Export SOCP

[zoharl]

 What should I do in the assign and afterwards for x,t in solve_yalmip_dual_export() to return the same values as in solve_yalmip_dual()?

https://www.dropbox.com/s/gnxw8lifu7nfc80/yalmip2d.zip

[Johan Löfberg]

Since export uses the same crappy setup as a standard call, it introduces dummy variables which must be skipped. They are placed last so that is easily done

assign(recover(recoverymodel.used_variables),res.sol.itr.xx(1:length(recoverymodel.used_variables)));

[zoharl]

This might get the primal variables, but I get NaN for the dual (x, t).

[Johan Löfberg]

That's probably going to be messy. Reverse engineering from callmosek and solvesdp leads to something along the lines of

D_struc = (sol.itr.suc-sol.itr.slc);
setduals(Constraints,D_struc,internalmodel.K);

I think the easiest is to to just wait for my update (I'll probably make a new release tomorrow). Mosek solves the problem in solve_yalmip in 2s while the model you generate via solve_yalmip_dual_export takes 3.5

[zoharl]

Messy by it works.

Even with your update (not urgent), I still need a working export. Erling, for example, prefers running mosekopt.

[zoharl]

Right, but I won't need the duals...