opt.get_initial_step error in python

[jerry-ban]

hi, I am getting an error while running optimization with the step size function in python environment, google search return nothing useful so looking for help here
.
I am trying to get the initial step size, based on an feasible starting solution x0. here by the example from NLOPT tutorial;

    _def run_offcial_example(self):
        opt = nlopt.opt(nlopt.LD_MMA, 2)
        opt.set_lower_bounds([-float('inf'), 0])
        opt.set_min_objective(self._official_myfunc)
        opt.add_inequality_constraint(lambda x, grad: self._official_myconstraint(x, grad, 2, 0), 1e-8)
        opt.add_inequality_constraint(lambda x, grad: self._official_myconstraint(x, grad, -1, 1), 1e-8)
        opt.set_xtol_rel(1e-4)
        print("\noptimization starts!")
        x0 = [1.234, 5.678]
        step_size = opt.get_initial_step(x0)
        x = opt.optimize(x0)

        print("\noptimization finished!\n")_

everyting works fine without the step size call: step_size = opt.get_initial_step(x0)
the step size call will shows error as follows:

return _nlopt.opt_get_initial_step(self, *args)
NotImplementedError: Wrong number or type of arguments for overloaded function 'opt_get_initial_step'.
 Possible C/C++ prototypes are:
   nlopt::opt::get_initial_step(std::vector< double,std::allocator< double > > &) const
   nlopt::opt::get_initial_step() const
   nlopt::opt::get_initial_step(std::vector< double,std::allocator< double > > const &,std::vector< double,std::allocator< double > > &) const

anything wrong here? thank you

[jschueller]

actually the in-place version is called, maybe it could be swapped with the value return variant (and renamed get_initial_step_in_place), but that would break the api
note that only value return variant is useful for the swig bindings