R devel - Feb 2004 - apply on logical data frame together with all (PR#6560)

Full_Name: Roland Puntaier
Version: 1.8.1
OS: Windows XP
Submission from: (NULL) (62.99.238.78)


I have a data frame with some columns being logical.
I wanted to calculate a column that is an AND combination of the other logical
columns.
I tried to use
  apply(df,1,all)
It did not work because of the implicit string conversion.
So I made the functions
  All<-function(...) all(as.logical(...))#because all cannot be used with
apply
  Any<-function(...) any(as.logical(...))#because any cannot be used with
apply
Now 
  apply(df,1,All)
seemed to work,
but produced some NAs where there was no reason for them.
I debugged apply and found that some logical values are transformed to "
TRUE"
instead of "TRUE". I did not follow that up closer. Instead I worked
around it
by doing the string conversion myself and then call apply(dfc,1,All)

roland.puntaier@chello.at writes:
> Full_Name: Roland Puntaier
> Version: 1.8.1
> OS: Windows XP
> Submission from: (NULL) (62.99.238.78)
> 
> 
> I have a data frame with some columns being logical.
> I wanted to calculate a column that is an AND combination of the other
logical
> columns.
> I tried to use
>   apply(df,1,all)
> It did not work because of the implicit string conversion.
> So I made the functions
>   All<-function(...) all(as.logical(...))#because all cannot be used
with apply
>   Any<-function(...) any(as.logical(...))#because any cannot be used
with apply
> Now 
>   apply(df,1,All)
> seemed to work,
> but produced some NAs where there was no reason for them.
> I debugged apply and found that some logical values are transformed to
" TRUE"
> instead of "TRUE". I did not follow that up closer. Instead I
worked around it
> by doing the string conversion myself and then call apply(dfc,1,All)
It's as documented (i.e. no bug) and the five-letter field for TRUE is
consequence of using format(). 

Behaviour has been changed in r-devel though, since you weren't the
first to be annoyed by it....

Meanwhile, I believe that do.call("cbind",df) is a better workaround.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907

On Mon,  9 Feb 2004 18:13:29 +0100 (CET), roland.puntaier@chello.at
wrote :
>Full_Name: Roland Puntaier
>Version: 1.8.1
>OS: Windows XP
>Submission from: (NULL) (62.99.238.78)
>
>
>I have a data frame with some columns being logical.
>I wanted to calculate a column that is an AND combination of the other
logical
>columns.
>I tried to use
>  apply(df,1,all)
>It did not work because of the implicit string conversion.
>So I made the functions
>  All<-function(...) all(as.logical(...))#because all cannot be used with
apply
>  Any<-function(...) any(as.logical(...))#because any cannot be used with
apply
>Now 
>  apply(df,1,All)
>seemed to work,
>but produced some NAs where there was no reason for them.
Here's a reproducible example of this:
> x <- c(FALSE, TRUE)
> y <- x
> df <- data.frame(x,y)
> df
      x     y
1 FALSE FALSE
2  TRUE  TRUE
> apply(df, 1, all)
Error in all(..., na.rm = na.rm) : incorrect argument
type
> All <- function(...) all(as.logical(...))
> apply(df, 1, All)
    1     2 
FALSE    NA 

The good news is that r-devel (to become 1.9.0) gives the correct
result because of this change:

    o   as.matrix.data.frame() now coerces an all-logical data frame
        to a logical matrix.

The bad news is that r-patched does not.  I guess the workaround while
you're waiting for 1.9 would be to write All as follows:

All<-function(x) all(as.logical(gsub(" ", "", x)))

Thanks for the report.

Duncan Murdoch