zhichun-project/CTF/backend/server0.py

import asyncio
import random
import heapq
import math
from lib.game_engine import GameMap, run_game_server
import threading
from collections import defaultdict

# 全局世界模型初始化
world = GameMap(show_gap_in_msec=1.0)
lock = threading.Lock()

# 游戏状态管理
player_roles = {}  # 玩家角色分配 {'player_name': 'attacker'/'defender'}
flag_assignments = {}  # flag分配 {'player_name': (x, y)}
attack_tasks = {}  # 攻击任务状态 {'player_name': 'attacking'/'returning'/'hiding'}
defender_positions = {}  # 防守位置分配 {'player_name': (x, y)}


def start_game(req):
    """游戏开始初始化"""
    global player_roles, flag_assignments, attack_tasks, defender_positions
    world.init(req)
    print("Start Game!!")

    # 重置所有全局状态
    player_roles = {}
    flag_assignments = {}
    attack_tasks = {}
    defender_positions = {}

    print(f"Game Started! Side: {'Left' if world.is_on_left(list(world.my_team_target)[0]) else 'Right'}")


def a_star(start, goal, dangerous_positions=None, hard_obstacles=None, opponents=None):
    """
    基于权重的寻路算法，区分软限制和硬限制
    危险位置：软限制（权重按距离opponent远近递增）
    对手位置和障碍物：硬限制（完全不能通过）
    """
    if start == goal:
        return [start]

    # 硬限制：墙壁、对手位置等完全不能通过的位置
    hard_blocked = set(world.walls)
    if hard_obstacles:
        hard_blocked.update(hard_obstacles)

    # 软限制：危险位置，权重按距离opponent远近递增
    dangerous_set = set(dangerous_positions) if dangerous_positions else set()

    # 获取opponents位置列表用于距离计算
    opponent_positions = []
    if opponents:
        opponent_positions = [(o["posX"], o["posY"]) if isinstance(o, dict) else o for o in opponents]
    elif hard_obstacles:
        opponent_positions = list(hard_obstacles)

    # 4个方向移动：上下左右
    directions = [(0, -1), (0, 1), (-1, 0), (1, 0)]  # 上, 下, 左, 右

    def is_valid(pos):
        """检查位置是否有效（不在在硬限制区域）"""
        x, y = pos
        return (0 <= x < world.width and 0 <= y < world.height and
                pos not in hard_blocked)

    def get_move_cost(pos):
        """获取移动到某位置的代价，危险区域权重按十字型阶梯递增，最小100"""
        if pos in dangerous_set and opponent_positions:
            # 计算到最近opponent的距离
            min_distance_to_opponent = min(
                calculate_manhattan_distance(pos, opp_pos)
                for opp_pos in opponent_positions
            )
            # 十字型阶梯权重：以对手为中心，每层距离对应一个权重阶梯，最小100
            if min_distance_to_opponent == 0:
                return 50000  # 对手位置：50000
            elif min_distance_to_opponent == 1:
                return 10000  # 距离1：10000
            elif min_distance_to_opponent == 2:
                return 5000  # 距离2：5000
            elif min_distance_to_opponent == 3:
                return 2000  # 距离3：2000
            elif min_distance_to_opponent == 4:
                return 1000  # 距离4：1000
            elif min_distance_to_opponent == 5:
                return 500  # 距离5：500
            elif min_distance_to_opponent == 6:
                return 300  # 距离6：300
            else:
                return 100  # 距离7+：最小100
        elif pos in dangerous_set:
            return 100  # 默认危险位置最小权重100
        
        if is_in_my_territory(pos):
            return 1
        else:
            return 5
        # return 1  # 正常代价

    def get_manhattan_distance(pos1, pos2):
        """计算曼哈顿距离作为启发式函数"""
        return abs(pos1[0] - pos2[0]) + abs(pos1[1] - pos2[1])

    # A*搜索主循环
    open_set = []
    heapq.heappush(open_set, (0, start))  # (f_score, pos)
    came_from = {}
    g_score = {start: 0}
    f_score = {start: get_manhattan_distance(start, goal)}
    closed_set = set()

    while open_set:
        current_f, current_pos = heapq.heappop(open_set)

        if current_pos in closed_set:
            continue
        closed_set.add(current_pos)

        if current_pos == goal:
            # 重构路径
            path = []
            pos = current_pos
            while pos in came_from:
                path.append(pos)
                pos = came_from[pos]
            path.append(start)
            return path[::-1]

        # 探索所有方向
        for direction in directions:
            next_pos = (current_pos[0] + direction[0], current_pos[1] + direction[1])

            if is_valid(next_pos) and next_pos not in closed_set:
                # 计算移动到下一位置的代价
                move_cost = get_move_cost(next_pos)
                tentative_g = g_score[current_pos] + move_cost

                if next_pos not in g_score or tentative_g < g_score[next_pos]:
                    came_from[next_pos] = current_pos
                    g_score[next_pos] = tentative_g
                    f_score[next_pos] = tentative_g + get_manhattan_distance(next_pos, goal)
                    heapq.heappush(open_set, (f_score[next_pos], next_pos))

    return []  # 无路径


def calculate_manhattan_distance(pos1, pos2):
    """计算曼哈顿距离"""
    return abs(pos1[0] - pos2[0]) + abs(pos1[1] - pos2[1])


def is_in_my_territory(pos):
    """检查位置是否在我方领域"""
    my_targets = list(world.list_targets(mine=True))
    if not my_targets:
        return True
    my_side_is_left = world.is_on_left(my_targets[0])
    return world.is_on_left(pos) == my_side_is_left


def is_in_enemy_territory(pos):
    """检查位置是否在敌方营地（敌方半场）"""
    return not is_in_my_territory(pos)


def get_midline_x():
    """获取地图中线X坐标（独立函数，避免重复计算）"""
    return world.width // 2


def is_about_to_cross_midline(current_pos, next_pos):
    """判断玩家是否即将越过中线（从己方半场进入敌方半场）"""
    # 获取我方半场标识（增加异常处理，防止崩溃）
    my_targets = list(world.list_targets(mine=True))
    if not my_targets:
        return False
    my_side_is_left = world.is_on_left(my_targets[0])
    midline_x = get_midline_x()

    # 我方在左侧：当前x < 中线，下一步x >= 中线 → 即将越界
    if my_side_is_left:
        current_in_my_territory = current_pos[0] < midline_x
        next_in_enemy_territory = next_pos[0] >= midline_x
    # 我方在右侧：当前x > 中线，下一步x <= 中线 → 即将越界
    else:
        current_in_my_territory = current_pos[0] > midline_x
        next_in_enemy_territory = next_pos[0] <= midline_x

    return current_in_my_territory and next_in_enemy_territory


def has_enemy_opposite(current_pos, next_pos, opponents, offset=1):
    """
    【核心修改】仅检测中线对面「贴着中线」的位置是否有敌方玩家
    贴着中线定义：敌方玩家的X坐标与中线X坐标的差值的绝对值 ≤ 1（即紧邻中线）
    同时保留Y轴小幅偏移（可调整offset），确保精准检测
    """
    # 前置校验：敌方玩家列表为空时，直接返回False
    if not opponents:
        return False
    # 1. 获取中线X坐标，定义「贴着中线」的判定条件
    midline_x = get_midline_x()
    STICK_TO_MIDLINE_THRESHOLD = 1  # 贴着中线：X坐标与中线差值≤1
    # 2. 即将越界位置的Y轴（判定基准）
    target_y = next_pos[1]

    # 遍历敌方玩家（增加数据有效性判断）
    for opponent in opponents:
        # 校验敌方玩家数据完整性，防止KeyError
        if not all(key in opponent for key in ["name", "posX", "posY"]):
            continue
        opp_pos = (opponent["posX"], opponent["posY"])

        # 【关键判定1】敌方玩家必须「贴着中线」（X坐标紧邻中线）
        if abs(opp_pos[0] - midline_x) > STICK_TO_MIDLINE_THRESHOLD:
            continue  # 非贴着中线的敌方，直接跳过
        # 【关键判定2】Y轴在允许偏移范围内（与即将越界位置的Y轴接近）
        if abs(opp_pos[1] - target_y) > offset:
            continue
        # 【关键判定3】敌方玩家在敌方半场（与玩家即将进入的区域一致，中线两侧）
        if is_in_my_territory(opp_pos):
            continue
        # 【关键判定4】敌方玩家与越界位置在中线两侧（严格对面）
        opp_side_left = opp_pos[0] < midline_x
        next_side_left = next_pos[0] < midline_x
        if opp_side_left == next_side_left:
            continue  # 同一侧，非对面

        # 满足所有条件（贴着中线+Y轴接近+敌方半场+中线两侧），判定为有风险
        print(f"[越界安全检测] 发现中线对面贴着中线的敌方玩家 {opponent['name']} 位置 {opp_pos}，禁止越界")
        return True
    return False


# 寻找敌方营地内的安全躲避位置
def find_safe_hide_position_in_enemy_territory(player_pos, opponents, enemy_flags):
    """
    在敌方营地内寻找安全躲避位置（远离所有敌方玩家，优先靠近目标旗帜）
    :param player_pos: 玩家当前位置
    :param opponents: 敌方玩家列表
    :param enemy_flags: 敌方旗帜列表
    :return: 安全躲避位置 / None
    """
    if not opponents:
        return player_pos  # 无敌方时，当前位置即为安全

    # 定义安全距离：与所有敌方玩家的曼哈顿距离≥3格
    SAFE_DISTANCE = 3
    enemy_territory_positions = []
    opponent_positions = [(o["posX"], o["posY"]) for o in opponents if all(key in o for key in ["posX", "posY"])]

    # 遍历敌方营地内所有有效位置（非墙壁）
    for x in range(world.width):
        for y in range(world.height):
            pos = (x, y)
            # 仅保留敌方营地+非墙壁位置
            if is_in_enemy_territory(pos) and pos not in world.walls:
                # 检查是否与所有敌方玩家保持安全距离
                is_safe = True
                for opp_pos in opponent_positions:
                    if calculate_manhattan_distance(pos, opp_pos) < SAFE_DISTANCE:
                        is_safe = False
                        break
                if is_safe:
                    enemy_territory_positions.append(pos)

    if not enemy_territory_positions:
        print("[躲避逻辑] 敌方营地无安全位置，优先向中线撤退")
        # 无安全位置时，返回靠近中线的位置
        midline_x = get_midline_x()
        retreat_pos = min(
            [(x, y) for x in range(world.width) for y in range(world.height) if
             is_in_enemy_territory((x, y)) and (x, y) not in world.walls],
            key=lambda p: abs(p[0] - midline_x)
        )
        return retreat_pos

    # 优先选择靠近目标旗帜的安全位置（为后续拿旗做准备）
    if enemy_flags:
        target_flag_positions = [(f["posX"], f["posY"]) for f in enemy_flags]
        # 按「距离最近目标旗帜」排序安全位置
        enemy_territory_positions.sort(
            key=lambda p: min(calculate_manhattan_distance(p, flag_pos) for flag_pos in target_flag_positions)
        )
    else:
        # 无旗帜时，按「距离当前玩家最近」排序
        enemy_territory_positions.sort(
            key=lambda p: calculate_manhattan_distance(p, player_pos)
        )

    # 返回最优安全躲避位置
    best_hide_pos = enemy_territory_positions[0]
    print(f"[躲避逻辑] 找到最优安全躲避位置: {best_hide_pos}（远离敌方，靠近目标旗帜）")
    return best_hide_pos


# 判断是否需要躲避敌方（进攻者在敌方营地时）
def need_hide_from_enemy(player_pos, opponents):
    """
    判断进攻者在敌方营地时是否需要躲避敌方
    :param player_pos: 玩家当前位置
    :param opponents: 敌方玩家列表
    :return: True=需要躲避（敌方过近）, False=无需躲避（安全状态）
    """
    if not opponents or not is_in_enemy_territory(player_pos):
        return False  # 无敌方 / 不在敌方营地，无需躲避

    # 危险距离：与任意敌方玩家曼哈顿距离<3格，判定为需要躲避
    DANGER_DISTANCE = 3
    opponent_positions = [(o["posX"], o["posY"]) for o in opponents if all(key in o for key in ["posX", "posY"])]
    for opp_pos in opponent_positions:
        if calculate_manhattan_distance(player_pos, opp_pos) < DANGER_DISTANCE:
            print(f"[躲避逻辑] 检测到敌方玩家过近（距离<{DANGER_DISTANCE}格），需要优先躲避")
            return True

    return False


def calculate_safe_and_dangerous_positions(my_pos, opponents):
    """
    计算安全位置和危险位置
    安全位置：己方player能在任意对方player之前提前一轮到达的位置
    己方半场的位置都是安全位置
    其余位置都是危险位置（包括周围一格范围）
    """
    safe_positions = set()
    base_dangerous_positions = set()

    print(f"正在计算从位置 {my_pos} 出发的安全/危险位置...")

    # 遍历地图上所有有效位置
    for x in range(world.width):
        for y in range(world.height):
            pos = (x, y)

            # 跳过墙壁
            if pos in world.walls:
                continue

            # 己方半场的位置都是安全位置
            if is_in_my_territory(pos):
                safe_positions.add(pos)
                continue

            # 计算己方到达该位置的距离
            my_distance = calculate_manhattan_distance(my_pos, pos)

            # 检查是否能比所有对手提前一轮到达
            is_safe = True
            for opponent in opponents:
                opponent_pos = (opponent["posX"], opponent["posY"])
                opponent_distance = calculate_manhattan_distance(opponent_pos, pos)

                # 如果对手能同时或更快到达，则不安全
                if opponent_distance <= my_distance:
                    is_safe = False
                    break

            if is_safe:
                safe_positions.add(pos)
            else:
                base_dangerous_positions.add(pos)

    # 将危险位置范围扩大一格
    expanded_dangerous_positions = set(base_dangerous_positions)
    for dangerous_pos in base_dangerous_positions:
        x, y = dangerous_pos
        # 检查周围8个方向的位置
        for dx in [-1, 0, 1]:
            for dy in [-1, 0, 1]:
                if dx == 0 and dy == 0:
                    continue  # 跳过自身位置

                neighbor_pos = (x + dx, y + dy)
                nx, ny = neighbor_pos

                # 检查边界和墙壁
                if (0 <= nx < world.width and 0 <= ny < world.height and
                        neighbor_pos not in world.walls and
                        not is_in_my_territory(neighbor_pos)):  # 己方半场仍然安全
                    expanded_dangerous_positions.add(neighbor_pos)

    # 从安全位置中移除被扩展为危险的位置（除了己方半场）
    final_safe_positions = set()
    for pos in safe_positions:
        if is_in_my_territory(pos) or pos not in expanded_dangerous_positions:
            final_safe_positions.add(pos)

    print(
        f"计算完成：基础危险位置 {len(base_dangerous_positions)} 个，扩展后危险位置 {len(expanded_dangerous_positions)} 个")
    print(f"最终安全位置 {len(final_safe_positions)} 个")

    return final_safe_positions, expanded_dangerous_positions


def find_closest_safe_position_to_target(my_pos, target_pos, safe_positions):
    """
    找到距离目标最近的安全位置
    这个位置必须能通过安全路径到达
    """
    if not safe_positions:
        return None

    print(f"寻找从 {my_pos} 到目标 {target_pos} 的最近安全位置...")

    # 如果目标本身就是安全位置，直接返回
    if target_pos in safe_positions:
        print(f"目标位置 {target_pos} 本身就是安全位置")
        return target_pos

    # 按距离目标的远近排序所有安全位置
    safe_positions_list = list(safe_positions)
    safe_positions_list.sort(key=lambda pos: calculate_manhattan_distance(pos, target_pos))

    # 找到第一个能通过安全路径到达的安全位置
    for safe_pos in safe_positions_list:
        # 检查从当前位置到这个安全位置的路径是否只经过安全位置
        path = a_star(my_pos, safe_pos, [])
        if path:
            # 检查路径上的所有位置是否都是安全的
            path_is_safe = all(pos in safe_positions for pos in path)
            if path_is_safe:
                print(f"找到最近安全位置: {safe_pos}, 距离目标 {calculate_manhattan_distance(safe_pos, target_pos)}")
                return safe_pos

    print("未找到可通过安全路径到达的安全位置")
    return None


def assign_player_roles(my_players, enemy_flags, my_flags, full_defense_mode=False):
    """
    动态分配玩家角色（攻击者/防守者）
    至少保持1个防守者，其余为攻击者
    当full_defense_mode=True时，所有玩家都转为防守者
    """
    global player_roles

    total_players = len(my_players)
    if total_players == 0:
        return

    # 完全防守模式：所有玩家都转为防守者
    if full_defense_mode:
        for p in my_players:
            player_roles[p["name"]] = "defender"
        print("进入完全防守模式，所有玩家均转为防守者")
        return

    # 计算需要的防守者数量
    min_defenders = 1
    enemy_threats = len([f for f in my_flags if f["canPickup"]])
    needed_defenders = max(min_defenders, min(enemy_threats, total_players // 2))

    # 清理已不存在的玩家角色
    current_players = {p["name"] for p in my_players}
    player_roles = {name: role for name, role in player_roles.items()
                    if name in current_players}

    # 为新玩家分配角色
    current_defenders = [name for name, role in player_roles.items()
                         if role == "defender"]
    current_attackers = [name for name, role in player_roles.items()
                         if role == "attacker"]

    unassigned = [p["name"] for p in my_players if p["name"] not in player_roles]

    # 如果防守者不足，从攻击者中转换
    while len(current_defenders) < needed_defenders and current_attackers:
        # 选择最适合防守的攻击者（距离我方目标最近的）
        my_targets = list(world.list_targets(mine=True))
        if my_targets:
            best_candidate = min(current_attackers,
                                 key=lambda name: min(calculate_manhattan_distance(
                                     (next(p for p in my_players if p["name"] == name)["posX"],
                                      next(p for p in my_players if p["name"] == name)["posY"]), target)
                                                      for target in my_targets))
            player_roles[best_candidate] = "defender"
            current_defenders.append(best_candidate)
            current_attackers.remove(best_candidate)

    # 分配未分配的玩家
    for name in unassigned:
        if len(current_defenders) < needed_defenders:
            player_roles[name] = "defender"
            current_defenders.append(name)
        else:
            player_roles[name] = "attacker"
            current_attackers.append(name)


def assign_defender_positions(defenders, my_players, my_flags, opponents, full_defense_mode=False):
    """
    为防守者分配防守位置 - 重构版本
    每轮前往距离任意opponent最近的flag，绝对不能到对方半场
    当full_defense_mode=True且防守者数量≥剩余旗帜数时，所有人员分开站在旗帜上不动
    """
    global defender_positions

    if not defenders:
        return

    my_targets = list(world.list_targets(mine=True))
    if not my_targets:
        return

    # 清空之前的分配
    defender_positions = {}

    # 只考虑我方半场的flags（绝对不能到对方半场）
    my_territory_flags = []
    for flag in my_flags:
        flag_pos = (flag["posX"], flag["posY"])
        if is_in_my_territory(flag_pos) and flag["canPickup"]:
            my_territory_flags.append(flag)

    print(f"我方半场可防守的flags: {[(f['posX'], f['posY']) for f in my_territory_flags]}")

    # 完全防守模式特殊处理
    if full_defense_mode and len(defenders) >= len(my_territory_flags) and my_territory_flags:
        print("完全防守模式：所有防守者分开站在旗帜上不动防守")
        # 为每个旗帜分配一个防守者
        assigned_flags = set()
        for i, defender_name in enumerate(defenders):
            # 循环分配旗帜
            flag_index = i % len(my_territory_flags)
            flag = my_territory_flags[flag_index]
            flag_pos = (flag["posX"], flag["posY"])

            # 确保每个旗帜只分配一个防守者
            if flag_pos not in assigned_flags:
                defender_positions[defender_name] = flag_pos
                assigned_flags.add(flag_pos)
                print(f"defender {defender_name} 被分配驻守flag {flag_pos}")
            else:
                # 如果旗帜已被分配，分配到目标区域巡逻
                target = my_targets[i % len(my_targets)]
                defender_positions[defender_name] = target
                print(f"defender {defender_name} 被分配到巡逻点 {target} (旗帜已被驻守)")
        return

    if not my_territory_flags:
        # 如果没有需要防守的flags，defenders在我方target区域巡逻
        print("没有需要防守的flags，defenders在我方领域巡逻")
        for i, defender_name in enumerate(defenders):
            target = my_targets[i % len(my_targets)]
            defender_positions[defender_name] = target
            print(f"defender {defender_name} 被分配到巡逻点 {target}")
        return

    # 为每个opponent找到距离它最近的我方flag（只检测不持有己方flag的opponent）
    threatened_flags = []
    for opponent in opponents:
        opponent_pos = (opponent["posX"], opponent["posY"])

        # 检查opponent是否持有我方flag
        if opponent.get("hasFlag", False):
            print(f"opponent {opponent['name']} 持有我方flag，跳过威胁检测")
            continue

        if my_territory_flags:
            # 找到距离这个opponent最近的我方flag
            closest_flag = min(my_territory_flags,
                               key=lambda f: calculate_manhattan_distance(
                                   (f["posX"], f["posY"]), opponent_pos))
            closest_flag_pos = (closest_flag["posX"], closest_flag["posY"])
            distance = calculate_manhattan_distance(closest_flag_pos, opponent_pos)

            # 检查opponent是否在我方半场
            is_opponent_in_my_territory = is_in_my_territory(opponent_pos)

            # 在己方半场的opponent优先级为0（最高优先级），否则按距离排序
            priority = 0 if is_opponent_in_my_territory else distance

            threatened_flags.append({
                'flag_pos': closest_flag_pos,
                'opponent': opponent,
                'distance': distance,
                'priority': priority,
                'in_my_territory': is_opponent_in_my_territory
            })

            territory_status = "己方半场" if is_opponent_in_my_territory else "敌方半场"
            print(
                f"opponent {opponent['name']} 在 {opponent_pos} ({territory_status}) 威胁flag {closest_flag_pos}, 距离: {distance}, 优先级: {priority}")

    # 按威胁优先级排序（距离最近的威胁优先处理）
    threatened_flags.sort(key=lambda x: x['priority'])

    # 为威胁最大的flags分配defenders，考虑时间优势
    assigned_flags = set()
    for i, defender_name in enumerate(defenders):
        if i < len(threatened_flags):
            threat = threatened_flags[i]
            new_flag_pos = threat['flag_pos']

            # 获取defender当前位置
            defender = next((p for p in my_players if p["name"] == defender_name), None)
            if not defender:
                continue

            defender_pos = (defender["posX"], defender["posY"])

            # 检查是否应该切换到新的flag
            should_switch = True

            # 如果defender之前已有目标，检查是否值得切换
            if defender_name in defender_positions:
                current_target = defender_positions[defender_name]

                # 计算defender到当前目标和新目标的距离
                distance_to_current = calculate_manhattan_distance(defender_pos, current_target)
                distance_to_new = calculate_manhattan_distance(defender_pos, new_flag_pos)

                # 计算opponent到新flag的距离
                opponent_to_new = threat['distance']

                # 只有当defender能在opponent前到达新flag，且新威胁更紧急时才切换
                if distance_to_new >= opponent_to_new:
                    should_switch = False
                    print(
                        f"defender {defender_name} 保持当前目标 {current_target}，因为无法及时赶到新威胁 {new_flag_pos}")
                    print(
                        f"  当前距离: {distance_to_current}, 新目标距离: {distance_to_new}, opponent距离: {opponent_to_new}")
                else:
                    print(f"defender {defender_name} 切换到新威胁 {new_flag_pos}，能够及时拦截")
                    print(f"  defender距离: {distance_to_new}, opponent距离: {opponent_to_new}")

            if should_switch:
                # 避免多个defender保护同一个flag（除非defender太多）
                if new_flag_pos not in assigned_flags or len(assigned_flags) >= len(my_territory_flags):
                    defender_positions[defender_name] = new_flag_pos
                    assigned_flags.add(new_flag_pos)
                    print(
                        f"defender {defender_name} 被分配保护flag {new_flag_pos}, 对抗opponent {threat['opponent']['name']} (距离: {threat['distance']})")
                else:
                    # 如果新flag已有人守护，保持当前目标或选择其他flag
                    if defender_name in defender_positions:
                        print(
                            f"defender {defender_name} 保持当前目标 {defender_positions[defender_name]}，新flag已有其他defender")
                    else:
                        remaining_flags = [f for f in my_territory_flags
                                           if (f["posX"], f["posY"]) not in assigned_flags]
                        if remaining_flags:
                            flag = remaining_flags[0]
                            flag_pos = (flag["posX"], flag["posY"])
                            defender_positions[defender_name] = flag_pos
                            assigned_flags.add(flag_pos)
                            print(f"defender {defender_name} 被分配到备选flag {flag_pos}")
                        else:
                            target = my_targets[i % len(my_targets)]
                            defender_positions[defender_name] = target
                            print(f"defender {defender_name} 被分配到巡逻点 {target}")
            else:
                defender_positions[defender_name] = current_target
        else:
            # 多余的defenders在target区域巡逻
            target = my_targets[i % len(my_targets)]
            defender_positions[defender_name] = target
            print(f"额外defender {defender_name} 被分配到巡逻点 {target}")

    print(f"最终defender分配: {defender_positions}")


def assign_attacker_targets(attackers, my_players, enemy_flags, opponents):
    """为攻击者分配攻击目标"""
    global flag_assignments, attack_tasks

    if not attackers or not enemy_flags:
        return

    my_targets = list(world.list_targets(mine=True))
    if not my_targets:
        return

    # 获取opponent位置集合
    opponent_positions_set = {(o["posX"], o["posY"]) for o in opponents}

    # 清理无效的分配
    active_attackers = set(attackers)
    flag_assignments = {name: pos for name, pos in flag_assignments.items()
                        if name in active_attackers}
    attack_tasks = {name: status for name, status in attack_tasks.items()
                    if name in active_attackers}

    # 检查并更新现有的flag分配
    attackers_need_reassign = []
    for attacker_name in attackers:
        player = next((p for p in my_players if p["name"] == attacker_name), None)
        if not player:
            continue

        # 如果玩家带flag，更新为返回状态
        if player["hasFlag"]:
            attack_tasks[attacker_name] = "returning"
            if attacker_name in flag_assignments:
                del flag_assignments[attacker_name]
            continue

        # 检查当前分配的flag上是否站着opponent
        if attacker_name in flag_assignments:
            assigned_flag = flag_assignments[attacker_name]
            if assigned_flag in opponent_positions_set:
                print(f"攻击者 {attacker_name} 的目标flag {assigned_flag} 上站着opponent，需要重新分配")
                del flag_assignments[attacker_name]
                if attacker_name in attack_tasks:
                    del attack_tasks[attacker_name]
                attackers_need_reassign.append(attacker_name)
            else:
                print(f"攻击者 {attacker_name} 的目标flag {assigned_flag} 安全")
        else:
            # 没有分配目标的攻击者
            attackers_need_reassign.append(attacker_name)

    # 为需要重新分配的攻击者选择new flag
    assigned_flags = set(flag_assignments.values())
    available_flags = [(f["posX"], f["posY"]) for f in enemy_flags
                       if (f["posX"], f["posY"]) not in assigned_flags]

    # 筛选出没有opponent站着的flag
    safe_flags = [flag_pos for flag_pos in available_flags
                  if flag_pos not in opponent_positions_set]

    print(f"可用flags: {available_flags}")
    print(f"安全flags（无opponent）: {safe_flags}")

    for attacker_name in attackers_need_reassign:
        player = next((p for p in my_players if p["name"] == attacker_name), None)
        if not player:
            continue

        player_pos = (player["posX"], player["posY"])

        if safe_flags:
            # 选择寻路分数最低的安全flag
            print(f"为攻击者 {attacker_name} 在 {player_pos} 从安全flags中选择目标...")

            # 计算当前玩家的危险位置
            safe_positions, dangerous_positions = calculate_safe_and_dangerous_positions(player_pos, opponents)

            best_flag = None
            best_score = float('inf')

            for flag_pos in safe_flags:
                # 使用权重寻路计算到达该flag的实际代价
                opponent_positions_list = [(o["posX"], o["posY"]) for o in opponents]
                path = a_star(player_pos, flag_pos, dangerous_positions, opponent_positions_list)

                if path:
                    # 计算路径的总权重代价
                    total_cost = 0
                    dangerous_set = set(dangerous_positions)

                    for pos in path:
                        if pos in dangerous_set:
                            total_cost += 10000  # 危险位置代价
                        else:
                            total_cost += 1  # 正常位置代价

                    print(f"  安全Flag {flag_pos}: 路径长度={len(path)}, 总代价={total_cost}")

                    if total_cost < best_score:
                        best_score = total_cost
                        best_flag = flag_pos
                else:
                    print(f"  安全Flag {flag_pos}: 无法到达")

            if best_flag:
                flag_assignments[attacker_name] = best_flag
                attack_tasks[attacker_name] = "attacking"
                safe_flags.remove(best_flag)
                print(f"  选择最优安全目标: {best_flag}, 代价: {best_score}")
            else:
                print(f"  WARNING: 攻击者 {attacker_name} 无法到达任何安全flag")
        else:
            # 没有安全flag，尝试选择有opponent占据的flag
            print(f"攻击者 {attacker_name} 没有安全flag可选，尝试选择有opponent占据的flag")

            occupied_flags = [flag_pos for flag_pos in available_flags
                              if flag_pos in opponent_positions_set]

            if occupied_flags:
                # 检查当前是否已经在攻击某个被占据的flag
                current_assignment = flag_assignments.get(attacker_name)
                if current_assignment in occupied_flags:
                    # 检查距离，如果靠近3格opponent还没让开就换目标
                    distance_to_flag = calculate_manhattan_distance(player_pos, current_assignment)
                    if distance_to_flag <= 3:
                        print(
                            f"  攻击者 {attacker_name} 距离被占据flag {current_assignment} 仅{distance_to_flag}格，opponent未让开，寻找替代目标")

                        # 寻找其他可选的flag
                        other_flags = [f for f in occupied_flags if f != current_assignment]
                        other_safe_flags = [f for f in available_flags if
                                            f not in opponent_positions_set and f != current_assignment]

                        if other_safe_flags:
                            # 优先选择安全flag
                            best_alternative = min(other_safe_flags,
                                                   key=lambda f: calculate_manhattan_distance(player_pos, f))
                            flag_assignments[attacker_name] = best_alternative
                            attack_tasks[attacker_name] = "attacking"
                            print(f"  切换到安全备选flag: {best_alternative}")
                        elif other_flags:
                            # 选择其他被占据的flag
                            best_alternative = min(other_flags,
                                                   key=lambda f: calculate_manhattan_distance(player_pos, f))
                            flag_assignments[attacker_name] = best_alternative
                            attack_tasks[attacker_name] = "attacking"
                            print(f"  切换到其他被占据flag: {best_alternative}")
                        else:
                            # 没有其他选择，返回己方半场
                            print(f"  无其他flag可选，攻击者 {attacker_name} 返回己方半场")
                            _assign_retreat_position(attacker_name, player_pos, opponent_positions_set)
                    else:
                        print(
                            f"  攻击者 {attacker_name} 继续攻击被占据flag {current_assignment}, 距离: {distance_to_flag}")
                else:
                    # 为新攻击者分配被占据的flag
                    best_occupied_flag = min(occupied_flags,
                                             key=lambda f: calculate_manhattan_distance(player_pos, f))
                    flag_assignments[attacker_name] = best_occupied_flag
                    attack_tasks[attacker_name] = "attacking"
                    print(f"  分配被占据flag: {best_occupied_flag}")
            else:
                # 没有任何可用flag，向己方半场撤退
                print(f"攻击者 {attacker_name} 没有任何flag可选，向己方半场撤退")
                _assign_retreat_position(attacker_name, player_pos, opponent_positions_set)


def _assign_retreat_position(attacker_name, player_pos, opponent_positions_set):
    """为攻击者分配撤退位置"""
    # 在己方半场选择一个随机安全位置
    my_territory_positions = []
    for x in range(world.width):
        for y in range(world.height):
            pos = (x, y)
            if (is_in_my_territory(pos) and pos not in world.walls and
                    pos not in opponent_positions_set):
                my_territory_positions.append(pos)

    if my_territory_positions:
        random_target = random.choice(my_territory_positions)
        flag_assignments[attacker_name] = random_target
        attack_tasks[attacker_name] = "retreating"
        print(f"  攻击者 {attacker_name} 选择随机撤退位置: {random_target}")
    else:
        print(f"  WARNING: 攻击者 {attacker_name} 无法找到安全撤退位置")


def plan_next_actions(req):
    """主要的游戏策略规划函数（优化：进攻者在敌方营地优先躲敌，再拿旗）"""
    if not world.update(req):
        return {}

    print(f"\n=== 游戏回合 {req.get('time', 'unknown')} ===")

    global player_roles, flag_assignments, attack_tasks, defender_positions

    # 获取游戏状态
    my_players = world.list_players(mine=True, inPrison=False, hasFlag=None)
    opponents = world.list_players(mine=False, inPrison=False, hasFlag=None)
    enemy_flags = world.list_flags(mine=False, canPickup=True)
    my_flags = world.list_flags(mine=True, canPickup=True)
    my_targets = list(world.list_targets(mine=True))

    # 计算双方剩余旗帜数量
    my_remaining_flags = len(my_flags)
    enemy_remaining_flags = len(enemy_flags)
    print(f"我方剩余旗帜: {my_remaining_flags}, 敌方剩余旗帜: {enemy_remaining_flags}")

    # 检查是否需要进入完全防守模式
    full_defense_mode = False
    if my_remaining_flags <= 3 and enemy_remaining_flags <= my_remaining_flags:
        full_defense_mode = True
        print(
            f"触发完全防守模式: 我方剩余旗帜={my_remaining_flags} ≤3 且 敌方剩余旗帜={enemy_remaining_flags} ≤ 我方旗帜数")

    print(f"我方玩家: {[(p['name'], p['posX'], p['posY'], p['hasFlag']) for p in my_players]}")
    print(f"敌方玩家: {[(p['name'], p['posX'], p['posY']) for p in opponents]}")
    print(f"敌方flags: {[(f['posX'], f['posY']) for f in enemy_flags]}")
    print(f"我方targets: {my_targets}")

    if not my_players or not my_targets:
        print("没有我方玩家或目标，返回空移动")
        return {}

    # 1. 动态角色分配（支持完全防守模式）
    assign_player_roles(my_players, enemy_flags, my_flags, full_defense_mode)
    print(f"角色分配: {player_roles}")

    # 2. 获取当前角色分配
    defenders = [name for name, role in player_roles.items() if role == "defender"]
    attackers = [name for name, role in player_roles.items() if role == "attacker"]
    print(f"防守者: {defenders}, 攻击者: {attackers}")

    # 3. 分配防守和攻击任务（防守任务支持完全防守模式）
    assign_defender_positions(defenders, my_players, my_flags, opponents, full_defense_mode)
    # 如果不是完全防守模式才分配攻击任务
    if not full_defense_mode:
        assign_attacker_targets(attackers, my_players, enemy_flags, opponents)
    else:
        # 完全防守模式下清空攻击任务
        flag_assignments = {}
        attack_tasks = {}
        print("完全防守模式：清空所有攻击任务")

    print(f"防守位置: {defender_positions}")
    print(f"攻击目标: {flag_assignments}")
    print(f"攻击任务状态: {attack_tasks}")

    # 4. 为每个玩家计算移动
    player_moves = {}

    for player in my_players:
        player_name = player["name"]
        current_pos = (player["posX"], player["posY"])
        is_attacker = player_name in attackers
        is_in_enemy = is_in_enemy_territory(current_pos)
        has_flag = player["hasFlag"]

        print(f"\n--- 处理玩家 {player_name} 位置 {current_pos} ---")
        print(
            f"[{player_name}] 身份: {'攻击者' if is_attacker else '防守者'}, 所在区域: {'敌方营地' if is_in_enemy else '己方营地'}, 持有旗帜: {has_flag}")

        # 【最高优先级】强制攻击逻辑：如果距离opponent仅1格且opponent在我方半场，无论什么状态都必须向opponent移动
        adjacent_opponent = None
        closest_distance = float('inf')

        for opponent in opponents:
            opponent_pos = (opponent["posX"], opponent["posY"])
            distance = calculate_manhattan_distance(current_pos, opponent_pos)

            # 检查是否相邻（距离为1）且opponent在我方半场
            if distance == 1 and is_in_my_territory(opponent_pos):
                # 如果有多个相邻入侵者，选择最近的一个
                if distance < closest_distance:
                    closest_distance = distance
                    adjacent_opponent = opponent
                    print(f"{player_name} 发现距离{distance}格的入侵者 {opponent['name']} 在我方半场 {opponent_pos}")

        # 强制攻击入侵者 - 无论玩家当前状态是什么（攻击者、防守者、带flag等）
        if adjacent_opponent:
            opponent_pos = (adjacent_opponent["posX"], adjacent_opponent["posY"])
            move = world.get_direction(current_pos, opponent_pos)
            if move:
                player_moves[player_name] = move
                print(
                    f"【强制攻击】{player_name} 无论当前状态如何，必须攻击入侵者 {adjacent_opponent['name']}: {current_pos} -> {opponent_pos}, 方向: {move}")
                continue
            else:
                print(f"ERROR: {player_name} 无法向入侵者移动 - 这不应该发生!")

        # 【优化核心逻辑】进攻者在敌方营地：优先判断是否需要躲避敌方
        if is_attacker and is_in_enemy and not has_flag:
            if need_hide_from_enemy(current_pos, opponents):
                # 1. 需要躲避：更新任务状态为躲避
                attack_tasks[player_name] = "hiding"
                # 2. 寻找敌方营地内的安全躲避位置
                hide_pos = find_safe_hide_position_in_enemy_territory(current_pos, opponents, enemy_flags)
                if hide_pos:
                    # 3. 规划前往躲避位置的路径
                    opponent_positions = [(o["posX"], o["posY"]) for o in opponents]
                    path = a_star(current_pos, hide_pos, [], opponent_positions)
                    if path and len(path) > 1:
                        next_pos = path[1]
                        move = world.get_direction(current_pos, next_pos)
                        if move:
                            player_moves[player_name] = move
                            print(
                                f"【优先躲避】{player_name} 前往安全躲避位置: {current_pos} -> {next_pos} (最终目标: {hide_pos})")
                            continue
                print(f"【躲避失败】{player_name} 无法找到躲避路径，执行原拿旗逻辑")

        # 确定目标位置（原有逻辑，躲避优先级高于此）
        target_pos = None
        task_type = "unknown"

        if player["hasFlag"]:
            # 带flag回家 - 选择加权分值最低的target
            print(f"{player_name} 正在选择最优回程目标（基于加权分值）...")

            # 计算当前位置的危险区域（用于加权计算）
            safe_positions, dangerous_positions = calculate_safe_and_dangerous_positions(current_pos, opponents)
            dangerous_set = set(dangerous_positions)

            best_target = None
            lowest_weighted_score = float('inf')

            for target in my_targets:
                # 计算到每个target的加权路径
                opponent_positions = [(o["posX"], o["posY"]) for o in opponents]
                path = a_star(current_pos, target, dangerous_positions, opponent_positions)

                if path:
                    # 计算路径的总加权代价
                    total_weighted_cost = 0
                    for pos in path:
                        if pos in dangerous_set:
                            total_weighted_cost += 10000  # 危险位置代价
                        else:
                            total_weighted_cost += 1  # 正常位置代价

                    print(f"  Target {target}: 路径长度={len(path)}, 加权代价={total_weighted_cost}")

                    if total_weighted_cost < lowest_weighted_score:
                        lowest_weighted_score = total_weighted_cost
                        best_target = target
                else:
                    print(f"  Target {target}: 无法到达")

            if best_target:
                target_pos = best_target
                print(f"{player_name} 选择最优回程目标: {target_pos}, 加权代价: {lowest_weighted_score}")
            else:
                # 如果所有target都无法到达，选择第一个作为备选
                target_pos = my_targets[0]
                print(f"WARNING: {player_name} 无法到达任何target，使用默认目标: {target_pos}")

            task_type = "returning_flag"
        elif player_name in defender_positions:
            # 防守者前往防守位置
            target_pos = defender_positions[player_name]
            task_type = "defending"

            # 完全防守模式且已到达旗帜位置则不移动
            if (full_defense_mode and
                    len(defenders) >= len(my_flags) and
                    current_pos == target_pos):
                print(f"{player_name} 已在防守位置 {target_pos}，完全防守模式下不移动")
                continue
        elif player_name in flag_assignments:
            # 攻击者前往目标flag（仅当无需躲避时执行）
            target_pos = flag_assignments[player_name]
            task_type = "attacking"

        print(f"{player_name} 任务类型: {task_type}, 目标位置: {target_pos}")

        if not target_pos or current_pos == target_pos:
            print(f"{player_name} 没有目标或已到达目标，跳过")
            continue

        # 检查是否在我方领域
        in_my_territory = is_in_my_territory(current_pos)
        print(f"{player_name} 是否在我方领域: {in_my_territory}")

        # 对于attacker，使用权重寻路策略（软性约束危险位置，硬性约束对手位置）
        if task_type == "attacking":
            print(f"攻击者 {player_name} 使用权重寻路策略")

            # 计算当前的危险位置
            safe_positions, dangerous_positions = calculate_safe_and_dangerous_positions(current_pos, opponents)
            print(f"{player_name} 检测到 {len(dangerous_positions)} 个危险位置")

            # 对手位置作为硬限制（完全不能通过）
            opponent_positions = [(o["posX"], o["posY"]) for o in opponents]
            print(f"{player_name} 硬限制对手位置: {opponent_positions}")

            # 直接向目标flag移动：危险位置软限制，对手位置硬限制
            path = a_star(current_pos, target_pos, dangerous_positions, opponent_positions, opponents)
            print(f"{player_name} 权重寻路路径: {path}")

            if not path:
                # 如果权重寻路失败，使用传统寻路作为备选
                print(f"{player_name} 权重寻路失败，尝试传统寻路")
                path = world.route_to(current_pos, target_pos, opponent_positions)
        else:
            # 对于defender或带flag的玩家，使用传统策略
            print(f"{player_name} 使用传统寻路策略")

            extra_obstacles = []
            if not in_my_territory:
                # 在敌方领域时避免与对手直接冲突
                extra_obstacles = [(o["posX"], o["posY"]) for o in opponents]
                print(f"{player_name} 在敌方领域，避开 {len(extra_obstacles)} 个对手位置")

            # 使用A*算法寻路
            path = a_star(current_pos, target_pos, extra_obstacles)

            # 如果A*失败，使用原始寻路算法作为备选
            if not path:
                print(f"{player_name} A*寻路失败，尝试原始BFS算法")
                path = world.route_to(current_pos, target_pos, extra_obstacles)

        print(f"{player_name} 最终路径: {path}")

        if len(path) > 1:
            next_pos = path[1]
            current_pos = (player["posX"], player["posY"])
            player_name = player["name"]

            # ===================== 嵌入：越界安全检测（无bug） =====================
            try:
                # 1. 检测是否即将越界
                about_to_cross = is_about_to_cross_midline(current_pos, next_pos)
                if about_to_cross:
                    # 2. 检测中线对面贴着中线的敌方
                    enemy_opposite = has_enemy_opposite(current_pos, next_pos, opponents)
                    if enemy_opposite:
                        # 3. 有敌方则禁止移动，原地待命
                        print(f"[{player_name}] 即将越界遇中线对面贴着中线的敌方，停止移动（安全防护）")
                        continue
            except Exception as e:
                # 捕获异常不影响程序运行
                print(f"[{player_name}] 越界检测临时异常：{e}，继续正常移动")
            # =====================================================================

            # 原有移动逻辑
            move = world.get_direction(current_pos, next_pos)
            print(f"{player_name} 下一步: {current_pos} -> {next_pos}, 方向: {move}")
            if move:
                player_moves[player_name] = move
            else:
                print(f"WARNING: {player_name} 无法确定移动方向!")
        else:
            print(f"WARNING: {player_name} 没有找到有效路径!")

    print(f"最终移动指令: {player_moves}")
    print("=" * 50)
    return player_moves


def game_over(req):
    print("Game Over!")
    world.show(force=True)


async def main():
    import sys
    if len(sys.argv) != 2:
        print(f"Usage: python3 {sys.argv[0]} <port>")
        print(f"Example: python3 {sys.argv[0]} 8080")
        sys.exit(1)

    port = int(sys.argv[1])
    print(f"AI backend running on port {port} ...")

    try:
        await run_game_server(port, start_game, plan_next_actions, game_over)
    except Exception as e:
        print(f"Server Stopped: {e}")
        sys.exit(1)


if __name__ == "__main__":
    asyncio.run(main())