Doubt about "nzval" field in factorization?

General Usage
1

I have the below MWE. I have two things:
1- The matrices of factor and factor1 are different. However, why the fifth element of factor.nzval=7 not 6. but the matrices of ``?

using SparseArrays, LinearAlgebra, KLU
A_Mat = spzeros(3,3);
A_Mat[diagind(A_Mat)[1:3]] .= 1.0;
A_Mat[2,1] = 9; A_Mat[3,1] = 10; A_Mat[1,3] = 6;
b_Mat = zeros(3); b_Mat[1,1] = 2.0; b_Mat[2,1] = 3; b_Mat[3,1] = 4;
x_Mat = zeros(3);
factor = klu(A_Mat);
A_Mat[1,3] = 7;
factor1 = klu(A_Mat);
ldiv!(x_Mat, factor, b_Mat)
julia> factor.nzval
6-element Vector{Float64}:
  1.0
  9.0
 10.0
  1.0
  7.0
  1.0

julia> factor1.nzval
6-element Vector{Float64}:
  1.0
  9.0
 10.0
  1.0
  7.0
  1.0

2- In the REPL, I can see all the matrices of factor, for example factor.L, but I cannot see them in the workspace (variables/global(main)) in VS code. Any reason for that?

2

factor.nzval and factor1.nzval work, but they are not properties of KLUFactorization. factor.L.nzval and factor1.L.nzval give the different results that are expected.

The way KLUFactorization getproperty function is defined passes unrecognized values to getfield on the structure, which has an nzval (the purpose of which I hadn’t investigated).

Hope this clear up something. Maybe a PR is necessary not to allow access to every KLUFactorization field with getproperty.

2 Likes
3

Thanks for your reply and assistance.

They are properties of A_Mat, right?

What is PR?

4

Short for Pull Request (on GitHub). The way package development is often organized.

1 Like