After writing my own eigenvalue vex code I noticed in VOPs that there is an Eigenvalue VOP (and also a vex function). No help card available but I guess it work like this:

- Put in a matrix 3x3
return:
- The number of roots,
- in an array the real value of the roots,
- and/or the imaginary value of the roots (or the imaginary part of the roots)

Is this correct?

I think sidefx should publish few basic examples for every node.
it will solve lots of questions.

from connecting print node and playing with it
seems you right :-) .

At least for the r values that seems to be the case. You can test it by creating some geometry and then multiply the eigenvectors (i assume thats what they are) by the matrix. They should all point into the same direction. For other vectors this is not the case, so it must be true that the r values are the eigenvectors.

Ah, I must have gotten something muddled up there. It's eigenvectors in eigenvalue disguise. I got to do some checking up. I made vectors out of values, and I'm not sure that is correct (maybe it is, it seems to work anyway).

Edit: I checked; they must be the values because their sum equals the trace of the matrix and det(M-value*I)=0. I also tested with an orthogonal rotation matrix and I get the expected imaginary values (then the matrix is “maximally” non-symmetric, and so you should get -i and i, which is the case). So yes, you seem to be right with your assumptions.