[R] Issue with the Matrix package

[Martin Maechler]

>>>>> "Tony" == Tony Chiang <tchiang at fhcrc.org>
>>>>>     on Fri, 4 May 2007 00:07:04 +0100 writes:

    Tony> Hi all,

    Tony> I am wondering if this is a bug in the Matrix package
    Tony> or if it something that I am just getting wrong...

    Tony> here is an example: 
    [..............]

It's a bug.

A shorter example - w/o dimnames - and showing what the semantic
really is for traditional matrices :

------------------------------------------------------------
> a <- matrix(0,4,4); a[c(1,2,1), 2] <- 1:3; a
     [,1] [,2] [,3] [,4]
[1,]    0    3    0    0
[2,]    0    2    0    0
[3,]    0    0    0    0
[4,]    0    0    0    0

> A <- Matrix(0,4,4); A[c(1,2,1), 2] <- 1:3; A
4 x 4 sparse Matrix of class "dgCMatrix"
            
[1,] . 4 . .
[2,] . 2 . .
[3,] . . . .
[4,] . . . .
------------------------------------------------------------

so we see that multiple assignments are supposed to happen
consecutively such that "the last wins".


    Tony> The documentation reads:

    Tony> " Most of the time, the function works via a traditional (_full_)
    Tony> 'matrix'.  However, 'Matrix(0, nrow,ncol)' directly constructs an
    Tony> "empty" sparseMatrix, as does 'Matrix(FALSE, *)'."

    Tony> So is this when an exception comes,
no

    Tony> and if so can someone explain to me why we get the 2?
    Tony> It would seem that it should just reassign the 1 to a
    Tony> 1 not add the number of times it is assigning a 1.

If you are interested: Things go via TsparseMatrix, i.e. the
triplet representation.  and there the convention is the
following: an index pair (i_0,j_0) can appear more than once. If
it does it *means* that all the 'x' values are summed up.

  ?dgTMatrix-class

does explain that, too.

Here's an example - using the auxiliary function I had posted a
while ago on R-help:

spMatrix <- function(nrow, ncol, i,j,x) {
    ## Purpose: User friendly construction of sparse "Matrix" from triple
    ## ----------------------------------------------------------------------
    ## Arguments: (i,j,x): 2 integer and 1 numeric vector of the same length:
    ##
    ##	The matrix M will have
    ##       M[i[k], j[k]] == x[k] , for k = 1,2,..., length(i)
    ##    and M[ i', j' ]  ==  0  for `` all other pairs (i',j')
    ## ----------------------------------------------------------------------
    ## Author: Martin Maechler, Date:  8 Jan 2007, 18:46
    dim <- c(as.integer(nrow), as.integer(ncol))
    ## The conformability of (i,j,x) with itself and with 'dim'
    ## will be checked automatically
    ## by an internal "validObject()" which is part of new(.):
    new("dgTMatrix", x = as.double(x), Dim = dim,
        ## our "Tsparse" Matrices use  0-based indices :
        i = as.integer(i - 1:1),
        j = as.integer(j - 1:1))
}

and now uses this lower-level construction of Tsparse Matrices:

> (A <- spMatrix(3,4, i= c(1:3, 1:2), j = c(2:4, 3:4), x = 1:5))
3 x 4 sparse Matrix of class "dgTMatrix"
            
[1,] . 1 4 .
[2,] . . 2 5
[3,] . . . 3

## ok

> (B <- spMatrix(3,4, i= c(1:3, 1:2), j = c(2:4, 2:3), x = 1:5))
3 x 4 sparse Matrix of class "dgTMatrix"
            
[1,] . 5 . .
[2,] . . 7 .
[3,] . . . 3

## oops!

> str(B)
Formal class 'dgTMatrix' [package "Matrix"] with 6 slots
  ..@ i       : int [1:5] 0 1 2 0 1
  ..@ j       : int [1:5] 1 2 3 1 2
  ..@ Dim     : int [1:2] 3 4
  ..@ Dimnames:List of 2
  .. ..$ : NULL
  .. ..$ : NULL
  ..@ x       : num [1:5] 1 2 3 4 5
  ..@ factors : list()

## which shows that you have 5 entries in B, with the "sum those
## with equal index" convention mentioned above.


Thank you for the report, Tony.
This will be fixed in the next release of Matrix.

Martin Maechler, ETH Zurich