Android中一次性获取Flow值效率低，如何优化代码？

架构老林

2026-7-8

优化方案及代码调整

核心问题分析

现有代码的主要低效点在于：Fragment中嵌套collect会导致每次isDynamicDataLoaded变为true时都创建新的Flow订阅，造成资源冗余；同时数据转换逻辑放在Fragment中违背MVVM职责划分；Room的Flow默认持续监听数据库变化，而你只需要一次性获取就绪后的数据。

优化后的ViewModel实现

将数据组合、状态判断、Model转换逻辑全部移至ViewModel，同时控制Flow仅在数据就绪后发射一次结果：

class StoreModeHomeViewModel : ViewModel() {
    // 组合后的UI数据Flow：仅在两个API完成后发射一次结果
    val combinedOffers: Flow<Pair<List<PersonalOfferModel>, List<FavoriteOfferModel>>> = combine(
        PersonalOffersRepository.instance.getPersonalOffers(),
        ProductsRepository.instance.getFavoriteOfferProductsLiveData().asFlow(),
        getRequestCompletionState()
    ) { personalOffers, favoriteDBProducts, isAllRequestsCompleted ->
        if (isAllRequestsCompleted) {
            // 完成实体到UI Model的转换
            val personalModels = personalOffers.map { PersonalOfferModel(it) }
            val favoriteModels = favoriteDBProducts.map { FavoriteOfferModel(Product.mapFromDB(it)) }
            personalModels to favoriteModels
        } else {
            null
        }
    }.filterNotNull() // 过滤未就绪的无效数据
    .take(1) // 仅发射一次结果，避免持续监听数据库

    // 对外暴露加载状态Flow
    val isLoading: Flow<Boolean> = getRequestCompletionState().map { !it }
        .distinctUntilChanged() // 避免重复发射相同状态

    private fun getRequestCompletionState(): Flow<Boolean> {
        val personalRequestDone = PersonalOffersRepository.instance.getOffersApiRequestStatus()
            .map { it?.status == ApiRequestStatus.COMPLETED }
            .distinctUntilChanged()
        
        val favoriteRequestDone = ProductsRepository.instance.getFavoriteOffersApiRequestStatusLiveData()
            .map { it?.status == ApiRequestStatus.COMPLETED }
            .distinctUntilChanged()

        return combine(personalRequestDone, favoriteRequestDone) { pDone, fDone ->
            pDone && fDone
        }
    }

    fun loadData(restCaller: RestCaller) = CoreApplication.instance?.applicationContext?.let {
        val workers = arrayListOf(        
            OneTimeWorkRequest.Builder(GetPersonalOffersWorker::class.java)
                .setInputData(restCaller.toData())
                .setCancelable()
                .build(),
            OneTimeWorkRequest.Builder(GetFavoriteOfferProductsWorker::class.java)
                .setInputData(restCaller.toData())
                .setCancelable()
                .build(),
        )
        WorkManager.getInstance(it)
            .beginUniqueWork("store_mode_data_task", ExistingWorkPolicy.KEEP, workers)
            .enqueue()
    }
}

优化后的Fragment实现

移除嵌套collect，直接订阅ViewModel提供的状态和数据Flow，逻辑更清晰：

lifecycleScope.launch {
    repeatOnLifecycle(Lifecycle.State.STARTED) {
        // 监听加载状态，控制进度条显示
        viewModel.isLoading.collect { isLoading ->
            binding.progressBar.visibility = if (isLoading) View.VISIBLE else View.GONE
        }
    }
}

lifecycleScope.launch {
    repeatOnLifecycle(Lifecycle.State.STARTED) {
        // 订阅就绪后的组合数据，更新UI
        viewModel.combinedOffers.collect { (personalModels, favoriteModels) ->
            personalOffersModelList.apply {
                clear()
                addAll(personalModels)
            }
            updateTabCounter(SwimLaneItemType.PERSONAL_OFFERS, personalModels.size)

            favoriteOffersModelList.apply {
                clear()
                addAll(favoriteModels)
            }
            updateTabCounter(SwimLaneItemType.FAVOURITE_OFFERS, favoriteModels.size)

            updateTabCounter(SwimLaneItemType.ALL, personalModels.size + favoriteModels.size)
            selectTabAt(selectedTab)
            updateOrderPickUpItemView()
        }
    }
}

关键优化点说明

移除嵌套订阅：避免重复创建Flow收集器，减少资源浪费和UI重复更新
ViewModel承担数据处理：将实体转Model的逻辑从Fragment移至ViewModel，符合MVVM架构职责
控制Flow发射次数：使用take(1)仅获取一次就绪后的数据，无需持续监听数据库变化
合并状态与数据流：直接在ViewModel中结合API请求状态和数据Flow，确保仅在数据就绪时发射有效结果
去重状态发射：用distinctUntilChanged过滤重复的加载状态，减少不必要的UI刷新

内容的提问来源于stack exchange，提问作者Taimoor Khan