Providing known adjoint solutions

[kswartz92]

I'm solving a problem with a known adjoint solution. I was wondering if it would be possible/desirable to provide an interface to supply adjoint solutions, rather request the physics model to solve for them.

[btalamini]

We discussed this offline. Let me try to summarize where things stand:

• Currently, there is an accessor to the adjoint field that gives a mutable reference, so there is a workaround for the requested feature.
• There is a bigger question about whether this is a wise choice. We may remove the mutability.
• The cases currently known in which this feature would be useful are all linear problems, hence the solve for the adjoint field is inexpensive. It may not be worth the trouble to allow the adjoint solution to be set just to avoid a trivial solve.