Should we use DenseMatrix instead of AbstractMatrix?

[amontoison]

I think we should use DenseMatrix instead of AbstractMatrix to ensure that the matrix can be represented as long vector in decompress! when needed.
Example:

function decompress!(A::SparseMatrixCSC, B::AbstractMatrix, result::ColumnColoringResult)
    (; compressed_indices) = result
    S = result.bg.S2
    check_same_pattern(A, S)
    nzA = nonzeros(A)
    for k in eachindex(nzA, compressed_indices)
        nzA[k] = B[compressed_indices[k]]
    end
    return A
end

[gdalle]

No, this would be restrictive without benefits because all AbstractArrays support linear indexing

[amontoison]

I didn't know that it was supported by all AbstractArray.
Thanks for the reference Guillaume!