標(biāo)簽： android browser download
簡介
當(dāng)我們用瀏覽器點(diǎn)開一個(gè)下載鏈接，然后去下載，從宏觀上認(rèn)識，有下載進(jìn)度的實(shí)時(shí)更新和界面的跳轉(zhuǎn)。整個(gè)過程中，主要涉及到以下過程。瀏覽器點(diǎn)擊下載按鈕，瀏覽器分發(fā)下去一個(gè)下載請求，跳轉(zhuǎn)界面的同時(shí)在DownloadProvider進(jìn)程中去真正的下載數(shù)據(jù)以及更新數(shù)據(jù)庫，在界面上監(jiān)聽數(shù)據(jù)庫的變化，去實(shí)時(shí)更新相關(guān)進(jìn)度。全過程中，Browser進(jìn)程負(fù)責(zé)分發(fā)下載請求，DownloadProvider進(jìn)程負(fù)責(zé)真正的下載操作。
目前而言，主要有兩種結(jié)構(gòu)，C-S和B-S結(jié)構(gòu)。對于Browser來說，主要在于對Webview這個(gè)控件的認(rèn)識，底層的內(nèi)核實(shí)現(xiàn)也是非常復(fù)雜，這里我們不做討論。對于一個(gè)瀏覽器鏈接，webkit底層會(huì)去解析，同時(shí)也會(huì)判斷這個(gè)鏈接屬于什么類型。比如我們今天的這個(gè)下載鏈接，Browser就有專門的下載監(jiān)聽器去回調(diào)執(zhí)行這個(gè)action，下面我們會(huì)詳細(xì)分析。
WebView控件簡單介紹
WebView控件提供了一個(gè)內(nèi)嵌的瀏覽器試圖，用于顯示本地的html或網(wǎng)路上的網(wǎng)頁。
并且比較強(qiáng)大的是，還可以直接跟js相互調(diào)用。
WebView有兩個(gè)方法：setWebChromeClient和setWebClient
WebChromeClient：主要處理解析，渲染網(wǎng)頁等瀏覽器做的事情，也是輔助WebView處理Javascript 的對話框，網(wǎng)站圖標(biāo)，網(wǎng)站title，加載進(jìn)度等
WebViewClient ：就是幫助WebView處理各種通知、請求事件的。
Browser下載的時(shí)序圖。
這里寫圖片描述
下面來詳細(xì)分析具體的代碼實(shí)現(xiàn)細(xì)節(jié)，時(shí)序圖是更加細(xì)節(jié)的步驟，這里我們著重分析下面的流程。
Step 1：Tab.setWebView
void setWebView(WebView w, boolean restore) {
....
mMainView = w;
// attach the WebViewClient, WebChromeClient and DownloadListener
if (mMainView != null) {
mMainView.setWebViewClient(mWebViewClient);
mMainView.setWebChromeClient(mWebChromeClient);
mMainView.setDownloadListener(mDownloadListener);
....
}
}
這個(gè)方法定義在packages/apps/Browser/src/com/android/browser/Tab.java
瀏覽器是用過Webview來顯示UI。這里設(shè)置了一個(gè)WebView對象，然后setWebViewClient和setWebChromeClient主要設(shè)置了對頁面加載以及js的處理。這里我們只分析setDownloadListener這個(gè)監(jiān)聽，首先要理解一點(diǎn)，對于WebView上的一個(gè)下載按鈕，它的事件是怎么處理的，瀏覽器如何判斷這個(gè)是下載？以上其實(shí)瀏覽器內(nèi)核已經(jīng)處理，瀏覽器內(nèi)核是根據(jù)指定的url判斷該鏈接是否是一個(gè)下載鏈接，如果點(diǎn)擊的是一個(gè)下載鏈接，那么最終會(huì)回調(diào)到該監(jiān)聽器中去處理，具體底層實(shí)現(xiàn)比較復(fù)雜，暫不作討論。
Tab(WebViewController wvcontroller, WebView w, Bundle state) {
/// M: add for save page
....
mDownloadListener = new BrowserDownloadListener() {
public void onDownloadStart(String url, String userAgent,
String contentDisposition, String mimetype, String referer,
long contentLength) {
/// M: add for fix download page url
mCurrentState.mIsDownload = true;
mWebViewController.onDownloadStart(Tab.this, url, userAgent, contentDisposition,
mimetype, referer, contentLength);
}
};
....
setWebView(w);
....
}
這個(gè)方法定義在packages/apps/Browser/src/com/android/browser/Tab.java
分析Tab的構(gòu)造方法，這里主要看BrowserDownloadListener這個(gè)對象。當(dāng)點(diǎn)擊了下載按鈕，則會(huì)去回調(diào)BrowserDownloadListener的onDownloadStart方法，這個(gè)最終是委托給了mWebViewController去處理。
Step 2：WebViewController.onDownloadStart
@Override
public void onDownloadStart(Tab tab, String url, String userAgent,
String contentDisposition, String mimetype, String referer,
long contentLength) {
....
DownloadHandler.onDownloadStart(mActivity, url, userAgent,
contentDisposition, mimetype, referer, false, contentLength);
...
}
這個(gè)方法定義在packages/apps/Browser/src/com/android/browser/Controller.java
WebViewController是一個(gè)接口，Controller是它的具體實(shí)現(xiàn)，在onDownloadStart方法中，實(shí)現(xiàn)比較簡單，直接是將參數(shù)委托給DownloadHandler的靜態(tài)方法onDownloadStart去進(jìn)一步處理。
在這里，參數(shù)：
url下載的網(wǎng)址鏈接
userAgent瀏覽器userAgent信息
mimetype下載內(nèi)容的type類型
contentLength下載內(nèi)容大小
Step 3：DownloadHandler.onDownloadStart
這個(gè)方法定義在packages/apps/Browser/src/com/android/browser/DownloadHandler.java
實(shí)現(xiàn)很簡單，直接將參數(shù)繼續(xù)傳遞到onDownloadStartNoStream方法。
Step 4：DownloadHandler.onDownloadStartNoStream
/*package */
public static void onDownloadStartNoStream(Activity activity,
String url, String userAgent, String contentDisposition,
String mimetype, String referer, boolean privateBrowsing, long contentLength) {
....
// java.net.URI is a lot stricter than KURL so we have to encode some
// extra characters. Fix for b 2538060 and b 1634719
WebAddress webAddress;
try {
webAddress = new WebAddress(url);
webAddress.setPath(encodePath(webAddress.getPath()));
} catch (Exception e) {
// This only happens for very bad urls, we want to chatch the
// exception here
Log.e(LOGTAG, "Exception trying to parse url:" + url);
return;
}
String addressString = webAddress.toString();
Uri uri = Uri.parse(addressString);
final DownloadManager.Request request = new DownloadManager.Request(uri);
request.setMimeType(mimetype);
// let this downloaded file be scanned by MediaScanner - so that it can
// show up in Gallery app, for example.
request.allowScanningByMediaScanner();
request.setDescription(webAddress.getHost());
// XXX: Have to use the old url since the cookies were stored using the
// old percent-encoded url.
String cookies = CookieManager.getInstance().getCookie(url, privateBrowsing);
request.addRequestHeader("cookie", cookies);
request.addRequestHeader("User-Agent", userAgent);
request.addRequestHeader("Referer", referer);
request.setNotificationVisibility(
DownloadManager.Request.VISIBILITY_VISIBLE_NOTIFY_COMPLETED);
request.setUserAgent(userAgent);
final DownloadManager manager = (DownloadManager)
activity.getSystemService(Context.DOWNLOAD_SERVICE);
new Thread("Browser download") {
public void run() {
manager.enqueue(request);
}
}.start();
/// M: Add to start Download activity. @{
Intent pageView = new Intent(DownloadManager.ACTION_VIEW_DOWNLOADS);
pageView.setFlags(Intent.FLAG_ACTIVITY_NEW_TASK);
activity.startActivity(pageView);
/// @}
}
這個(gè)方法定義在packages/apps/Browser/src/com/android/browser/DownloadHandler.java
在該方法中，主要做了三件事
1.將下載信息url,minetype等封裝成一個(gè)Request對象，供后續(xù)使用。
2.獲取一個(gè)DownloadManager對象，將前面封裝的Request對象，安排到下載隊(duì)列
3.開始下載的同時(shí)，去跳轉(zhuǎn)UI界面，同步顯示UI信息。
這里我們重點(diǎn)分析數(shù)據(jù)流程這塊，接下來分析enqueue這個(gè)方法的具體實(shí)現(xiàn)。
Step 5：DownloadManager.enqueue
public long enqueue(Request request) {
ContentValues values = request.toContentValues(mPackageName);
Uri downloadUri = mResolver.insert(Downloads.Impl.CONTENT_URI, values);
if (downloadUri != null) {
long id = Long.parseLong(downloadUri.getLastPathSegment());
return id;
}
return -1;
}
這個(gè)方法定義在frameworks/base/core/java/android/app/DownloadManager.java
首先toContentValues將Request的信息要存數(shù)據(jù)庫的字段轉(zhuǎn)化為一個(gè)ContentValues對象，以上幾步都是在Browser進(jìn)程中進(jìn)行的，接下來insert方法，通過uri開始最終跨進(jìn)程請求去插入數(shù)據(jù)。這里Downloads.Impl.CONTENT_URI為content://downloads/my_downloads，從pacakges/providers/DownloadProvider的清單文件中很容易知道最終是調(diào)用了DownloadProvider的insert方法去插入數(shù)據(jù)。
pacakges/providers/DownloadProvider的清單文件如下：
....
<provider android:name=".DownloadProvider"
android:authorities="downloads" android:exported="true">
....
Step 6：DownloadProvider.insert
@Override
public Uri insert(final Uri uri, final ContentValues values) {
....
long rowID = db.insert(DB_TABLE, null, filteredValues);
if (rowID == -1) {
Log.d(Constants.TAG, "couldn't insert into downloads database");
return null;
}
insertRequestHeaders(db, rowID, values);
notifyContentChanged(uri, match);
// Always start service to handle notifications and/or scanning
final Context context = getContext();
context.startService(new Intent(context, DownloadService.class));
return ContentUris.withAppendedId(Downloads.Impl.CONTENT_URI, rowID);
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadProvider.java
insert方法即是往DB_TABLE（downloads）表中插入了一條數(shù)據(jù)。接下來在insert方法最后啟動(dòng)DownloadService，這幾步都是在DownloadProvider進(jìn)程中進(jìn)行的。接下來會(huì)有兩條主線。
1，在DownloadProvider進(jìn)程中啟動(dòng)的這個(gè)DownloadService繼續(xù)執(zhí)行。
2，返回到Step 4 Browser進(jìn)程的中的DownloadHandler.onDownloadStartNoStream方法中去跳轉(zhuǎn)界面。
這里我們不討論UI界面，接下來分析DownloadService的操作。
Step 7：DownloadService.onCreate
@Override
public void onCreate() {
super.onCreate();
....
mUpdateThread = new HandlerThread(TAG + "-UpdateThread");
mUpdateThread.start();
mUpdateHandler = new Handler(mUpdateThread.getLooper(), mUpdateCallback);
mScanner = new DownloadScanner(this);
mNotifier = new DownloadNotifier(this);
mNotifier.cancelAll();
mObserver = new DownloadManagerContentObserver();
getContentResolver().registerContentObserver(Downloads.Impl.ALL_DOWNLOADS_CONTENT_URI,
true, mObserver);
....
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadService.java
第一次啟動(dòng)，首次執(zhí)行onCreate方法，創(chuàng)建一個(gè)HandlerThread工作線程，并注冊了一個(gè)監(jiān)聽數(shù)據(jù)庫改變的一個(gè)DownloadManagerContentObserver對象，監(jiān)聽的uri為"content://downloads/all_downloads"，第2個(gè)參數(shù)為true，表示可以同時(shí)匹配其派生的Uri。接下來進(jìn)入onStartCommand方法，在onStartCommand方法中繼續(xù)執(zhí)行enqueueUpdate方法。
public void enqueueUpdate() {
if (mUpdateHandler != null) {
mUpdateHandler.removeMessages(MSG_UPDATE);
mUpdateHandler.obtainMessage(MSG_UPDATE, mLastStartId, -1).sendToTarget();
}
}
這個(gè)方法執(zhí)行很簡單，首先是移除掉之前所有的MSG_UPDATE消息，然后再重新發(fā)送一個(gè)MSG_UPDATE消息，接下來分析Handler這個(gè)消息的回調(diào)實(shí)現(xiàn)。
private Handler.Callback mUpdateCallback = new Handler.Callback() {
@Override
public boolean handleMessage(Message msg) {
Process.setThreadPriority(Process.THREAD_PRIORITY_BACKGROUND);
final int startId = msg.arg1;
final boolean isActive;
synchronized (mDownloads) {
isActive = updateLocked();
}
if (msg.what == MSG_FINAL_UPDATE) {
mNotifier.dumpSpeeds();
}
if (isActive) {
// Still doing useful work, keep service alive. These active
// tasks will trigger another update pass when they're finished.
// Enqueue delayed update pass to catch finished operations that
// didn't trigger an update pass; these are bugs.
enqueueFinalUpdate();
} else {
// No active tasks, and any pending update messages can be
// ignored, since any updates important enough to initiate tasks
// will always be delivered with a new startId.
if (stopSelfResult(startId)) {
if (DEBUG_LIFECYCLE) Log.v(TAG, "Nothing left; stopped");
getContentResolver().unregisterContentObserver(mObserver);
mScanner.shutdown();
mUpdateThread.quit();
}
}
return true;
}
};
這個(gè)方法處理的邏輯比較多，先整體上認(rèn)識這個(gè)，主要有updateLocked方法主要負(fù)責(zé)具體的下載實(shí)現(xiàn)，它的返回值是一個(gè)boolean類型，用以判斷當(dāng)前下載是否是激活狀態(tài)，也就是是否有下載任務(wù)。接下來如果判斷isActive為true，則會(huì)去執(zhí)行enqueueFinalUpdate方法。
private void enqueueFinalUpdate() {
mUpdateHandler.removeMessages(MSG_FINAL_UPDATE);
mUpdateHandler.sendMessageDelayed(
mUpdateHandler.obtainMessage(MSG_FINAL_UPDATE, mLastStartId, -1),
5 * MINUTE_IN_MILLIS);
}
從這里我們可以看出，這個(gè)回調(diào)其實(shí)是當(dāng)有下載任務(wù)的時(shí)候，會(huì)一直的循環(huán)執(zhí)行下去，用以保證下載的任務(wù)的連續(xù)性，如果有中斷，則會(huì)重新啟動(dòng)。
下面我們來分析updateLocked的具體實(shí)現(xiàn)，是如何將下載任務(wù)放入線程中去執(zhí)行的，又是怎么知道有哪些下載任務(wù)的。
Step 8 ：DownloadService.updateLocked
private boolean updateLocked() {
...
final Cursor cursor = resolver.query(Downloads.Impl.ALL_DOWNLOADS_CONTENT_URI,
null, null, null, null);
try {
final DownloadInfo.Reader reader = new DownloadInfo.Reader(resolver, cursor);
final int idColumn = cursor.getColumnIndexOrThrow(Downloads.Impl._ID);
while (cursor.moveToNext()) {
final long id = cursor.getLong(idColumn);
staleIds.remove(id);
DownloadInfo info = mDownloads.get(id);
if (info != null) {
updateDownload(reader, info, now);
} else {
info = insertDownloadLocked(reader, now);
}
if (info.mDeleted) {
// Delete download if requested, but only after cleaning up
if (!TextUtils.isEmpty(info.mMediaProviderUri)) {
resolver.delete(Uri.parse(info.mMediaProviderUri), null, null);
}
deleteFileIfExists(info.mFileName);
resolver.delete(info.getAllDownloadsUri(), null, null);
} else {
// Kick off download task if ready
final boolean activeDownload = info.startDownloadIfReady(mExecutor);
// Kick off media scan if completed
final boolean activeScan = info.startScanIfReady(mScanner);
isActive |= activeDownload;
isActive |= activeScan;
}
// Keep track of nearest next action
nextActionMillis = Math.min(info.nextActionMillis(now), nextActionMillis);
}
} finally {
cursor.close();
}
// Clean up stale downloads that disappeared
for (Long id : staleIds) {
deleteDownloadLocked(id);
}
...
return isActive;
}
這個(gè)方法的實(shí)現(xiàn)分為幾步：
1.查詢downloads表中的所有記錄，接著將其封裝成一個(gè)DownloadInfo對象。
2.顯然第一次DownloadInfo的info是空值，接下來insertDownloadLocked會(huì)根據(jù)Cursor去新建一個(gè)DownloadInfo信息。
3.DownloadInfo緩存的管理，將DownloadInfo緩存至mDownloads中管理。這里有個(gè)小的判斷分支，如果info.mDeleted為true，則刪除掉這條下載記錄，并且對應(yīng)的文件也將被刪除，其實(shí)屬于邏輯控制，跟下載無太大關(guān)系，不用太糾結(jié)。
4.對于一個(gè)新的下載，info.mDeleted顯然是false，所以會(huì)進(jìn)入到到else語句，調(diào)用DownloadInfo的startDownloadIfReady方法開始下載。
我們先分析insertDownloadLocked新建一個(gè)下載任務(wù)DownloadInfo的流程
private DownloadInfo insertDownloadLocked(DownloadInfo.Reader reader, long now) {
final DownloadInfo info = reader.newDownloadInfo(this, mSystemFacade, mNotifier);
mDownloads.put(info.mId, info);
if (Constants.LOGVV) {
Log.v(Constants.TAG, "processing inserted download " + info.mId);
}
return info;
}
這個(gè)方法中，調(diào)用DownloadInfo.Reader去新建一個(gè)下載任務(wù)，從前面可以看出，這個(gè)reader對象是由數(shù)據(jù)庫Cursor進(jìn)行封裝的，具體分析reader.newDownloadInfo方法
public DownloadInfo newDownloadInfo(
Context context, SystemFacade systemFacade, DownloadNotifier notifier)
final DownloadInfo info = new DownloadInfo(context, systemFacade, notifier);
updateFromDatabase(info);
readRequestHeaders(info);
return info;
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadInfo.java
Reader是DownloadInfo的一個(gè)靜態(tài)內(nèi)部類，這個(gè)方法中，首先是new了一個(gè)DownloadInfo對象，然后調(diào)用updateFromDatabase去更新DownloadInfo的一些屬性值。實(shí)現(xiàn)比較簡單，就是根據(jù)前面的Cursor對象，獲取數(shù)據(jù)庫的一些字段值保存在DownloadInfo中。
從這里我們可以看出，數(shù)據(jù)庫中所有的信息都會(huì)封裝成一個(gè)下載DownloadInfo，那么它是通過什么來判斷當(dāng)前數(shù)據(jù)是否是需要下載的任務(wù)呢？顯然如果這個(gè)url對應(yīng)的任務(wù)已經(jīng)被下載完成了，那么肯定是不需要再次下載的。接下來我們繼續(xù)往下走，進(jìn)入到startDownloadIfReady這個(gè)方法。
Step 9：DownloadInfo.startDownloadIfReady
public boolean .startDownloadIfReady(ExecutorService executor) {
synchroized (this) {
final boolean isReady = isReadyToDownload();
final boolean isActive = mSubmittedTask != null && !mSubmittedTask.isDone();
if (isReady && !isActive) {
if (mStatus != Impl.STATUS_RUNNING) {
mStatus = Impl.STATUS_RUNNING;
ContentValues values = new ContentValues();
values.put(Impl.COLUMN_STATUS, mStatus);
mContext.getContentResolver().update(getAllDownloadsUri(), values, null, null);
}
mTask = new DownloadThread(mContext, mSystemFacade, mNotifier, this);
mSubmittedTask = executor.submit(mTask);
}
return isReady;
}
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadInfo.java
我們先分析isReadyToDownload這個(gè)方法。
private boolean isReadyToDownload() {
....
switch (mStatus) {
case 0: // status hasn't been initialized yet, this is a new download
case Downloads.Impl.STATUS_PENDING: // download is explicit marked as ready to start
case Downloads.Impl.STATUS_RUNNING: // download interrupted (process killed etc) while
// running, without a chance to update the database
return true;
case Downloads.Impl.STATUS_WAITING_FOR_NETWORK:
case Downloads.Impl.STATUS_QUEUED_FOR_WIFI:
return checkCanUseNetwork(mTotalBytes) == NetworkState.OK;
case Downloads.Impl.STATUS_WAITING_TO_RETRY:
// download was waiting for a delayed restart
final long now = mSystemFacade.currentTimeMillis();
return restartTime(now) <= now;
case Downloads.Impl.STATUS_DEVICE_NOT_FOUND_ERROR:
// is the media mounted?
return Environment.getExternalStorageState().equals(Environment.MEDIA_MOUNTED);
/// M: Because OMA DL spec, if insufficient memory, we
/// will show to user but not retry.
//case Downloads.Impl.STATUS_INSUFFICIENT_SPACE_ERROR:
// should check space to make sure it is worth retrying the download.
// but thats the first thing done by the thread when it retries to download
// it will fail pretty quickly if there is no space.
// so, it is not that bad to skip checking space availability here.
//return true;
/// M: Add for fix alp00406729, file already exist but user do not operation. @{
case Downloads.Impl.STATUS_FILE_ALREADY_EXISTS_ERROR:
return false;
/// @}
}
return false;
}
一切都明白了，這里就是根據(jù)mStatus這個(gè)字段，來判斷這個(gè)任務(wù)是否需要下載，也解決了我們之前的疑問，返回值為true的才會(huì)去執(zhí)行下載，我們可以回頭看看Browser里面當(dāng)時(shí)insert一條下載記錄的時(shí)候，是沒有插入mStatus這個(gè)字段的，所以對于一個(gè)新任務(wù)這里mStatus為默認(rèn)值即0，整個(gè)返回值為true。
接下來分析isActive這個(gè)boolean值，它主要用來標(biāo)識當(dāng)前DownloadInfo是否在線程中去執(zhí)行了，保證一個(gè)DownloadInfo只執(zhí)行一次，對于新任務(wù)，顯然初始化的時(shí)候mSubmittedTask為null。
接下來進(jìn)入if語句，先update數(shù)據(jù)庫中的COLUMN_STATUS字段置為STATUS_RUNNING。然后新建一個(gè)DownloadThread，放入到ExecutorService線程池中去執(zhí)行，這樣一個(gè)下載鏈接就正式開始下載了。接下來分析下載讀寫文件以及更新數(shù)據(jù)庫的動(dòng)作。
Step 10：new DownloadThread
@Override
public void run() {
....
if (DownloadInfo.queryDownloadStatus(mContext.getContentResolver(), mId)
== Downloads.Impl.STATUS_SUCCESS) {
logDebug("Already finished; skipping");
return;
}
....
executeDownload();
....
}
這個(gè)方法定義在
DownloadThread是一個(gè)Runnable對象，這里我們關(guān)注構(gòu)造方法中的第4個(gè)參數(shù)，即DownloadInfo，將DownloadInfo這個(gè)對象的信息，傳給DownloadThread的成員變量，還有DownloadInfoDelta對象，最后用于更新下載進(jìn)度數(shù)據(jù)庫信息，我們后續(xù)分析。這樣就完全得到了這條下載信息的內(nèi)容。接下來去執(zhí)行DownloadThread的run方法，在新的線程中進(jìn)行下載。在run方法的實(shí)現(xiàn)中，首先是再次確認(rèn)這個(gè)任務(wù)是需要下載的，否則直接return，線程結(jié)束，然后如果需要下載則去調(diào)用executeDownload方法去執(zhí)行。
private void executeDownload() throws StopRequestException {
.....
URL url;
try {
// TODO: migrate URL sanity checking into client side of API
url = new URL(mInfoDelta.mUri);
} catch (MalformedURLException e) {
throw new StopRequestException(STATUS_BAD_REQUEST, e);
}
int redirectionCount = 0;
while (redirectionCount++ < Constants.MAX_REDIRECTS) {
// Open connection and follow any redirects until we have a useful
// response with body.
HttpURLConnection conn = null;
try {
checkConnectivity();
conn = (HttpURLConnection) url.openConnection();
conn.setInstanceFollowRedirects(false);
conn.setConnectTimeout(DEFAULT_TIMEOUT);
conn.setReadTimeout(DEFAULT_TIMEOUT);
addRequestHeaders(conn, resuming);
final int responseCode = conn.getResponseCode();
switch (responseCode) {
case HTTP_OK:
....
/// @}
transferData(conn);
return;
.....
}
throw new StopRequestException(STATUS_TOO_MANY_REDIRECTS, "Too many redirects");
}
在executeDownload方法中根據(jù)url創(chuàng)建一個(gè)HttpURLConnection連接。然后判斷getResponseCode網(wǎng)絡(luò)端返回值。這里我們分析HTTP_OK的情況。在HTTP_OK：接下來調(diào)用transferData(conn);傳入的參數(shù)為這個(gè)HttpURLConnection這個(gè)連接。
Step 11：DownloadThread.transferData
private void transferData(HttpURLConnection conn) throws StopRequestException {
....
DrmManagerClient drmClient = null;
ParcelFileDescriptor outPfd = null;
FileDescriptor outFd = null;
InputStream in = null;
OutputStream out = null;
try {
try {
in = conn.getInputStream();
} catch (IOException e) {
throw new StopRequestException(STATUS_HTTP_DATA_ERROR, e);
}
try {
outPfd = mContext.getContentResolver()
.openFileDescriptor(mInfo.getAllDownloadsUri(), "rw");
outFd = outPfd.getFileDescriptor();
if (DownloadDrmHelper.isDrmConvertNeeded(mInfoDelta.mMimeType)) {
drmClient = new DrmManagerClient(mContext);
out = new DrmOutputStream(drmClient, outPfd, mInfoDelta.mMimeType);
} else {
out = new ParcelFileDescriptor.AutoCloseOutputStream(outPfd);
}
// Pre-flight disk space requirements, when known
if (mInfoDelta.mTotalBytes > 0) {
final long curSize = Os.fstat(outFd).st_size;
final long newBytes = mInfoDelta.mTotalBytes - curSize;
StorageUtils.ensureAvailableSpace(mContext, outFd, newBytes);
// We found enough space, so claim it for ourselves
Os.posix_fallocate(outFd, 0, mInfoDelta.mTotalBytes);
}
// Move into place to begin writing
Os.lseek(outFd, mInfoDelta.mCurrentBytes, OsConstants.SEEK_SET);
} catch (ErrnoException e) {
throw new StopRequestException(STATUS_FILE_ERROR, e);
} catch (IOException e) {
throw new StopRequestException(STATUS_FILE_ERROR, e);
}
// Start streaming data, periodically watch for pause/cancel
// commands and checking disk space as needed.
transferData(in, out, outFd);
....
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadThread.java
在這個(gè)方法中，獲取一個(gè)該url對應(yīng)的網(wǎng)絡(luò)輸入流對象InputStream，同時(shí)根據(jù)uri構(gòu)造一個(gè)文件描述符，進(jìn)而構(gòu)建一個(gè)輸出流OutputStream對象，最后到重載的transferData方法，將輸入輸出流，以及文件描述符傳入transferData開始存儲文件。
private void transferData(InputStream in, OutputStream out, FileDescriptor outFd)
throws StopRequestException {
final byte buffer[] = new byte[Constants.BUFFER_SIZE];
while (true) {
checkPausedOrCanceled();
int len = -1;
try {
len = in.read(buffer);
} catch (IOException e) {
throw new StopRequestException(
STATUS_HTTP_DATA_ERROR, "Failed reading response: " + e, e);
}
if (len == -1) {
break;
}
try {
// When streaming, ensure space before each write
if (mInfoDelta.mTotalBytes == -1) {
final long curSize = Os.fstat(outFd).st_size;
final long newBytes = (mInfoDelta.mCurrentBytes + len) - curSize;
StorageUtils.ensureAvailableSpace(mContext, outFd, newBytes);
}
out.write(buffer, 0, len);
mMadeProgress = true;
mInfoDelta.mCurrentBytes += len;
updateProgress(outFd);
} catch (ErrnoException e) {
throw new StopRequestException(STATUS_FILE_ERROR, e);
} catch (IOException e) {
throw new StopRequestException(STATUS_FILE_ERROR, e);
}
}
.....
}
真正開始下載都是在這段code中，首先checkPausedOrCanceled方法檢查是否有取消下載請求，如果有直接進(jìn)入catch語句跳過，下載結(jié)束。如果沒有取消，則執(zhí)行while語句，執(zhí)行輸入輸出流的讀寫操作。每一次讀寫的同時(shí)都會(huì)執(zhí)行updateProgress方法，顯然該方法是用來更新進(jìn)度的，下面具體來分析。
Step 12：DownloadThread.updateProgress
private void updateProgress(FileDescriptor outFd) throws IOException, StopRequestException {
....
final long bytesDelta = currentBytes - mLastUpdateBytes;
final long timeDelta = now - mLastUpdateTime;
if (bytesDelta > Constants.MIN_PROGRESS_STEP && timeDelta > Constants.MIN_PROGRESS_TIME) {
// fsync() to ensure that current progress has been flushed to disk,
// so we can always resume based on latest database information.
outFd.sync();
//mInfoDelta.writeToDatabaseOrThrow();
mInfoDelta.writeToDatabaseWithoutModifyTime();
mLastUpdateBytes = currentBytes;
mLastUpdateTime = now;
}
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadThread.java
總共做了兩件事，第一，調(diào)用outFd.sync強(qiáng)制所有系統(tǒng)緩沖區(qū)與基礎(chǔ)設(shè)備同步，第二調(diào)用mInfoDelta的writeToDatabaseWithoutModifyTime去更新數(shù)據(jù)庫操作，即將當(dāng)前進(jìn)度，下載了多少update到數(shù)據(jù)庫。
Step 13：DownloadInfoDelta.writeToDatabaseWithoutModifyTime
public void writeToDatabaseWithoutModifyTime() throws StopRequestException {
final ContentValues values = new ContentValues();
values.put(Downloads.Impl.COLUMN_URI, mUri);
values.put(Downloads.Impl._DATA, mFileName);
values.put(Downloads.Impl.COLUMN_MIME_TYPE, mMimeType);
values.put(Downloads.Impl.COLUMN_STATUS, mStatus);
values.put(Downloads.Impl.COLUMN_FAILED_CONNECTIONS, mNumFailed);
values.put(Constants.RETRY_AFTER_X_REDIRECT_COUNT, mRetryAfter);
values.put(Downloads.Impl.COLUMN_TOTAL_BYTES, mTotalBytes);
values.put(Downloads.Impl.COLUMN_CURRENT_BYTES, mCurrentBytes);
values.put(Constants.ETAG, mETag);
values.put(Downloads.Impl.COLUMN_ERROR_MSG, mErrorMsg);
if (mContext.getContentResolver().update(mInfo.getAllDownloadsUri(),
values, Downloads.Impl.COLUMN_DELETED + " == '0'", null) == 0) {
throw new StopRequestException(STATUS_CANCELED, "Download deleted or missing!");
}
}
}
這個(gè)方法定義在packages/providers/DownloadProvider/src/com/android/providers/downloads/DownloadThread.java
DownloadInfoDelta是DownloadThread的一個(gè)內(nèi)部類，主要用于更新數(shù)據(jù)庫進(jìn)度操作，這個(gè)方法中此時(shí)uri為"content://downloads/all_downloads/id",對應(yīng)DownloadProvider的update方法去更新數(shù)據(jù)庫，而此時(shí)又會(huì)回調(diào)至DowbloadService中的DownloadManagerContentObserver監(jiān)聽中，因?yàn)榇藭r(shí)對應(yīng)uri數(shù)據(jù)庫內(nèi)容已經(jīng)改變。至此，整個(gè)updateLocked方法執(zhí)行完畢。
簡單分析DownloadManagerContentObserver內(nèi)容，可以看出這個(gè)目的還是保證了下載的連續(xù)性，只要每次有下載數(shù)據(jù)更新，則會(huì)循環(huán)檢測，以確保下載任務(wù)的連續(xù)性。
private class DownloadManagerContentObserver extends ContentObserver {
public DownloadManagerContentObserver() {
super(new Handler());
}
@Override
public void onChange(final boolean selfChange) {
enqueueUpdate();
}
}
至此，整個(gè)下載過程已經(jīng)結(jié)束，至于UI界面的更新情況，則只需要監(jiān)聽數(shù)據(jù)庫中的數(shù)據(jù)變化，或者在有下載任務(wù)時(shí)候，間隔一段時(shí)間去數(shù)據(jù)庫查詢進(jìn)度信息，更新進(jìn)度即可。
對于下載界面，自4.4之后，都是BrowserActivity->DownloadList->DocumentActivity，而且對于DocumentUI正是采用的一段時(shí)間查詢數(shù)據(jù)庫，更新的方式，這里我們也不討論了。
