如何取消正在运行的 LiveData 协程块

Question

通过使用 LiveData 的最新版本“androidx.lifecycle:lifecycle-livedata-ktx:2.2.0-alpha03”，我使用 LiveData 的新构建块（LiveData +协程），它使用 Retrofit 执行同步网络调用，并相应地更新 ViewModel 中的不同标志（isLoading、isError）。我在“查询”LiveData 上使用 Transforamtions.switchMap，因此每当 UI 中的“查询”发生变化时，“搜索产品”代码就会使用 Transformations.switchMap 开始执行。一切正常，除了我想在“查询”LiveData 发生变化时取消之前的改造调用。目前我看不到任何方法可以做到这一点。任何帮助，将不胜感激。

class ProductSearchViewModel : ViewModel() {
    val completableJob = Job()
    private val coroutineScope = CoroutineScope(Dispatchers.IO + completableJob)

    // Query Observable Field
    val query: MutableLiveData<String> = MutableLiveData()

    // IsLoading Observable Field
    private val _isLoading = MutableLiveData<Boolean>()
    val isLoading: LiveData<Boolean> = _isLoading


    val products: LiveData<List<ProductModel>> = query.switchMap { q ->
        liveData(context = coroutineScope.coroutineContext) {
            emit(emptyList())
            _isLoading.postValue(true)
            val service = MyApplication.getRetrofitService()
            val response = service?.searchProducts(q)
            if (response != null && response.isSuccessful && response.body() != null) {
                _isLoading.postValue(false)
                val body = response.body()
                if (body != null && body.results != null) {
                    emit(body.results)
                }
            } else {
                _isLoading.postValue(false)
            }
        }
    }
}

Answer 1

你可以通过两种方式解决这个问题：

方法#1（简单方法）

就像 Mel 在他的 answer 中解释的那样，您可以在 switchMap 之外保留对作业实例的引用，并在 switchMap 中返回新的 liveData 之前取消该作业的实例。

class ProductSearchViewModel : ViewModel() {

    // Job instance
    private var job = Job()

    val products = Transformations.switchMap(_query) {
        job.cancel() // Cancel this job instance before returning liveData for new query
        job = Job() // Create new one and assign to that same variable

        // Pass that instance to CoroutineScope so that it can be cancelled for next query
        liveData(CoroutineScope(job + Dispatchers.IO).coroutineContext) { 
            // Your code here
        }
    }

    override fun onCleared() {
        super.onCleared()
        job.cancel()
    }
}

方法#2（不是那么干净，但独立且可重复使用）

由于liveData {} 构建器块在协程范围内运行，您可以使用CompletableDeffered 和协程launch 构建器的组合来暂停该liveData 块并手动观察query liveData 以启动网络请求的作业。

class ProductSearchViewModel : ViewModel() {

    private val _query = MutableLiveData<String>()

    val products: LiveData<List<String>> = liveData {
        var job: Job? = null // Job instance to keep reference of last job

        // LiveData observer for query
        val queryObserver = Observer<String> {
            job?.cancel() // Cancel job before launching new coroutine
            job = GlobalScope.launch {
                // Your code here
            }
        }

        // Observe query liveData here manually
        _query.observeForever(queryObserver)

        try {
            // Create CompletableDeffered instance and call await.
            // Calling await will suspend this current block 
            // from executing anything further from here
            CompletableDeferred<Unit>().await()
        } finally {
            // Since we have called await on CompletableDeffered above, 
            // this will cause an Exception on this liveData when onDestory
            // event is called on a lifeCycle . By wrapping it in 
            // try/finally we can use this to know when that will happen and 
            // cleanup to avoid any leaks.
            job?.cancel()
            _query.removeObserver(queryObserver)
        }
    }
}

您可以在此demo project 中下载并测试运行这两种方法

编辑：更新方法 #1 以在 onCleared 方法上添加作业取消，正如 yasir 在 cmets 中指出的那样。

Answer 2

当父作用域被取消时，改造请求应该被取消。

class ProductSearchViewModel : ViewModel() {
    val completableJob = Job()
    private val coroutineScope = CoroutineScope(Dispatchers.IO + completableJob)

    /**
     * Adding job that will be used to cancel liveData builder.
     * Be wary - after cancelling, it'll return a new one like:
     *
     *     ongoingRequestJob.cancel() // Cancelled
     *     ongoingRequestJob.isActive // Will return true because getter created a new one
     */
    var ongoingRequestJob = Job(coroutineScope.coroutineContext[Job])
        get() = if (field.isActive) field else Job(coroutineScope.coroutineContext[Job])

    // Query Observable Field
    val query: MutableLiveData<String> = MutableLiveData()

    // IsLoading Observable Field
    private val _isLoading = MutableLiveData<Boolean>()
    val isLoading: LiveData<Boolean> = _isLoading


    val products: LiveData<List<ProductModel>> = query.switchMap { q ->
        liveData(context = ongoingRequestJob) {
            emit(emptyList())
            _isLoading.postValue(true)
            val service = MyApplication.getRetrofitService()
            val response = service?.searchProducts(q)
            if (response != null && response.isSuccessful && response.body() != null) {
                _isLoading.postValue(false)
                val body = response.body()
                if (body != null && body.results != null) {
                    emit(body.results)
                }
            } else {
                _isLoading.postValue(false)
            }
        }
    }
}

那么你需要在需要的时候取消ongoingRequestJob。下次触发liveData(context = ongoingRequestJob) 时，由于它会返回一个新作业，它应该可以正常运行。您只需要在需要的地方取消它，即在query.switchMap 函数范围内。