def play_game(agent1, agent2, k, mod, deb):
global debug_
debug_ = deb
debug("Game begin")
debug("Agent1 initializing...")
agent1.reinit()
debug("Agent1 init done")
debug("Agent2 initializing...")
agent2.reinit()
debug("Agent2 init done")
l1, r1, l2, r2 = 1, 1, 1, 1
turn = 0
step = 0
se = set()
step_limit = k * k * k
while not (end(l1, r1, l2, r2)) and step < step_limit:
se.add((l1, r1, l2, r2, turn))
action_list = generate_available_action(k, l1, r1, l2, r2, turn, se,
mod)
if len(action_list) == 0:
break
if turn == 0:
action = agent1.next_step(l1, r1, l2, r2, action_list)
l1, r1, l2, r2, _ = generate_from_action(k, l1, r1, l2, r2, 0,
action_list[action][0],
action_list[action][1])
#print(action_list)
else:
action = agent2.next_step(l2, r2, l1, r1, action_list)
l2, r2, l1, r1, _ = generate_from_action(k, l2, r2, l1, r1, 0,
action_list[action][0],
action_list[action][1])
step = step + 1
turn = 1 - turn
debug(
"Step %d, Player %d use action (%s,%s), get state (%d,%d,%d,%d)" %
(step, 2 - turn, ma[action_list[action][0]],
ma[action_list[action][1]], l1, r1, l2, r2))
if step >= step_limit:
result = 0
debug("Draw")
elif turn == 0:
result = 2
debug("Agent 2 wins")
else:
result = 1
debug("Agent 1 wins")
return result, step
def action():
op1 = int(request.args["op"]) // 2
op2 = int(request.args["op"]) % 2
game_id = request.args["gameid"]
if game_list[game_id]["history"][-1][0] < 0:
return redirect(
url_for("human_agent",
gameid=game_id,
step=len(game_list[game_id]["history"])))
state = request.args["state"].replace("(", "").replace(")", "").replace(
" ", "").split(",")
l1 = int(state[0])
r1 = int(state[1])
l2 = int(state[2])
r2 = int(state[3])
turn = int(state[4])
# print("haha1", l1, r1, l2, r2, turn)
l1, r1, l2, r2, turn = generate_from_action(k, l1, r1, l2, r2, turn, op1,
op2)
# print("haha2", l1, r1, l2, r2, turn)
step = len(game_list[game_id]["history"]) + 2
game_list[game_id]["history"].append((l1, r1, l2, r2, turn))
return redirect(url_for("human_agent", gameid=game_id, step=step))
def simulate(self, p):
pse = set(list(self.se))
if self.mod == 2:
pse.add(p.state)
depth = p.depth
l1, r1, l2, r2, turn = p.state
lose = False
while not (end(l1, r1, l2, r2)) and depth <= self.k**2:
action_list = generate_available_action(self.k, l1, r1, l2, r2,
turn, pse, self.mod)
if len(action_list) == 0:
lose = True
break
op1, op2 = action_list[random.randint(0, len(action_list) - 1)]
l1, r1, l2, r2, turn = generate_from_action(
self.k, l1, r1, l2, r2, turn, op1, op2)
depth += 1
if self.mod == 2:
pse.add((l1, r1, l2, r2, turn))
if end(l1, r1, l2, r2) or lose:
if turn == p.state[4]:
return 0
else:
return 2
else:
return 1
def generate_next(self, l1, r1, l2, r2, turn):
action_list = generate_available_action(self.k, l1, r1, l2, r2, turn,
set(), self.mod)
for op1, op2 in action_list:
news = generate_from_action(self.k, l1, r1, l2, r2, turn, op1, op2)
self.next[(l1, r1, l2, r2, turn)].append(news)
self.pre[news].append((l1, r1, l2, r2, turn))
def next_step(self, l1, r1, l2, r2, action_list):
if self.mod == 2:
return self.agent.next_step(l1, r1, l2, r2, action_list)
s = self.reform_state(l1, r1, l2, r2, 0)
lose_s = []
none_s = []
win_s = []
for idx in range(0, len(action_list)):
op1, op2 = action_list[idx]
ns = generate_from_action(self.k, l1, r1, l2, r2, 0, op1, op2)
nexts = self.reform_state(ns[0], ns[1], ns[2], ns[3], ns[4])
if self.f[nexts] is None:
none_s.append(idx)
elif self.f[nexts] is False:
lose_s.append(idx)
else:
win_s.append(idx)
if len(lose_s) > 0:
return lose_s[random.randint(0, len(lose_s) - 1)]
elif len(none_s) > 0:
return none_s[random.randint(0, len(none_s) - 1)]
else:
return win_s[random.randint(0, len(win_s) - 1)]
def next_step(self, l1, r1, l2, r2, action_list):
s = (l1, r1, l2, r2, 0)
lose_s = []
none_s0 = []
none_s1 = []
win_s = []
for idx in range(0, len(action_list)):
op1, op2 = action_list[idx]
nexts = generate_from_action(self.k, l1, r1, l2, r2, 0, op1, op2)
if self.f[nexts] is None:
if nexts in self.next_cnt.keys():
if self.next_cnt[nexts][0]:
none_s0.append(idx)
else:
none_s1.append(idx)
else:
none_s1.append(idx)
elif self.f[nexts] is False:
lose_s.append(idx)
else:
win_s.append(idx)
if len(lose_s) > 0:
return lose_s[random.randint(0, len(lose_s) - 1)]
elif len(none_s0) > 0:
return none_s0[random.randint(0, len(none_s0) - 1)]
elif len(none_s1) > 0:
return none_s1[random.randint(0, len(none_s1) - 1)]
else:
return win_s[random.randint(0, len(win_s) - 1)]
def expand(self, p):
l1, r1, l2, r2, turn = p.state
action_list = generate_available_action(self.k, l1, r1, l2, r2, turn,
self.se, self.mod)
p.next = []
for op1, op2 in action_list:
news = generate_from_action(self.k, l1, r1, l2, r2, turn, op1, op2)
p.next.append(Node(news, p.depth + 1))
result = self.simulate(p)
# print(result)
p.n += 1
p.v += result
return result
def check(self, l1, r1, l2, r2, turn, action_list):
for a in range(0, len(action_list)):
news = generate_from_action(self.k, l1, r1, l2, r2, turn,
action_list[a][0], action_list[a][1])
if news != self.root.next[a].state:
raise NotImplementedError
def gamer(game_id, fb, method, mod):
f = open(os.path.join(app.config["RESULT_DIR"], game_id + ".txt"), "w")
if fb == 0:
print("human vs %s" % method, file=f)
else:
print("%s vs human" % method, file=f)
print("Mod %d" % mod, file=f)
agent = generate_agent(method)(k, mod)
step_limit = 75
wait_second = 0
max_wait_second = 100
while True:
l = len(game_list[game_id]["history"])
if l % 2 != game_list[game_id]["turn"]:
wait_second += 1
if wait_second > max_wait_second:
game_list[game_id]["history"].append((-6, -6, -6, -6, -6))
print("human leaves, game ends", file=f)
break
time.sleep(1)
continue
print("from machine", game_list[game_id]["history"])
state = game_list[game_id]["history"][-1]
wait_second = 0
if state[0] == -4:
print("human grant lose, game ends", file=f)
break
if state[0] == -2 or state[0] == -7:
print("human loses, game ends", file=f)
break
game_list[game_id]["set"].add(state)
l1, r1, l2, r2, turn = state
print("%d,%d,%d,%d" % (l1, r1, l2, r2), file=f)
action_list = generate_available_action(k, l1, r1, l2, r2, turn,
game_list[game_id]["set"], mod)
if len(action_list) == 0 or end(l1, r1, l2, r2):
game_list[game_id]["history"].append((-1, -1, -1, -1, -1))
print("machine loses,game ends", file=f)
break
if turn == 0:
action = agent.next_step(l1, r1, l2, r2, action_list)
l1, r1, l2, r2, turn = generate_from_action(
k, l1, r1, l2, r2, turn, action_list[action][0],
action_list[action][1])
else:
action = agent.next_step(l2, r2, l1, r1, action_list)
l1, r1, l2, r2, turn = generate_from_action(
k, l1, r1, l2, r2, turn, action_list[action][0],
action_list[action][1])
print("%d,%d,%d,%d" % (l1, r1, l2, r2), file=f)
game_list[game_id]["set"].add((l1, r1, l2, r2, turn))
game_list[game_id]["history"].append((l1, r1, l2, r2, turn))
if end(l1, r1, l2, r2):
game_list[game_id]["history"].append((-2, -2, -2, -2, -2))
print("human loses,game ends", file=f)
break
if len(game_list[game_id]["history"]) >= step_limit:
game_list[game_id]["history"].append((-3, -3, -3, -3, -3))
print("draw, game ends", file=f)
break
f.close()
print("Game %s ends" % game_id)