def test_update_state_return_new_state(self):
# Given
state = State([])
# When
new_state = state.update_function(None)
# Assert
assert isinstance(new_state, State)
assert new_state is not state
def test_state_method_update_should_return_new_state(self):
# Given
update = lambda x, n: (e + n for e in x)
state = State([0, 1], update_function=update)
# When
new_state = state.update(1)
# Assert
assert new_state is not state
assert new_state.to_list() == [1, 2]
def test_states_are_iterable(self):
# Given
state1 = State([1, 2, 3])
state2 = State([0, 1])
my_list = []
# When
a, b, c = state1
for element in state2:
my_list.append(element)
# Assert
assert a == 1
assert b == 2
assert c == 3
assert my_list == [0, 1]
def test_run_full_train_experiment(self):
# Given
state = State([0, 20]).update_function(update_state)
agent = Agent(state, RandomQFunction(), state_to_actions)
experiment = Experiment(agent)
# When
experiment.run()
# Assert
assert experiment.agent.state[1] == 0
def test_update_should_call_update_function(self):
# Given
mock_function = MagicMock()
state = State([], update_function=mock_function)
agent = Agent(state=state)
# When
action = "Do nothing"
agent.update_state(action=action)
# Assert
mock_function.assert_called_once_with([], "Do nothing")
def test_run_full_train_experiment_with_record(self):
# Given
state = State([0, 20]).update_function(update_state)
agent = Agent(state, RandomQFunction(), state_to_actions)
recorder = Recorder()
experiment = Experiment(agent)
# When
experiment.run(recorder=recorder)
# Assert
assert len(list(recorder.all_transitions)) > 0
def test_state_should_cast_to_list(self):
# Given
state1 = State([1, 2, 3])
state2 = State([])
# Assert
assert state1.to_list() == [1, 2, 3]
assert state2.to_list() == []
def test_zero_q_function_should_return_vector(self):
# Given
q_function = ZeroQFunction()
state = State([])
two_actions = Actions("Do Nothing", "Do Something")
three_actions = Actions("Do Nothing", "Do Something",
"Do Something else")
# When
two_q_values = q_function.evaluate(state, two_actions)
three_q_values = q_function.evaluate(state, three_actions)
# Check
assert len(two_q_values) == 2
assert len(three_q_values) == 3
def state_and_actions_to_matrix(state: State, actions: Actions):
return [state.to_list() + [action] for action in actions]
def update_state(state, action):
position, speed = state
new_speed = max(speed - action, 0)
new_position = position + speed
return State((new_position, new_speed))