jetpack之paging源码解析

参考资料：
源码：https://mp.weixin.qq.com/s/aoBAuiuSRvfyCpnHGYrpwA
设计思路：https://mp.weixin.qq.com/s/QoKwTm0x6hssCzocnDp_1A

如果没有现在这么爱较真的心情，直接查看以上地址完全能够理解，而且写得也很好，以下也是参考这篇文章的。（我就是个搬代码的瓦匠工，如有雷同，时间在我前都算我借鉴的，以下字是我一行行敲的）

非常重要的原理图（网上拷贝）以及源码结构（自行剪切）

原理图基本上说明了paging分页的工作原理。源码确实类名比较多，根据原理图以及实际操作来按顺序讲解。

类介绍
这里先不做介绍，按照PagingWithNetworkSample demo走一遍，这个有的手机因为某些你懂得原因访问网络失败，那就重新找个demo，或者用pagingsample demo将就下。

PagingWithNetWorkSample案例：( kotlin编写的代码)

这里面有三个：
1.network only（by item）以这条为主线进行讲解
2.network only （by page）不讲，雷同by item
3.db+network（ roomdatabase）作为晋级来讲；

(1) network only(by item)
用户点击 NETWORK ONLY(BY ITEM) 按钮引发的血案…（我网络请求失败了，大爷的，那咱直接分析代码）：

默认输入KEY_SUBREDDIT 值导致调用repository.postsOfSubreddit(it, 30)方法

SubRedditViewModel.kt:

 companion object {
        const val KEY_SUBREDDIT = "subreddit"
        const val DEFAULT_SUBREDDIT = "androiddev"
    }

    init {
        if (!savedStateHandle.contains(KEY_SUBREDDIT)) {
            savedStateHandle.set(KEY_SUBREDDIT, DEFAULT_SUBREDDIT)
        }
    }

    private val repoResult = savedStateHandle.getLiveData<String>(KEY_SUBREDDIT).map {
        repository.postsOfSubreddit(it, 30) *注释①*
    }

repository.postsOfSubreddit方法逻辑如下
先SubRedditViewModel 方法调用val repo = ServiceLocator.instance(this@RedditActivity).getRepository(repoType)；

然后在ServiceLocator类中instance方法实例化DefaultServiceLocator（继承ServiceLocator）并且getRepository()方法，根据type调用InMemoryByItemRepository,其实以上代码注释①调用的就是这里的postsOfSubreddit方法。

RedditActivity.kt:

private val model: SubRedditViewModel by viewModels {
        object : AbstractSavedStateViewModelFactory(this, null) {
            override fun <T : ViewModel?> create(
                    key: String,
                    modelClass: Class<T>,
                    handle: SavedStateHandle
            ): T {
                val repoTypeParam = intent.getIntExtra(KEY_REPOSITORY_TYPE, 0)
                val repoType = RedditPostRepository.Type.values()[repoTypeParam]
                val repo = ServiceLocator.instance(this@RedditActivity)
                        .getRepository(repoType)
                @Suppress("UNCHECKED_CAST")
                return SubRedditViewModel(repo, handle) as T
            }
        }
    }
    
ServiceLocator.kt:

fun instance(context: Context): ServiceLocator {
            synchronized(LOCK) {
                if (instance == null) {
                    instance = DefaultServiceLocator(
                            app = context.applicationContext as Application,
                            useInMemoryDb = false)
                }
                return instance!!
            }
        }
        ...
        override fun getRepository(type: RedditPostRepository.Type): RedditPostRepository {
        return when (type) {
            RedditPostRepository.Type.IN_MEMORY_BY_ITEM -> InMemoryByItemRepository(
                    redditApi = getRedditApi(),
                    networkExecutor = getNetworkExecutor())
            RedditPostRepository.Type.IN_MEMORY_BY_PAGE -> InMemoryByPageKeyRepository(
                    redditApi = getRedditApi(),
                    networkExecutor = getNetworkExecutor())
            RedditPostRepository.Type.DB -> DbRedditPostRepository(
                    db = db,
                    redditApi = getRedditApi(),
                    ioExecutor = getDiskIOExecutor())
        }
    }
    
InMemoryByItemRepository.kt:

@MainThread
    override fun postsOfSubreddit(subReddit: String, pageSize: Int): Listing<RedditPost> {
        val sourceFactory = SubRedditDataSourceFactory(redditApi, subReddit, networkExecutor)

        // We use toLiveData Kotlin ext. function here, you could also use LivePagedListBuilder
        val livePagedList = sourceFactory.toLiveData(
                // we use Config Kotlin ext. function here, could also use PagedList.Config.Builder
                config = Config(
                        pageSize = pageSize,
                        enablePlaceholders = false,
                        initialLoadSizeHint = pageSize * 2),
                // provide custom executor for network requests, otherwise it will default to
                // Arch Components' IO pool which is also used for disk access
                fetchExecutor = networkExecutor)
				...
    }

Config是配置文件：这里配置每页请求个数，占位符为false（一般都为false，这个对我而言目前用不上，而且用起来不复杂，这里不表），首次请求为page*2

（源码）PagedListConfig.kt
@Suppress("FunctionName")
fun Config(
    pageSize: Int,
    prefetchDistance: Int = pageSize,
    enablePlaceholders: Boolean = true,
    initialLoadSizeHint: Int =
            pageSize * PagedList.Config.Builder.DEFAULT_INITIAL_PAGE_MULTIPLIER,
    maxSize: Int = PagedList.Config.MAX_SIZE_UNBOUNDED
): PagedList.Config {
    return PagedList.Config.Builder()
            .setPageSize(pageSize)
            .setPrefetchDistance(prefetchDistance)
            .setEnablePlaceholders(enablePlaceholders)
            .setInitialLoadSizeHint(initialLoadSizeHint)
            .setMaxSize(maxSize)
            .build()
}

SubRedditDataSourceFactory继承 DataSource.Factory，工厂用于创建DataSource，

class SubRedditDataSourceFactory(
        private val redditApi: RedditApi,
        private val subredditName: String,
        private val retryExecutor: Executor) : DataSource.Factory<String, RedditPost>() {
    val sourceLiveData = MutableLiveData<ItemKeyedSubredditDataSource>()
    override fun create(): DataSource<String, RedditPost> {*注释②