Matrices being transposed

[kockahonza]

Hi again, I'm sorry to be setting up yet another issue but I think this one is likely directly related to this package. I'm using the package to symbolically solve for eigenvalues of a matrix of Symbolics values using the "Eigensystem" wolfram command and I have noticed that the solutions returned are incorrect and I'm pretty sure correspond to the eigensystem of the matrix transpose. I suspect this has to do with julia using column major, as when I do the same call on the transpose of the matrix I get exactly the same solutions as mathematica reports in its own gui.

[eswagel]

Could you give code with the example you used? It may be as simple a fix as adding a transpose.

[kockahonza]

Sure, I think this is the simplest demonstration:
in Julia

julia> @variables a, b, c, d
4-element Vector{Num}:
 a
 b
 c
 d

julia> mat = [a b; c d]
2×2 Matrix{Num}:
 a  b
 c  d

julia> SymbolicsMathLink.expr_to_mathematica(mat)
W`List[List[a, c], List[b, d]]`

copy pasting that into Mathematica for example as

aa = List[List[a, c], List[b, d]];
MatrixForm[aa]

shows the matrix transposed.

In terms of a fix, that is precisely what I have done for my code, just add a transpose call. Idk how that does in terms of efficiency etc. especially as the SymbolicsMathLink doesn't actually accept what tranpose returns but it needs to be made into a concrete matrix via for example a collect call. After reconsidering I don't think would have been caused by Julia using column major but more like a small bug somewhere in the code that constructs the List objects, fixing that would probably be the best fix.

[eswagel]

Good catch. I pushed a change to ensure the matrix is converted in row-major order. It won't be any less efficient than before.