Parametric sensitivity calculation in the Nlpsol base class

[jaeandersson]

The NLP solver base class has now an output for parametric sensitivities, see #566. This should be calculated during the optimization (possibly using the fact that it's an auxilliary output for the Hessian evaluation).

As a by-product of this calculation, correct multipliers could be calculated, which should be possible without extra overhead. This would resolve for example #667.

[ghorn]

what's the status of this?

[jaeandersson]

Not implemented, but relatively easy to implement. Suggested way forward: get it to work for SQPMethod (or other NLP solver), then move implementation to the base class.

[jaeandersson]

Postponed to 2.1.