def test_step(): qttt = Qttt() qttt.step((1, 0), 1) qttt.step((1, 0), 2) assert (qttt.board[0].entangled_marks == [1, 2]) assert (qttt.board[0].entangled_blocks == [1, 1]) assert (qttt.board[1].entangled_marks == [1, 2]) assert (qttt.board[1].entangled_blocks == [0, 0]) for i in range(2, 9): assert (qttt.board[i].entangled_marks == []) assert (qttt.board[i].entangled_blocks == [])
def test_get_free_QBlock_ids(): # one circle with four elements qttt = Qttt() qttt.step((0, 1), 1) qttt.step((1, 2), 2) qttt.step((2, 3), 3) qttt.step((0, 3), 4) assert ((qttt.get_free_QBlock_ids() == range(0, 9)).all()) qttt1, qttt2 = qttt.get_all_possible_collapse((0, 3), 4) qttt1.propagate_qttt_to_ttt() assert ((qttt1.get_free_QBlock_ids() == range(4, 9)).all()) qttt2.propagate_qttt_to_ttt() assert ((qttt2.get_free_QBlock_ids() == range(4, 9)).all())
def test_to_tensor(): s_env = EnvForDeepRL() result_tensor = torch.zeros(11, 9) _ = s_env.step(Qttt(), (0, 4))[0] result_tensor[1, [0, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (1, 4))[0] result_tensor[2, [1, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (2, 4))[0] result_tensor[3, [2, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (3, 4))[0] result_tensor[4, [3, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (5, 4))[0] result_tensor[5, [5, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (6, 4))[0] result_tensor[6, [6, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (7, 4))[0] result_tensor[7, [7, 4]] = 1 _ = s_env.step(s_env.get_valid_moves()[1][0], (8, 4))[0] result_tensor[8, [8, 4]] = 1 assert (result_tensor.reshape(11, 3, 3) == s_env.qttt.to_tensor()).all() _ = s_env.step(s_env.get_valid_moves()[1][0], (0, 4))[0] result_tensor = torch.zeros(11, 9) result_tensor[1, 0] = 1 result_tensor[2, 1] = 1 result_tensor[3, 2] = 1 result_tensor[4, 3] = 1 result_tensor[5, 5] = 1 result_tensor[6, 6] = 1 result_tensor[7, 7] = 1 result_tensor[8, 8] = 1 result_tensor[9, 4] = 1 result_tensor[0][:] = 1 # update collapsed_qttt even when it's done! s_env.step(s_env.collapsed_qttts[1], None) assert (result_tensor.reshape(11, 3, 3) == s_env.qttt.to_tensor()).all() s_env.change_to_even_pieces_view() result_tensor = torch.zeros(11, 9) result_tensor[2, 0] = 1 result_tensor[3, 1] = 1 result_tensor[4, 2] = 1 result_tensor[5, 3] = 1 result_tensor[6, 5] = 1 result_tensor[7, 6] = 1 result_tensor[8, 7] = 1 result_tensor[9, 8] = 1 result_tensor[10, 4] = 1 result_tensor[0][:] = 1 assert (result_tensor.reshape(11, 3, 3) == s_env.qttt.to_tensor()).all()
def test_view_of_env1(): # r 1 p 0 s_env = EnvForDeepRL() # r 2 p 1 _ = s_env.step(Qttt(), (0, 1))[0] # r 3 p 0 _ = s_env.step(s_env.collapsed_qttts[0], (0, 1))[0] # change when it has already been normal view before_view_change = deepcopy(s_env) s_env.change_to_normal_view() # r 3 p 0 assert before_view_change == s_env # odd piece round change to even view, everything except player id +1 # r 4 p 0 s_env.change_to_even_pieces_view() _, collapsed_qttts, _ = s_env.get_valid_moves() assert s_env.round_ctr == 4 assert s_env.current_player_id == 0 assert (collapsed_qttts[1].ttt.board == np.array([2, 3, 0, 0, 0, 0, 0, 0, 0])).all() # in converted even piece view, we continue to play # r 5 p 1 _ = s_env.step(s_env.collapsed_qttts[1], (3, 4))[0] assert s_env.round_ctr == 5 assert s_env.current_player_id == 1 assert (s_env.collapsed_qttts[0].ttt.board == np.array([2, 3, 0, 0, 0, 0, 0, 0, 0])).all() assert s_env.collapsed_qttts[0].board[3].entangled_blocks == [4] assert s_env.collapsed_qttts[0].board[3].entangled_marks == [4] # in odd piece round of converted view env, it will back to normal view, even piece round # r 4 p 1 s_env.change_to_normal_view() before_view_change = deepcopy(s_env) assert s_env.round_ctr == 4 assert s_env.current_player_id == 1 assert (s_env.qttt.ttt.board == np.array([1, 2, 0, 0, 0, 0, 0, 0, 0])).all() assert s_env.qttt.board[3].entangled_blocks == [4] assert s_env.qttt.board[3].entangled_marks == [3] s_env.change_to_even_pieces_view() assert s_env == before_view_change
def test_has_won(): ttt = Qttt.ttt() assert (ttt.has_won() == (False, "NO")) ttt.board = np.array([ 0, 1, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL ]) assert (ttt.has_won() == (False, "NO")) ttt.board = np.array([ 1, 2, Qttt.ttt.EMPTY_BLOCK_VAL, 3, 4, Qttt.ttt.EMPTY_BLOCK_VAL, 5, 6, Qttt.ttt.EMPTY_BLOCK_VAL ]) assert (ttt.has_won() == (True, "ODD_EVEN_WIN")) ttt.board = np.array([ 1, 2, Qttt.ttt.EMPTY_BLOCK_VAL, 3, 4, Qttt.ttt.EMPTY_BLOCK_VAL, 7, 6, Qttt.ttt.EMPTY_BLOCK_VAL ]) assert (ttt.has_won() == (True, "EVEN_ODD_WIN")) ttt.board = np.array([3, 2, 9, 4, 1, 8, 5, 6, 7]) assert (ttt.has_won() == (True, "ODD_WIN")) ttt.board = np.array([ 1, 3, 2, Qttt.ttt.EMPTY_BLOCK_VAL, 4, 6, Qttt.ttt.EMPTY_BLOCK_VAL, 8, 5 ]) assert (ttt.has_won() == (False, "NO")) ttt.board = np.array([1, 3, 2, 7, 4, 6, Qttt.ttt.EMPTY_BLOCK_VAL, 8, 5]) assert (ttt.has_won() == (True, "TIE")) ttt.board = np.array([3, 2, 1, 4, 9, 8, 5, 6, 7]) assert (ttt.has_won() == (True, "ODD_WIN")) ttt.board = np.array([ 2, 4, 6, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL ]) assert (ttt.has_won() == (True, "EVEN_WIN"))
def test_get_all_possible_collapse(): # one circle with four elements qttt = Qttt() qttt.step((0, 1), 1) qttt.step((1, 2), 2) qttt.step((2, 3), 3) qttt.step((0, 3), 4) qttt1, qttt2 = qttt.get_all_possible_collapse((0, 3), 4) # check qttt1 elements expectation_marks = [4, 1, 2, 3] for i in range(4): assert (qttt1.board[i].entangled_blocks == []) assert (qttt1.board[i].entangled_marks == [expectation_marks[i]]) # check qttt2 elements expectation_marks = [1, 2, 3, 4] for i in range(4): assert (qttt2.board[i].entangled_blocks == []) assert (qttt2.board[i].entangled_marks == [expectation_marks[i]]) # test to propagate to ttt qttt1.propagate_qttt_to_ttt() assert ((qttt1.ttt.board == np.array([ 4, 1, 2, 3, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL ])).all()) qttt2.propagate_qttt_to_ttt() assert ((qttt2.ttt.board == np.array([ 1, 2, 3, 4, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL, Qttt.ttt.EMPTY_BLOCK_VAL ])).all()) ''' take a look at special example 3 2 1,9 4 1,2,3 8 4,5,6 7,8,9 5 6 7 ''' s_qttt = Qttt() s_qttt.step((2, 4), 1) s_qttt.step((1, 4), 2) s_qttt.step((0, 4), 3) s_qttt.step((3, 4), 4) s_qttt.step((6, 4), 5) s_qttt.step((7, 4), 6) s_qttt.step((8, 4), 7) s_qttt.step((5, 4), 8) s_qttt.step((2, 4), 9) # s_qttt.visualize_board() qttt1, qttt2 = s_qttt.get_all_possible_collapse((2, 4), 9) ''' Finally, we get 2 possible collapsed Qttt state: 3 2 9 3 2 1 4 1 8 or 4 9 8 5 6 7 5 6 7 ''' # check qttt1 elements expectation_marks = [3, 2, 9, 4, 1, 8, 5, 6, 7] for i in range(4): assert (qttt1.board[i].entangled_blocks == []) assert (qttt1.board[i].entangled_marks == [expectation_marks[i]]) # check qttt2 elements expectation_marks = [3, 2, 1, 4, 9, 8, 5, 6, 7] for i in range(4): assert (qttt2.board[i].entangled_blocks == []) assert (qttt2.board[i].entangled_marks == [expectation_marks[i]]) # test to propagate to ttt qttt1.propagate_qttt_to_ttt() # qttt1.ttt.visualize_board() assert ((qttt1.ttt.board == np.array([3, 2, 9, 4, 1, 8, 5, 6, 7])).all()) qttt2.propagate_qttt_to_ttt() assert ((qttt2.ttt.board == np.array([3, 2, 1, 4, 9, 8, 5, 6, 7])).all())
def test_has_cycle(): # one circle with two element qttt = Qttt() qttt.step((2, 3), 1) assert (not qttt.has_cycle) qttt.step((2, 3), 2) assert qttt.has_cycle # one circle with four elements qttt2 = Qttt() qttt2.step((0, 1), 1) assert (not qttt2.has_cycle) qttt2.step((1, 2), 2) assert (not qttt2.has_cycle) qttt2.step((2, 3), 3) assert (not qttt2.has_cycle) qttt2.step((0, 3), 4) assert qttt2.has_cycle # test no cycle qttt = Qttt() qttt.step((0, 1), 1) assert (not qttt.has_cycle) qttt.step((2, 3), 2) assert (not qttt.has_cycle) # test no cycle qttt = Qttt() qttt.step((2, 4), 1) assert (not qttt.has_cycle) qttt.step((1, 4), 2) assert (not qttt.has_cycle)
def test_step(): # one circle with two element env = Env() _ = env.step(Qttt(), (1, 0))[0] free_qblock_id_lists, collapsed_qttts, collapse_choice = env.get_valid_moves() assert (len(free_qblock_id_lists) == 1) assert (len(free_qblock_id_lists) == len(collapsed_qttts)) assert ((free_qblock_id_lists[0] == range(9)).all()) assert (collapse_choice == ()) agent_move = (1, 0) _ = env.step(collapsed_qttts[0], agent_move)[0] free_qblock_id_lists, collapsed_qttts, collapse_choice = env.get_valid_moves() assert (collapse_choice == agent_move) assert ((free_qblock_id_lists[0] == range(2, 9)).all()) assert ((free_qblock_id_lists[1] == range(2, 9)).all()) assert (len(free_qblock_id_lists) == 2) assert (len(free_qblock_id_lists) == len(collapsed_qttts)) assert ((collapsed_qttts[0].ttt.board[:2] == np.array([1, 2])).all()) assert ((collapsed_qttts[1].ttt.board[:2] == np.array([2, 1])).all()) assert ((collapsed_qttts[0].ttt.board[:2] == np.array([1, 2])).all()) assert ((collapsed_qttts[1].ttt.board[:2] == np.array([2, 1])).all()) _ = env.step(collapsed_qttts[0], (2, 3))[0] free_qblock_id_lists, collapsed_qttts, collapse_choice = env.get_valid_moves() assert (collapse_choice == ()) assert (len(free_qblock_id_lists) == 1) assert (len(free_qblock_id_lists) == len(collapsed_qttts)) ''' take a look at special example 3 2 1,9 4 1,2,3 8 4,5,6 7,8,9 5 6 7 ''' s_env = Env() _ = s_env.step(Qttt(), (2, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (1, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (0, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (3, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (6, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (7, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (8, 4))[0] _ = s_env.step(s_env.get_valid_moves()[1][0], (5, 4))[0] free_qblock_id_lists, collapsed_qttts, _ = s_env.get_valid_moves() assert (len(free_qblock_id_lists) == 1) assert (len(free_qblock_id_lists) == len(collapsed_qttts)) assert ((free_qblock_id_lists[0] == range(9)).all()) assert ((collapsed_qttts[0].ttt.board == (np.ones(9) * Qttt.ttt.EMPTY_BLOCK_VAL)).all()) params = s_env.step(s_env.get_valid_moves()[1][0], (2, 4)) free_qblock_id_lists, collapsed_qttts, _ = s_env.get_valid_moves() assert (not free_qblock_id_lists[0]) assert (not free_qblock_id_lists[1]) assert ((collapsed_qttts[0].ttt.board == np.array([3, 2, 9, 4, 1, 8, 5, 6, 7])).all()) assert ((collapsed_qttts[1].ttt.board == np.array([3, 2, 1, 4, 9, 8, 5, 6, 7])).all()) assert (params[1] == 10) assert (params[2] == 0) assert (not params[3]) params = s_env.step(collapsed_qttts[1], None) assert (params[2] == -1) assert (params[3])
def test_view_of_env2(): # r 1 p 0 s_env = EnvForDeepRL() # r 2 p 1 s_env.step(Qttt(), (0, 4)) # r 3 p 0 s_env.step(s_env.get_valid_moves()[1][0], (1, 4)) # r 4 p 1 s_env.step(s_env.get_valid_moves()[1][0], (2, 4)) # r 5 p 0 s_env.step(s_env.get_valid_moves()[1][0], (3, 4)) # r 6 p 1 s_env.step(s_env.get_valid_moves()[1][0], (5, 4)) # r 7 p 0 s_env.step(s_env.get_valid_moves()[1][0], (6, 4)) # r 8 p 1 s_env.step(s_env.get_valid_moves()[1][0], (7, 4)) assert s_env.player_id == 1 assert s_env.round_ctr == 8 before_view_change = deepcopy(s_env) # r 8 p 1 s_env.change_to_even_pieces_view() # during even piece round nothing should change assert before_view_change.player_id == s_env.player_id assert before_view_change.round_ctr == s_env.round_ctr assert before_view_change.qttt == s_env.qttt assert before_view_change.collapse_choice == s_env.collapse_choice # r 9 p 0 _ = s_env.step(s_env.get_valid_moves()[1][0], (8, 4))[0] assert s_env.player_id == 0 assert s_env.round_ctr == 9 # round for odd piece # r 10 p 0 s_env.change_to_even_pieces_view() assert s_env.player_id == 0 assert s_env.round_ctr == 10 _, collapsed_qttts, _ = s_env.get_valid_moves() for i in range(3): assert collapsed_qttts[0].board[i].entangled_marks == [i + 2] assert collapsed_qttts[0].board[i].entangled_blocks == [4] for i in range(5, 9): assert collapsed_qttts[0].board[i].entangled_marks == [i + 1] assert collapsed_qttts[0].board[i].entangled_blocks == [4] assert collapsed_qttts[0] == s_env.qttt assert collapsed_qttts[0].board[4].entangled_blocks == [0, 1, 2, 3, 5, 6, 7, 8] assert collapsed_qttts[0].board[4].entangled_marks == [2, 3, 4, 5, 6, 7, 8, 9] assert (collapsed_qttts[0].ttt.board == np.ones(9) * Qttt.ttt.EMPTY_BLOCK_VAL).all() # r 9 p 0 s_env.change_to_normal_view() assert s_env.player_id == 0 assert s_env.round_ctr == 9 _, collapsed_qttts, _ = s_env.get_valid_moves() for i in range(3): assert collapsed_qttts[0].board[i].entangled_marks == [i + 1] assert collapsed_qttts[0].board[i].entangled_blocks == [4] for i in range(5, 9): assert collapsed_qttts[0].board[i].entangled_marks == [i] assert collapsed_qttts[0].board[i].entangled_blocks == [4] assert collapsed_qttts[0].board[4].entangled_blocks == [0, 1, 2, 3, 5, 6, 7, 8] assert collapsed_qttts[0].board[4].entangled_marks == [1, 2, 3, 4, 5, 6, 7, 8] assert (collapsed_qttts[0].ttt.board == np.ones(9) * Qttt.ttt.EMPTY_BLOCK_VAL).all() # r 10 p 1 s_env.step(s_env.get_valid_moves()[1][0], [0, 4]) _, collapsed_qttts, _ = s_env.get_valid_moves() for i in range(3): assert collapsed_qttts[1].board[i].entangled_marks == [i + 1] assert collapsed_qttts[1].board[i].entangled_blocks == [] for i in range(5, 9): assert collapsed_qttts[1].board[i].entangled_marks == [i] assert collapsed_qttts[1].board[i].entangled_blocks == [] assert collapsed_qttts[1].board[4].entangled_marks == [9] assert collapsed_qttts[1].board[4].entangled_blocks == [] assert (collapsed_qttts[1].ttt.board == np.array([1, 2, 3, 4, 9, 5, 6, 7, 8])).all() # r 10 p 1 s_env.change_to_even_pieces_view() _, collapsed_qttts, _ = s_env.get_valid_moves() for i in range(3): assert collapsed_qttts[1].board[i].entangled_marks == [i + 1] assert collapsed_qttts[1].board[i].entangled_blocks == [] for i in range(5, 9): assert collapsed_qttts[1].board[i].entangled_marks == [i] assert collapsed_qttts[1].board[i].entangled_blocks == [] assert collapsed_qttts[1].board[4].entangled_marks == [9] assert collapsed_qttts[1].board[4].entangled_blocks == [] assert (collapsed_qttts[1].ttt.board == np.array([1, 2, 3, 4, 9, 5, 6, 7, 8])).all()