miss something here?

 > sessionInfo()
R version 2.11.1 RC (2010-05-29 r52140)
x86_64-apple-darwin9.8.0

locale:
[1] C/UTF-8/C/C/C/C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

Thanks,

Philippe
-- 
..............................................<?}))><........
  ) ) ) ) )
( ( ( ( (    Prof. Philippe Grosjean
  ) ) ) ) )
( ( ( ( (    Numerical Ecology of Aquatic Systems
  ) ) ) ) )   Mons University, Belgium
( ( ( ( (
..............................................................

The problem is that, underneath the factors are actually numbers (1
and 2), where as, if you extract the levels and then get the logical,
it converts them to strings and then to logicals.  I run into this
problem ALL THE TIME with numerics in a dataset.  Consider the
following:
> factor(c(3,6,5,2,7,8,4))
[1] 3 6 5 2 7 8 4
Levels: 2 3 4 5 6 7 8
> as.numeric(factor(c(3,6,5,2,7,8,4)))
[1] 2 5 4 1 6 7 3
> as.numeric(as.character(factor(c(3,6,5,2,7,8,4))))
[1] 2 3 4 5 6 7 8

as.logical converts all non-zeros to TRUE, and 0 to
false:
> as.logical(0:10)
 [1] FALSE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE

Hope that helps,
Sam

On Sun, Aug 15, 2010 at 5:32 AM, Philippe Grosjean
<phgrosjean at sciviews.org> wrote:
> Hello,
>
> According to ?as.logical:
>
> "as.logical attempts to coerce its argument to be of logical type. For
> factors, this uses the levels (labels)."
>
> However,
>
>> as.logical(factor(c("FALSE", "TRUE")))
> [1] TRUE TRUE
>
> Shouldn't it be the same as:
>
>> as.logical(levels(factor(c("FALSE", "TRUE"))))
> [1] FALSE ?TRUE
>
> according to the documentation? Did I miss something here?
>
>> sessionInfo()
> R version 2.11.1 RC (2010-05-29 r52140)
> x86_64-apple-darwin9.8.0
>
> locale:
> [1] C/UTF-8/C/C/C/C
>
> attached base packages:
> [1] stats ? ? graphics ?grDevices utils ? ? datasets ?methods ? base
>
> Thanks,
>
> Philippe
> --
> ..............................................<?}))><........
> ?) ) ) ) )
> ( ( ( ( ( ? ?Prof. Philippe Grosjean
> ?) ) ) ) )
> ( ( ( ( ( ? ?Numerical Ecology of Aquatic Systems
> ?) ) ) ) ) ? Mons University, Belgium
> ( ( ( ( (
> ..............................................................
>
> ______________________________________________
> R-help at r-project.org mailing list
> 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.
>

Begin forwarded message:
> From: Peter Dalgaard <pdalgd at gmail.com>
> Date: August 15, 2010 3:43:31 PM EDT
> To: phgrosjean at sciviews.org
> Cc: r-help at r-project.org
> Subject: Re: [R] as.logical(factor) behaviour
>
> Philippe Grosjean wrote:
>> Thank you, but I already know that. I am not surprised by this  
>> behavior,
>> but by an inconsistency between that behavior and the documentation  
>> that
>> says "For factors, this uses the levels (labels).", which it
does
>> not.
>> Best,
>
> My gut feeling say that the docs need to be changes. The current
> situation is a bit silly because de facto, as.logical(f) is always  
> TRUE
> (except  for NA handling). However, fixing the code to match the docs
> would instead give NA in nearly all cases, and I suspect that has
> potential to upset code all over the place.
>
> It is easy enough to try and see what happens:
>
>> as.logical.factor <- function(x)as.logical(levels(x))[x]
>> as.logical(factor(0))
> [1] NA
>> as.logical(factor(FALSE))
> [1] FALSE
>> as.logical(factor(TRUE))
> [1] TRUE
>> as.logical(factor(1))
> [1] NA
>> as.logical("0")
> [1] NA
>
What about changing it to behave thusly:

 > as.logical.factor <- function(vec)  
as.logical(as.numeric(factor(vec))-1 )

 > as.logical(as.numeric(factor(c("TRUE", "FALSE",
NA)))-1 )
[1]  TRUE FALSE    NA
 > as.logical(as.numeric(factor(c(TRUE, FALSE, NA)))-1 )
[1]  TRUE FALSE    NA
>

-- 

David Winsemius, MD
West Hartford, CT