def test_high_oob(self):
snake = core.Snake(np.array([5, 5]), 3, core.RIGHT)
state = core.GameState(snake, 6, 6, 0)
self.assertTrue(state.is_playable())
state.update(core.RIGHT)
self.assertFalse(state.is_playable())
snake = core.Snake(np.array([5, 5]), 3, core.RIGHT)
state = core.GameState(snake, 6, 6, 0)
state.update(core.DOWN)
self.assertFalse(state.is_playable())
def test_state_matrix_correct(self):
snake = core.Snake(np.array([5, 5]), 3, core.RIGHT)
state = core.GameState(snake, 10, 10)
matrix = state.to_matrix()
food = state.food()
self._check_matrix(snake, matrix, food)
def test_multiple_food_items(self):
snake = core.Snake(np.array([5, 5]), 1, core.RIGHT)
state = core.GameState(snake, 10, 10, food_max=5)
self.assertEqual(5, len(state.food()))
state.update(core.RIGHT)
self.assertEqual(5, len(state.food()))
def test_food_regenerated_upon_eating(self):
snake = core.Snake(np.array([5, 5]), 1, core.RIGHT)
state = core.GameState(snake, 10, 10, seed=10, food_max=5)
self.assertEqual(5, len(state.food()))
for _ in range(4):
state.update(core.UP)
self.assertEqual(5, len(state.food()))
def test_state_becomes_unplayable_on_update(self):
snake = core.Snake(np.array([2, 1]), 3, core.LEFT)
state = core.GameState(snake, 5, 5)
self.assertTrue(state.is_playable())
state.update(core.LEFT)
self.assertTrue(state.is_playable())
state.update(core.LEFT)
self.assertFalse(state.is_playable())
assert_snake_has_position(snake, [[0, 1], [1, 1], [2, 1]])
def test_punishes_game_over(self):
snake = core.Snake(np.array([5, 5]), 2, core.LEFT)
state = core.GameState(snake, 10, 10)
for _ in range(4):
state.update(core.UP)
state.update(core.UP)
self.assertFalse(state.is_playable())
self.assertGreaterEqual(
-1, trainer.determine_reward(False, state.is_playable(), False))
def test_game_state_becomes_invalid_after_drought(self):
snake = core.Snake(np.array([5, 5]), 1, core.LEFT)
state = core.GameState(snake, 10, 10, food_max=0, max_drought=10)
moves = cycle([core.DOWN, core.RIGHT, core.UP, core.LEFT])
for _ in range(10):
move = next(moves)
state.update(move)
self.assertTrue(state.is_playable())
state.update(core.UP)
self.assertFalse(state.is_playable())
def test_game_state_does_not_degrade_normally(self):
snake = core.Snake(np.array([5, 5]), 1, core.LEFT)
state = core.GameState(snake, 10, 10, food_max=0)
# Execute 1,000 moves to check against a drought.
for i in range(250):
state.update(core.DOWN)
state.update(core.RIGHT)
state.update(core.UP)
state.update(core.LEFT)
self.assertTrue(state.is_playable())
def test_snake_grows_from_state(self):
# TODO(matthew-c21): Update this to match any changes to the seeding in GameState.
snake = core.Snake(np.array([5, 5]), 2, core.LEFT)
state = core.GameState(snake, 10, 10, seed=10)
for _ in range(4):
state.update(core.UP)
for _ in range(2):
state.update(core.LEFT)
self.assertEqual(3, len(snake))
def test_rewards_eating(self):
# Recall the 10 seed from test_core.
snake = core.Snake(np.array([5, 5]), 2, core.LEFT)
state = core.GameState(snake, 10, 10, seed=10)
has_eaten = False
for _ in range(4):
has_eaten = state.update(core.UP)
actual_reward = trainer.determine_reward(state.is_playable(), 0,
has_eaten)
self.assertLess(0, actual_reward)
def test_punish_self_intersection(self):
snake = core.Snake(np.array([5, 5]), 5, core.LEFT)
state = core.GameState(snake, 10, 10)
state.update(core.DOWN)
state.update(core.RIGHT)
old_state = state.to_matrix()
state.update(core.UP)
new_state = state.to_matrix()
self.assertGreaterEqual(
-1, trainer.determine_reward(state.is_playable(), 0, False))
def test_not_eating_gives_negative_distance(self):
snake = core.Snake(np.array([5, 5]), 2, core.LEFT)
state = core.GameState(snake, 10, 10, seed=10)
old_state = state.to_matrix()
state.update(core.UP)
new_state = state.to_matrix()
# snake at (5, 4). Food at (5, 1).
self.assertEqual(
-3.0,
trainer.determine_reward(True, state.min_distance_to_food(),
False))
def test_board_does_not_make_changes_after_game_over(self):
snake = core.Snake(np.array([1, 1]), 1, core.LEFT)
state = core.GameState(snake, 5, 5)
self.assertTrue(state.is_playable())
state.update(core.LEFT)
self.assertFalse(state.is_playable())
assert_snake_has_position(snake, [[0, 1]])
for i in range(10):
state.update(core.LEFT)
assert_snake_has_position(snake, [[0, 1]])
def test_setting_food_items_persistent(self):
snake = core.Snake(np.array([5, 5]), 1, core.RIGHT)
state = core.GameState(snake, 10, 10, food_max=5)
food = core._food_item(np.array([6, 5]), 100)
state.set_food([food])
self.assertEqual(5, len(state.food()))
self.assertTrue(any(x is food for x in state.food()))
state.update(core.RIGHT)
self.assertTrue(all(x is not food for x in state.food()))
self.assertEqual(5, len(state.food()))
def test_multiple_legal_moves(self):
snake = core.Snake(np.array([5, 5]), 3, core.LEFT)
state = core.GameState(snake, 10, 10, 0)
self.assertTrue(state.is_playable())
snake.move(core.LEFT)
self.assertTrue(state.is_playable())
snake.move(core.UP)
self.assertTrue(state.is_playable())
snake.move(core.RIGHT)
self.assertTrue(state.is_playable())
snake.move(core.UP)
self.assertTrue(state.is_playable())
assert_snake_has_position(snake, [[5, 3], [5, 4], [4, 4]])
matrix = state.to_matrix()
self._check_matrix(snake, matrix)