GradientProblemSolver::Options does not support update_state_every_iteration

[bchretien]

While implementing a callback for GradientProblemSolver, I realized that update_state_every_iteration is not supported for that kind of solver. Yet, GradientProblemSolver::Options::callbacks's documentation hints at the contrary (here):

// Callbacks are executed in the order that they are specified in
// this vector. By default, parameter blocks are updated only at
// the end of the optimization, i.e when the Minimizer
// terminates. This behaviour is controlled by
// update_state_every_variable. If the user wishes to have access
// to the update parameter blocks when his/her callbacks are
// executed, then set update_state_every_iteration to true.

Are there any plans to support this option, or am I missing something? Otherwise, the documentation should probably be updated. Also, update_state_every_variable is referenced, but I'm assuming this is a typo/legacy name for update_state_every_iteration (it is also present in some other parts of the documentation).

This was tested with Ceres 1.12.0.

[sandwichmaker]

@bchretien thanks for reporting this, yes this is a bug in gradientproblemsolver, even though the underlying machinery actually supports this. You can either submit a patch or I can look into it soonish.

[bchretien]

@sandwichmaker thanks for the quick reply, I probably won't have the time to work on a patch during the next few weeks, but I'll keep you posted if that changes.

[bchretien]

I think I solved the issue (cf. this). I'll forward this to Gerrit when I find the time.

[sandwichmaker]

@bchretien that CL looks pretty good. Please do consider sending it to gerrit, I will review as soon as I can.

[sandwichmaker]

@bchretien gentle ping :)