I have written about the size of created sparse matrices with Matrix R package here. After discussion with the developers of the package, I realized that one should not create sparse matrices in a way presented here i.e. setting up the matrix and then filling it is NOT OK. Instead one should either start with sparse design matrices and create the crossprods e.g.
> tmp <- factor(rep(letters[1:3], each=4))
 a a a a b b b b c c c c
Levels: a b c
> (x <- as(tmp, "sparseMatrix"))
3 x 12 sparse Matrix of class "dgCMatrix"
a 1 1 1 1 . . . . . . . .
b . . . . 1 1 1 1 . . . .
c . . . . . . . . 1 1 1 1
3 x 3 sparse Matrix of class "dsCMatrix"
a b c
a 4 . .
b . 4 .
c . . 4
or directly via the "triplets":
i <- as.integer(c(1, 2, 3))
j <- as.integer(c(1, 2, 3))
x <- c(4, 4, 4)
new("dsTMatrix", Dim=c(3L, 3L), i=(i - 1L), j=(j - 1L), x=x)
The above code should of course be more "automagical" but shows the point. Additionally, creating a diagonal matrix is of course easy (there is a special class for them in Matrix), but the above code works in general setting.