Assign "sparse" values to the diagonal of a sparse Matrix in R

I want to assign empty or sparse values into a sparse matrix in R. Specifically I want to remove all values in the diagonal of a sparse matrix from the Matrix package. I am currently only able to assign non-sparse 0's into the diagonal and then remove them with the drop0 function.

Is there a way to assign empty values directly into the diagonal of a sparse matrix?

The following code demonstrates the problem:

    library(Matrix)
    m       <- Diagonal(10) 
    diag(m) <- 0
    m
    drop0(m)

I would like to skip the drop0(m) step. In my real analysis I have off-diagonal values that I want to keep for a network analysis, so I am not really interested in a empty matrix but a way to assign "sparse" or "empty" values into a sparse matrix.

Solution

I assume you have indeed a matrix of class ddiMatrix and want to create an empty matrix:

m1 <- as(m, "generalMatrix")
m1@i <- integer(0)
m1@p <- m1@p * 0L
m1@x <- numeric(0)
#10 x 10 sparse Matrix of class "dgCMatrix"
#                         
# [1,] . . . . . . . . . .
# [2,] . . . . . . . . . .
# [3,] . . . . . . . . . .
# [4,] . . . . . . . . . .
# [5,] . . . . . . . . . .
# [6,] . . . . . . . . . .
# [7,] . . . . . . . . . .
# [8,] . . . . . . . . . .
# [9,] . . . . . . . . . .
#[10,] . . . . . . . . . .

Of course, you could also simply construct a new empty sparse matrix of these dimensions.

If you have a CsparseMatrix with off-diagonal values, coerce it to a TsparseMatrix and remove all i, j, x values where i == j (or skip the coercion and calculate the correct indices from i and p which I find difficult).

m1 <- as(m, "generalMatrix")

m2 <- as(m1, "TsparseMatrix")

del <- m2@i == m2@j
m2@i <- m2@i[!del]
m2@j <- m2@j[!del]
m2@x <- m2@x[!del]