app.py

import gradio as gr
import time
import math
from collections import deque, defaultdict

# --- 設定參數 ---
MAX_RECENT_EVENTS = 1000   # 暫存區大小 (達到此數量觸發壓縮)
GRID_PRECISION = 4        # 網格精度 (4小數點約=11公尺)
MAX_GRID_INTENSITY = 20.0 # 單一網格最大亮度 (避免單點過曝)

# --- 資料結構 ---
# 1. 近期事件 (Delta Queue): 用於快速同步
# 格式: {'t': time, 'u': user_id, 'l': [lat, lng], 's': speed}
recent_events = deque()

# 2. 歷史網格 (Snapshot Grid): 儲存壓縮後的長期記憶
# Key: (lat_round, lng_round), Value: intensity
archived_grid = defaultdict(float)

# 3. 狀態版本控制
snapshot_version = 0
last_snapshot_time = time.time()

def get_intensity_by_speed(speed_kmh):
    """
    根據速度決定迷霧驅散強度 (亮度)
    < 5 km/h (步行/停留): 強度 1.0
    < 20 km/h (跑步/腳踏車): 強度 0.5
    > 20 km/h (車輛): 強度 0.1
    """
    if speed_kmh is None: return 0.5
    if speed_kmh < 5: return 1.0
    if speed_kmh < 20: return 0.5
    return 0.1

def compress_events_to_grid(events_list):
    """將事件列表壓縮進網格"""
    global archived_grid
    for event in events_list:
        lat, lng = event['l']
        speed = event['s']
        
        # 網格化座標
        grid_key = (round(lat, GRID_PRECISION), round(lng, GRID_PRECISION))
        
        # 累加亮度
        intensity = get_intensity_by_speed(speed)
        archived_grid[grid_key] += intensity
        
        # 限制最大亮度 (防止單點無限疊加)
        if archived_grid[grid_key] > MAX_GRID_INTENSITY:
            archived_grid[grid_key] = MAX_GRID_INTENSITY

def trigger_snapshot():
    """檢查並執行快照壓縮"""
    global recent_events, snapshot_version, last_snapshot_time
    
    # 如果暫存區超過閾值，將最舊的一半資料壓縮進歷史網格
    if len(recent_events) > MAX_RECENT_EVENTS:
        events_to_archive = []
        count_to_remove = int(MAX_RECENT_EVENTS / 2)
        
        for _ in range(count_to_remove):
            events_to_archive.append(recent_events.popleft())
            
        compress_events_to_grid(events_to_archive)
        
        snapshot_version += 1
        last_snapshot_time = time.time()
        print(f"[Server] Snapshot Updated. Ver: {snapshot_version}, Grid Points: {len(archived_grid)}")

def sync_game_state(client_last_time, client_ver, user_id, lat, lng, speed):
    """
    主 API: 處理客戶端同步請求
    """
    current_time = time.time()
    
    # 1. 記錄用戶上傳的位置 (如果有)
    if lat is not None and lng is not None:
        # 過濾異常數據 (例如 GPS 飄移瞬間移動 > 150km/h)
        if speed is None or speed < 150: 
            new_event = {
                "t": current_time,
                "u": user_id,
                "l": [lat, lng],
                "s": speed
            }
            recent_events.append(new_event)
    
    # 2. 檢查是否需要觸發壓縮
    trigger_snapshot()
    
    # 3. 準備回傳資料
    response = {
        "server_time": current_time,
        "snapshot_version": snapshot_version,
        "sync_type": "delta",
        "payload": []
    }
    
    # 判斷同步策略
    # 如果客戶端版本落後 (需要全量更新) 或 客戶端剛加入 (ver = -1)
    # if client_ver < snapshot_version:
    if client_ver < 0:
        response["sync_type"] = "snapshot"
        # 回傳壓縮後的網格數據: [[lat, lng, intensity], ...]
        response["payload"] = [[k[0], k[1], v] for k, v in archived_grid.items()]
        
        # 同時附上還沒被壓縮的近期事件，確保數據不出現斷層
        response["recent_events"] = list(recent_events)
    else:
        # 增量更新: 只回傳 client_last_time 之後發生的事件
        # 注意: 這裡回傳的是未壓縮的原始事件
        delta = [e for e in recent_events if e['t'] > client_last_time]
        response["payload"] = delta

    return response

# --- Gradio 介面設定 ---
with gr.Blocks() as demo:
    gr.Markdown("## City Scout Game API Server")
    gr.Markdown("此 Space 為遊戲後端，請使用前端 HTML 連接。")
    
    with gr.Row(visible=False): # 隱藏 UI，僅作為 API 使用
        # Inputs
        in_time = gr.Number(label="Client Last Time")
        in_ver = gr.Number(label="Client Snapshot Version")
        in_uid = gr.Textbox(label="User ID")
        in_lat = gr.Number(label="Lat")
        in_lng = gr.Number(label="Lng")
        in_speed = gr.Number(label="Speed (km/h)")
        
        # Outputs
        out_json = gr.JSON(label="Sync Response")
        
    # API 按鈕
    btn_sync = gr.Button("Sync")
    btn_sync.click(
        fn=sync_game_state,
        inputs=[in_time, in_ver, in_uid, in_lat, in_lng, in_speed],
        outputs=out_json,
        api_name="sync" # 重要: 前端 client.predict("/sync") 用這個名字
    )

# 啟動並允許跨域 (CORS)
demo.queue(max_size=50).launch(cors_allowed_origins=["*"])