Communicate f-value back to caller?

[anriseth]

The linesearches communicates the step length back to the caller, but not the objective value.
We should pass the objective value back as well to stop unnecessary calls of the objective function in Optim and NLOpt NLSolve

[cortner]

But see Optim.jl #288

[anriseth]

If some version of what's discussed there is implemented, I guess this will be automatically sorted.
Currently, Optim is passing the newest information to linesearches via the LineSearchResults, but does not use the result when updating f_x. I think this line could be changed to use
the latest information in state.lsr:
https://github.com/JuliaOpt/Optim.jl/blob/11c935e3689ce05e8ef75ea30fff7f592a26f56d/src/optimize.jl#L195

[cortner]

I agree

[cortner]

note that some line searches will also have gradient information that should also be communicated back.

[anriseth]

Hopefully this can be sorted out after JuliaNLSolvers/Optim.jl#337

[pkofod]

This can be closed now, right?

[anriseth]

This can be closed now, right?

Yep, should be fixed with #37