datamllab / rlcard / 315

pending completion

Build # 315

Build Type

push

travis-ci

Committed by

web-flow

Commit Message

<a href="https://github.com/datamllab/rlcard/commit/e1b11d327">Merge pull request #17 from datamllab/rule_doudizhu

Rule of kicker for doudizhu

Former-commit-id: </a><a class="double-link" href="https://github.com/datamllab/rlcard/commit/17a43f709">17a43f709</a>

Run Details

22 of 22 new or added lines in 1 file covered. (100.0%)

3293 of 3448 relevant lines covered (95.5%)

0.96 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

96.92

/rlcard/envs/doudizhu.py

import numpy as np

from rlcard.utils.utils import *
from rlcard.envs.env import Env
from rlcard.games.doudizhu.game import DoudizhuGame as Game
from rlcard.games.doudizhu.utils import SPECIFIC_MAP, CARD_RANK_STR
from rlcard.games.doudizhu.utils import ACTION_LIST, ACTION_SPACE
from rlcard.games.doudizhu.utils import encode_cards


class DoudizhuEnv(Env):
    ''' Doudizhu Environment
    '''

    def __init__(self, allow_step_back=False):
        super().__init__(Game(allow_step_back), allow_step_back)
        self.state_shape = [6, 5, 15]

    def extract_state(self, state):
        ''' Encode state

        Args:
            state (dict): dict of original state

        Returns:
            numpy array: 6*5*15 array
                         6 : current hand
                             the union of the other two players' hand
                             the recent three actions
                             the union of all played cards
        '''
        obs = np.zeros((6, 5, 15), dtype=int)
        for index in range(6):
            obs[index][0] = np.ones(15, dtype=int)
        encode_cards(obs[0], state['current_hand'])
        encode_cards(obs[1], state['others_hand'])
        for i, action in enumerate(state['trace'][-3:]):
            if action[1] != 'pass':
                encode_cards(obs[4-i], action[1])
        if state['played_cards'] is not None:
            encode_cards(obs[5], state['played_cards'])

        extrated_state = {'obs': obs, 'legal_actions': self.get_legal_actions()}
        return extrated_state

    def get_payoffs(self):
        ''' Get the payoffs of players. Must be implemented in the child class.

        Returns:
            payoffs (list): a list of payoffs for each player
        '''
        return self.game.judger.judge_payoffs(self.game.round.landlord_id, self.game.winner_id)

    def decode_action(self, action_id):
        ''' Action id -> the action in the game. Must be implemented in the child class.

        Args:
            action_id (int): the id of the action

        Returns:
            action (string): the action that will be passed to the game engine.
        '''
        abstract_action = ACTION_LIST[action_id]
        # without kicker
        if '*' not in abstract_action:
            return abstract_action
        # with kicker
        legal_actions = self.game.state['actions']
        specific_actions = []
        kickers = []
        for legal_action in legal_actions:
            for abstract in SPECIFIC_MAP[legal_action]:
                main = abstract.strip('*')
                if abstract == abstract_action:
                    specific_actions.append(legal_action)
                    kickers.append(legal_action.replace(main, '', 1))
                    break
        # choose kicker with minimum score
        player_id = self.game.get_player_id()
        kicker_scores = []
        for kicker in kickers:
            score = 0
            for action in self.game.judger.playable_cards[player_id]:
                if kicker in action:
                    score += 1
            kicker_scores.append(score+CARD_RANK_STR.index(kicker[0]))
        min_index = 0
        min_score = kicker_scores[0]
        for index, score in enumerate(kicker_scores):
            if score < min_score:
                min_score = score
                min_index = index
        return specific_actions[min_index]

    def get_legal_actions(self):
        ''' Get all legal actions for current state

        Returns:
            legal_actions (list): a list of legal actions' id
        '''
        legal_action_id = []
        legal_actions = self.game.state['actions']
        if legal_actions:
            for action in legal_actions:
                for abstract in SPECIFIC_MAP[action]:
                    action_id = ACTION_SPACE[abstract]
                    if action_id not in legal_action_id:
                        legal_action_id.append(action_id)
        return legal_action_id

1	import numpy as np	1✔
2
3	from rlcard.utils.utils import *	1✔
4	from rlcard.envs.env import Env	1✔
5	from rlcard.games.doudizhu.game import DoudizhuGame as Game	1✔
6	from rlcard.games.doudizhu.utils import SPECIFIC_MAP, CARD_RANK_STR	1✔
7	from rlcard.games.doudizhu.utils import ACTION_LIST, ACTION_SPACE	1✔
8	from rlcard.games.doudizhu.utils import encode_cards	1✔
9
10
11	class DoudizhuEnv(Env):	1✔
12	''' Doudizhu Environment
13	'''
14
15	def __init__(self, allow_step_back=False):	1✔
16	super().__init__(Game(allow_step_back), allow_step_back)	1✔
17	self.state_shape = [6, 5, 15]	1✔
18
19	def extract_state(self, state):	1✔
20	''' Encode state
21
22	Args:
23	state (dict): dict of original state
24
25	Returns:
26	numpy array: 6515 array
27	6 : current hand
28	the union of the other two players' hand
29	the recent three actions
30	the union of all played cards
31	'''
32	obs = np.zeros((6, 5, 15), dtype=int)	1✔
33	for index in range(6):	1✔
34	obs[index][0] = np.ones(15, dtype=int)	1✔
35	encode_cards(obs[0], state['current_hand'])	1✔
36	encode_cards(obs[1], state['others_hand'])	1✔
37	for i, action in enumerate(state['trace'][-3:]):	1✔
38	if action[1] != 'pass':	1✔
39	encode_cards(obs[4-i], action[1])	1✔
40	if state['played_cards'] is not None:	1✔
41	encode_cards(obs[5], state['played_cards'])	1✔
42
43	extrated_state = {'obs': obs, 'legal_actions': self.get_legal_actions()}	1✔
44	return extrated_state	1✔
45
46	def get_payoffs(self):	1✔
47	''' Get the payoffs of players. Must be implemented in the child class.
48
49	Returns:
50	payoffs (list): a list of payoffs for each player
51	'''
52	return self.game.judger.judge_payoffs(self.game.round.landlord_id, self.game.winner_id)	1✔
53
54	def decode_action(self, action_id):	1✔
55	''' Action id -> the action in the game. Must be implemented in the child class.
56
57	Args:
58	action_id (int): the id of the action
59
60	Returns:
61	action (string): the action that will be passed to the game engine.
62	'''
63	abstract_action = ACTION_LIST[action_id]	1✔
64	# without kicker
65	if '*' not in abstract_action:	1✔
66	return abstract_action	1✔
67	# with kicker
68	legal_actions = self.game.state['actions']	1✔
69	specific_actions = []	1✔
70	kickers = []	1✔
71	for legal_action in legal_actions:	1✔
72	for abstract in SPECIFIC_MAP[legal_action]:	1✔
73	main = abstract.strip('*')	1✔
74	if abstract == abstract_action:	1✔
75	specific_actions.append(legal_action)	1✔
76	kickers.append(legal_action.replace(main, '', 1))	1✔
77	break	1✔
78	# choose kicker with minimum score
79	player_id = self.game.get_player_id()	1✔
80	kicker_scores = []	1✔
81	for kicker in kickers:	1✔
82	score = 0	1✔
83	for action in self.game.judger.playable_cards[player_id]:	1✔
84	if kicker in action:	1✔
85	score += 1	1✔
86	kicker_scores.append(score+CARD_RANK_STR.index(kicker[0]))	1✔
87	min_index = 0	1✔
88	min_score = kicker_scores[0]	1✔
89	for index, score in enumerate(kicker_scores):	1✔
90	if score < min_score:	1✔
91	min_score = score	×
92	min_index = index	×
93	return specific_actions[min_index]	1✔
94
95	def get_legal_actions(self):	1✔
96	''' Get all legal actions for current state
97
98	Returns:
99	legal_actions (list): a list of legal actions' id
100	'''
101	legal_action_id = []	1✔
102	legal_actions = self.game.state['actions']	1✔
103	if legal_actions:	1✔
104	for action in legal_actions:	1✔
105	for abstract in SPECIFIC_MAP[action]:	1✔
106	action_id = ACTION_SPACE[abstract]	1✔
107	if action_id not in legal_action_id:	1✔
108	legal_action_id.append(action_id)	1✔
109	return legal_action_id	1✔

datamllab / rlcard / 315

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous