[Rd] sort.int(S3object) strips class but not the is.object flag

William Dunlap wdunlap at tibco.com
Mon Apr 11 19:28:22 CEST 2011


> -----Original Message-----
> From: r-devel-bounces at r-project.org 
> [mailto:r-devel-bounces at r-project.org] On Behalf Of William Dunlap
> Sent: Monday, April 11, 2011 10:10 AM
> To: R-devel at r-project.org
> Subject: [Rd] sort.int(S3object) strips class but not the 
> is.object flag
> 
> If x has an S3 class then sort.int(x) returns a value
> without an S3 class but which has the is.object flag set,
> which, I think, causes identical() give a false/misleading
> report:
>  
>   > x <- structure(1:3, class="unrecognizedClass")
>   > y <- sort.int(x)
>   > t <- 1:3
>   > identical(y, t) # expect TRUE
>   [1] FALSE

The misleading is.object flag can also affect
the time it takes to sort objects (it looks like
xtfrm is only called when is.object reports TRUE):
  > system.time(for(i in 1:1e5) order(y)) # is.object(y) is TRUE
     user  system elapsed 
    12.07    0.00   11.28 
  > system.time(for(i in 1:1e5) order(t)) # is.object(t) is FALSE
     user  system elapsed 
     1.98    0.01    2.01

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com 
 
>   > identical(as.vector(y), as.vector(t)) # expect TRUE
>   [1] FALSE
>   > dput(y)
>   1:3
>   > dput(t)
>   1:3
>   > class(y)
>   [1] "integer"
>   > class(t)
>   [1] "integer"
>   > is.object(y)
>   [1] TRUE
>   > is.object(t)
>   [1] FALSE
> 
> The files made by
>   save(t, file="t.Rdata", compress=FALSE)
>   save(y, file="y.Rdata", compress=FALSE)
> differ in 2 places, where the first is presumably
> the name of the object:
>   % cmp -l y.Rdata t.Rdata
>   36 171 164
>   39   1   0
> (The problem persists after a save/load cycle.)
> 
> This is on R 2.12.2 on Linux.  Sorry, I don't have 2.13.0
> yet installed.
> 
> Bill Dunlap
> Spotfire, TIBCO Software
> wdunlap tibco.com 
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
> 



More information about the R-devel mailing list