def test_step_auto(self, auto_act_mock, act_mock, ai_auto_act_mock):
auto_act_mock.return_value = True
act_mock.return_value = 45
ai_auto_act_mock.return_value = False
self.env._step_auto()
expected_space = RenjuSpace(15 * 15)
expected_space.remove(45)
expected_state = RenjuState(RenjuBoard(15), PlayerColor.WHITE,
PlayerColor.BLACK).act(45)
self.assertIsInstance(self.env._policies[PlayerColor.BLACK],
InputPolicy)
self.assertIsInstance(self.env._policies[PlayerColor.WHITE],
RandomPolicy)
self.assertEquals(expected_space, self.env.action_space)
self.assertEquals(expected_state, self.env._state)
def test_reset(self, act_mock):
act_mock.return_value = 100
observation = self.env._reset()
expected_space = RenjuSpace(15 * 15)
expected_space.remove(100)
expected_state = RenjuState(RenjuBoard(15), PlayerColor.WHITE,
PlayerColor.BLACK).act(100)
expected_observation = np.array([0 for _ in range(15 * 15)])
np.put(expected_observation, 100, 1)
self.assertIsInstance(self.env._policies[PlayerColor.BLACK],
RandomPolicy)
self.assertIsInstance(self.env._policies[PlayerColor.WHITE],
InputPolicy)
self.assertEquals(expected_space, self.env.action_space)
self.assertEquals(expected_state, self.env._state)
self.assertTrue(all(expected_observation == observation))
def test_init(self):
players = ['input', 'random']
env = RenjuEnv(players, 15, True)
self.assertIsInstance(env._policies[PlayerColor.BLACK], InputPolicy)
self.assertIsInstance(env._policies[PlayerColor.WHITE], RandomPolicy)
self.assertEquals(RenjuSpace(15 * 15), env.action_space)
self.assertEquals(
RenjuState(RenjuBoard(15), PlayerColor.WHITE, PlayerColor.BLACK),
env._state)
def test_render_not_change_state(self, finish_mock):
finish_mock.return_value = True
self.env._render()
self.assertIsInstance(self.env._policies[PlayerColor.BLACK],
InputPolicy)
self.assertIsInstance(self.env._policies[PlayerColor.WHITE],
RandomPolicy)
self.assertEquals(RenjuSpace(15 * 15), self.env.action_space)
self.assertEquals(
RenjuState(RenjuBoard(15), PlayerColor.WHITE, PlayerColor.BLACK),
self.env._state)
def test_rewards_finished(self, finsih_mock, get_reward_mock,
get_opponent_reward_mock):
finsih_mock.return_value = True
get_reward_mock.return_value = 10.0
get_opponent_reward_mock.return_value = -5.0
rewards = self.env.get_rewards()
self.assertEquals([-5.0, 10.0], rewards)
# Also test that the state not changed
self.assertIsInstance(self.env._policies[PlayerColor.BLACK],
InputPolicy)
self.assertIsInstance(self.env._policies[PlayerColor.WHITE],
RandomPolicy)
self.assertEquals(RenjuSpace(15 * 15), self.env.action_space)
self.assertEquals(
RenjuState(RenjuBoard(15), PlayerColor.WHITE, PlayerColor.BLACK),
self.env._state)
def test_step_valid(self, ai_act_mock):
ai_act_mock.return_value = 191
actual_observation, actual_reward, actual_is_finish, actual_state_dict = self.env._step(
79)
expected_space = RenjuSpace(15 * 15)
expected_space.remove(79)
expected_space.remove(191)
expected_state = RenjuState(RenjuBoard(15), PlayerColor.WHITE,
PlayerColor.BLACK).act(79).act(191)
expected_observation = np.array([0 for _ in range(15 * 15)])
np.put(expected_observation, [79, 191], [1, 2])
self.assertIsInstance(self.env._policies[PlayerColor.BLACK],
InputPolicy)
self.assertIsInstance(self.env._policies[PlayerColor.WHITE],
RandomPolicy)
self.assertEquals(expected_space, self.env.action_space)
self.assertEquals(expected_state, self.env._state)
self.assertTrue(all(expected_observation == actual_observation))
self.assertEquals(0, actual_reward)
self.assertEquals(False, actual_is_finish)
self.assertEquals({'state': expected_state}, actual_state_dict)
def generate(self, space_size: int) -> DiscreteSpace:
return RenjuSpace(space_size)