def __init__(self, optimal_state, inputs=[topfarm.x_key, topfarm.y_key]):
        """
        Parameters
        ----------
        optimal_state : array_like, dim=(#wt,#inputs)
            optimal state array
        inputs : array_like
            list of input names
        """
        self.optimal_state = np.array(optimal_state)
        self.n_wt = self.optimal_state.shape[0]

        CostModelComponent.__init__(self, inputs, self.n_wt, self.cost)

    def cost(self, **kwargs):
        opt = self.optimal_state
        for i in self.input_keys:
            print(i, kwargs[i])        
        return np.sum([(kwargs[n] - opt[:, i])**2 for i, n in enumerate(self.input_keys)])

from openmdao.drivers.doe_generators import FullFactorialGenerator,\
    ListGenerator
from openmdao.drivers.doe_driver import DOEDriver
from topfarm.cost_models.dummy import DummyCost
import numpy as np
from topfarm.easy_drivers import EasyScipyOptimizeDriver
from topfarm.tests import npt
from topfarm import TopFarmProblem
from topfarm.constraint_components.spacing import SpacingConstraint
from topfarm.constraint_components.boundary import XYBoundaryConstraint
optimal = np.array([(0, 2, 4, 1), (4, 2, 1, 0)])
design_vars = {k: v for k, v in zip('xy', optimal.T)}
design_vars['z'] = (optimal[:, 2], 0, 4)
xyz_problem = TopFarmProblem(
    design_vars,
    cost_comp=DummyCost(optimal, ['x', 'y', 'z', 'type']),
    constraints=[SpacingConstraint(2), XYBoundaryConstraint([(0, 0), (4, 4)], 'square')],
    driver=EasyScipyOptimizeDriver(disp=False))
tf = TopFarmProblem(
    {'type': ([0, 0], 0, 1)},
    cost_comp=xyz_problem,
    driver=DOEDriver(FullFactorialGenerator(2)))
cost = tf.optimize()[0]
assert cost == 0

**Major change** As I was building a nested problem I have noticed that due to global partial declaration in the class CostModelComponent() the compute was being called unnecessarily for an integer input. So if the partials are approximated; for example the input variable 'type' would cause to call the compute n_wt times for no reason. You can run a test with below changes to see what i mean. To avoid this the local variable input_keys is now filtering type_key before declaring partials. And then in the compute_partials type_key is omitted as it is not declared. Not sure if we want to generalize this for user selected integer inputs. But for now it solves the main problem. **Minor change** I exposed the method as user input coz some components are capable of running complex step instead of finite difference. ----- Sample to recreate the problem I guess you can see this in different ways but I am writing the way I noticed it; First make two changes in dummy.py in /topfarm/cost_models to make sure partials are approximated and compute is printed out ``` def __init__(self, optimal_state, inputs=[topfarm.x_key, topfarm.y_key]): """ Parameters ---------- optimal_state : array_like, dim=(#wt,#inputs) optimal state array inputs : array_like list of input names """ self.optimal_state = np.array(optimal_state) self.n_wt = self.optimal_state.shape[0] CostModelComponent.__init__(self, inputs, self.n_wt, self.cost) def cost(self, **kwargs): opt = self.optimal_state for i in self.input_keys: print(i, kwargs[i]) return np.sum([(kwargs[n] - opt[:, i])**2 for i, n in enumerate(self.input_keys)]) ``` Copy this code taken from test anywhere in the computer and run. see the print on the screen for type- ``` from openmdao.drivers.doe_generators import FullFactorialGenerator,\ ListGenerator from openmdao.drivers.doe_driver import DOEDriver from topfarm.cost_models.dummy import DummyCost import numpy as np from topfarm.easy_drivers import EasyScipyOptimizeDriver from topfarm.tests import npt from topfarm import TopFarmProblem from topfarm.constraint_components.spacing import SpacingConstraint from topfarm.constraint_components.boundary import XYBoundaryConstraint optimal = np.array([(0, 2, 4, 1), (4, 2, 1, 0)]) design_vars = {k: v for k, v in zip('xy', optimal.T)} design_vars['z'] = (optimal[:, 2], 0, 4) xyz_problem = TopFarmProblem( design_vars, cost_comp=DummyCost(optimal, ['x', 'y', 'z', 'type']), constraints=[SpacingConstraint(2), XYBoundaryConstraint([(0, 0), (4, 4)], 'square')], driver=EasyScipyOptimizeDriver(disp=False)) tf = TopFarmProblem( {'type': ([0, 0], 0, 1)}, cost_comp=xyz_problem, driver=DOEDriver(FullFactorialGenerator(2))) cost = tf.optimize()[0] assert cost == 0 ```