移动端弱网、离线和重试策略

📶 移动网络不可靠：切换、丢包、延迟时刻发生。居安思危的设计是找用户体验的预先。

弱网条件下的策略

请求超时与重试

// OkHttp 配置
val client = OkHttpClient.Builder()
    .connectTimeout(10, TimeUnit.SECONDS)
    .readTimeout(30, TimeUnit.SECONDS)
    .retryOnConnectionFailure(true)
    .addInterceptor(RetryInterceptor(maxRetries = 3))
    .build()

// 指数退灯重试
class RetryInterceptor(val maxRetries: Int) : Interceptor {
    override fun intercept(chain: Interceptor.Chain): Response {
        var attempt = 0
        while (true) {
            try {
                return chain.proceed(chain.request())
            } catch (e: IOException) {
                if (++attempt > maxRetries) throw e
                Thread.sleep((2.0.pow(attempt) * 1000).toLong())  // 指数退灯
            }
        }
    }
}

---

离线缓存策略

数据分层缓存

// Room 数据库作为本地缓存
class UserRepository(
    private val api: UserApi,
    private val dao: UserDao
) {
    fun getUsers(): Flow<List<User>> = flow {
        // 1. 先发射缓存数据
        val cached = dao.getAll()
        if (cached.isNotEmpty()) emit(cached)
        
        // 2. 拉取网络数据
        try {
            val fresh = api.getUsers()
            dao.insertAll(fresh)   // 更新缓存
            emit(fresh)
        } catch (e: IOException) {
            if (cached.isEmpty()) throw e  // 缓存也没有才投错
        }
    }
}

缓存有效期设计

@Entity
data class CachedData(
    val id: String,
    val data: String,
    val updatedAt: Long = System.currentTimeMillis()
)

fun CachedData.isExpired(ttlMs: Long = 5 * 60 * 1000): Boolean {
    return System.currentTimeMillis() - updatedAt > ttlMs
}

---

请求队列与离线演播

// 将遇到网络错误的请求存入一个队列并当网络恢复后重试
class RequestQueue(private val db: AppDatabase) {
    
    suspend fun enqueue(request: PendingRequest) {
        db.requestDao().insert(request)
    }
    
    suspend fun processQueue() {
        val pending = db.requestDao().getPending()
        pending.forEach { req ->
            try {
                executeRequest(req)
                db.requestDao().markComplete(req.id)
            } catch (e: IOException) {
                db.requestDao().incrementRetryCount(req.id)
            }
        }
    }
}

---

网络状态监听

// 监听网络变化
val networkCallback = object : ConnectivityManager.NetworkCallback() {
    override fun onAvailable(network: Network) {
        // 网络已恢复，处理队列
        lifecycleScope.launch { requestQueue.processQueue() }
    }
    
    override fun onLost(network: Network) {
        // 网络断开，切换离线模式
        showOfflineMode()
    }
}

connectivityManager.registerDefaultNetworkCallback(networkCallback)

---

常见误区

• 所有接口都用同一个重试次数，忽略了隐影更新操作不能重试
• 缓存没有设置过期时间，导致展示过时数据
• 离线队列没有限制尽小重试次数，网络恢复后大量请求市发