R help - Feb 2011 - Wired behavior of a 2-by-2 matrix indicies

Dear R,

I found a very wired behavior for a 2-by-2 matrix, see this example
> A <- matrix(1:4, 2)
> idx4A <- matrix(1:4, 2)
> A[idx4A]
Error in A[idx4A] : subscript out of bounds

But other matrices are fine,
> B <- matrix(1:9, 3)
> idx4B <- matrix(1:9, 3)
> B[idx4B]
[1] 1 2 3 4 5 6 7 8 9


I can reproduce this for both 32bit windows and 64bit linux with R 2.12.1
and some earlier versions.

Can someone tell me whether this a bug or a feature or something I don't
know?

Thanks!

Feng

-- 
Feng Li
Department of Statistics
Stockholm University
106 91 Stockholm, Sweden
http://feng.li/

	[[alternative HTML version deleted]]

Hello,

On Sat, Feb 26, 2011 at 12:11 PM, Feng Li <m at feng.li>
wrote:
> Dear R,
>
> I found a very wired behavior for a 2-by-2 matrix, see this example
>
>> A <- matrix(1:4, 2)
>> idx4A <- matrix(1:4, 2)
>> A[idx4A]
> Error in A[idx4A] : subscript out of bounds
This shouldn't work. From the help for "[",

          When indexing arrays by ?[? a single argument ?i? can be a
          matrix with as many columns as there are dimensions of ?x?;
          the result is then a vector with elements corresponding to
          the sets of indices in each row of ?i?.

Since there are two dimensions to A, and two rows to idx4A, then this
form is used.
The first row of idx4A is c(1, 3) so the first value extracted is
A[1, 3]
which doesn't exist, thus the error message.
> But other matrices are fine,
>
>> B <- matrix(1:9, 3)
>> idx4B <- matrix(1:9, 3)
>> B[idx4B]
> [1] 1 2 3 4 5 6 7 8 9
The part that surprises me is that this works.
Apparently there is an implicit conversion to vector here because
dim(B) != ncol(idx4B)

What you seem to want is
A[as.vector(idx4A)]
and
B[as.vector(idx4B)]

but you neglected to tell us what you expected the result to be.

Sarah
-- 
Sarah Goslee
http://www.functionaldiversity.org