def test_full_column_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for i in range(AzulEnv.NUM_COLORS):
wall.reset()
wall.state[:, i] = True
for j in range(AzulEnv.NUM_COLORS):
self.assertEqual(wall.compute_build_reward(j, i), 12)
def test_single_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for i in range(AzulEnv.NUM_COLORS):
for j in range(AzulEnv.NUM_COLORS):
wall.reset()
wall.state[i, j] = True
self.assertEqual(wall.compute_build_reward(i, j), 1)
def test_full_row_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for i in range(AzulEnv.NUM_COLORS):
wall.reset()
wall.state[i, :] = True
for j in range(AzulEnv.NUM_COLORS):
self.assertEqual(wall.compute_build_reward(i, j), 7)
def test_break_one_tile(self):
wall = Wall(AzulEnv.NUM_COLORS)
self.assertEqual(wall.break_tiles(1), -1)
self.assertEqual(wall.break_tiles(1), -1)
self.assertEqual(wall.break_tiles(1), -2)
self.assertEqual(wall.break_tiles(1), -2)
self.assertEqual(wall.break_tiles(1), -2)
self.assertEqual(wall.break_tiles(1), -3)
self.assertEqual(wall.break_tiles(1), -3)
def test_full_color_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for color_idx in range(AzulEnv.NUM_COLORS):
wall.reset()
row_idx = np.arange(AzulEnv.NUM_COLORS)
column_idx = (row_idx + color_idx) % AzulEnv.NUM_COLORS
wall.state[row_idx, column_idx] = True
for i, j in zip(row_idx, column_idx):
self.assertEqual(wall.compute_build_reward(i, j), 11)
def test_adjecent_column_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for k in range(2, AzulEnv.NUM_COLORS):
for i in range(AzulEnv.NUM_COLORS):
for j in range(AzulEnv.NUM_COLORS - k + 1):
wall.reset()
wall.state[j:j + k, i] = True
for kk in range(k):
self.assertEqual(
wall.compute_build_reward(j + kk, i), k)
def test_adjecent_row_and_column_tile_reward(self):
wall = Wall(AzulEnv.NUM_COLORS)
for k in range(2, AzulEnv.NUM_COLORS):
for i in range(AzulEnv.NUM_COLORS - k + 1):
for j in range(AzulEnv.NUM_COLORS - k + 1):
wall.reset()
wall.state[i:i + k, j:j + k] = True
for ki in range(k):
for kj in range(k):
self.assertEqual(
wall.compute_build_reward(i + ki, j + kj),
k * 2)
def test_add_tiles_first_player_token(self):
wall = Wall(AzulEnv.NUM_COLORS)
first_player_token = False
reward = wall.add_tiles(0, 0, 1, first_player_token)
self.assertEqual(reward, 1)
first_player_token = True
reward = wall.add_tiles(2, 0, 1, first_player_token)
self.assertEqual(reward, 0)
first_player_token = True
reward = wall.add_tiles(2, 0, 1, first_player_token)
self.assertEqual(reward, -3)
def test_break_three_tiles(self):
wall = Wall(AzulEnv.NUM_COLORS)
self.assertEqual(wall.break_tiles(3), -4)
self.assertEqual(wall.break_tiles(3), -7)
self.assertEqual(wall.break_tiles(3), -3)
def test_break_two_tiles(self):
wall = Wall(AzulEnv.NUM_COLORS)
self.assertEqual(wall.break_tiles(2), -2)
self.assertEqual(wall.break_tiles(2), -4)
self.assertEqual(wall.break_tiles(2), -5)
self.assertEqual(wall.break_tiles(2), -3)
def test_break_five_tiles(self):
wall = Wall(AzulEnv.NUM_COLORS)
self.assertEqual(wall.break_tiles(5), -8)
self.assertEqual(wall.break_tiles(5), -6)
def __init__(self, factories, np_random):
self.factories = factories
self.np_random = np_random
self.wall = Wall(factories.num_colors)
self.score = 0