vector labels are not permuted properly in a call to sort() (R 2.1)

"AndyL" == Liaw, Andy <andy_liaw at merck.com>
    on Tue, 4 Oct 2005 13:51:11 -0400 writes:

    AndyL> The `problem' is that sort() does not doing anything special when given
    AndyL> a matrix: it only treat it as a vector.  After sorting, it copies
    AndyL> attributes of the original input to the output.  Since dimnames are
    AndyL> attributes, they get copied as is.

exactly. Thanks Andy.

And I think users would want this (copying of attributes) in
many cases; in particular for user-created attributes

?sort  really talks about sorting of vectors and factors;
       and it doesn't mention attributes explicitly at all
       {which should probably be improved}.

One could wonder if R should keep the dim & dimnames
attributes for arrays and matrices.  
S-plus (6.2) simply drops them {returning a bare unnames vector}
and that seems pretty reasonable to me.

At least the user would never make the wrong assumptions that
Greg made about ``matrix sorting''.


    AndyL> Try:

    >> y <- matrix(8:1, 4, 2, dimnames=list(LETTERS[1:4], NULL))
    >> y

    AndyL> [,1] [,2]
    AndyL> A    8    4
    AndyL> B    7    3
    AndyL> C    6    2
    AndyL> D    5    1

    >> sort(y)

    AndyL> [,1] [,2]
    AndyL> A    1    5
    AndyL> B    2    6
    AndyL> C    3    7
    AndyL> D    4    8

    AndyL> Notice the row names stay the same.  I'd argue that this is the correct
    AndyL> behavior.

    AndyL> Andy

    >> From: Greg Finak
    >> 
    >> Not sure if this is the correct forum for this, 

yes, R-devel is the proper forum.
{also since this is really a proposal for a change in R ...}

    >> but I've found what I  
    >> would consider to be a potentially serious bug to the 
    >> unsuspecting user.
    >> Given a numeric vector V with class labels in R,  the following calls
    >> 
    >> 1.

    >> > sort(as.matrix(V))

    >> 
    >> and
    >> 
    >> 2.

    >> >as.matrix(sort(V))

    >> 
    >> produce different ouput. The vector is sorted properly in 
    >> both cases,  
    >> but only 2. produces the correct labeling of the vector. The call to  
    >> 1. produces a vector with incorrect labels (not sorted).
    >> 
    >> Code:

    >> >X<-c("A","B","C","D","E","F","G","H")
    >> >Y<-rev(1:8)
    >> >names(Y)<-X
    >> > Y

    >> A B C D E F G H
    >> 8 7 6 5 4 3 2 1

    >> > sort(as.matrix(Y))

    >> [,1]
    >> A    1
    >> B    2
    >> C    3
    >> D    4
    >> E    5
    >> F    6
    >> G    7
    >> H    8

    >> > as.matrix(sort(Y))

    >> [,1]
    >> H    1
    >> G    2
    >> F    3
    >> E    4
    >> D    5
    >> C    6
    >> B    7
    >> A    8
    >>