webkit加载网页

来源：互联网发布：征途打鱼软件编辑：程序博客网时间：2024/05/16 09:21

一、读取url

以下内容部分引自http://blog.csdn.net/dlmu2001/article/details/5936122
http的协议细节实现并不需要WebCore来关注，WebCore要关注的是，如何设置请求的相关头部信息，如何获取服务器返回回来的响应体部数据。
WebKit中同http打交道的类主要是ResourceRequest，ResourceResonse，ResourceHandle*，ResourceHandleManager等，这里先介绍同请求信息维护相关的ResourceRequest.
ResourceRequest类的作用比较好理解，基本上就是维护http请求相关的信息（app或者内核都有可能设置这些信息），然后当WebCore发起http请求的时候，可以获取这些信息，调用curl的接口设置对应的http请求字段。在这些信息中，最常用到的是url。
ResourceRequest类继承于ResourceRequestBase类，绝大部分功能在ResourceRequestBase类中实现。
一）ResourceRequest构造：构造一个ResourceRequest对象只需要url参数就够了，这是比较简单的一个类，没有维护其它类的对象或者句柄。
ResourceRequest(const String& url)；
ResourceRequest(const KURL& url)；
二）ResourceRequest对象的创建与维护
1.当用户输入网址，触发qt事件，由MainWindow(WebKitBuild/Debug/QtTestBrowser/moc_mainwindow.cpp:87<--moc_launcherwindow.cpp:145)调用changeLocation函数（另有openLocation等）。
由MainWindow(Tools/QtTestBrowser/mainwindow.cpp:194)获取地址栏中的url(QString类型/usr/include/QtCore/qstring.h)，创建QUrl类型(/usr/include/QtCore/qurl.h url规格化类，包括host、port、scheme等成员变量)，加载页面(page()->mainFrame()->load(url) page()函数是全局函数？)。

最终调用QWebFrame::load(Source/WebKit/qt/Api/qwebframe.cpp:885)函数，在该函数中，会构造出ResourceRequest(Source/WebCore/platform/network/qt/ResourceRequest.h继承ResourceRequestBaseSource/WebCore/platform/network/ResourceRequestBase.cpp)对象，并将这个对象作为一个参数，调用FrameLoader对象的load函数(Source/WebCore/loader/FrameLoader.cpp:1450)。

FrameLoader类负责将documents加载到Frames。当你点击一个链接的时候，FrameLoader创建一个新的处于”policy”状态的DocumentLoader对象，等待WebKit客户端决定是否处理这个加载。通常WebKit客户端会指示FrameLoader将这个加载视为一个导航（navigation），而不是阻止加载等。

DocumentLoader类中会维护这个ResourceRequest(FrameLoader.cpp:1460)，policyChecker()
——>checkNavigationPolicy(Source/WebCore/loader/FrameLoader.cpp:1585)

1585 policyChecker()->checkNavigationPolicy(loader->request(), loader, formState,

1586 callContinueLoadAfterNavigationPolicy, this);//回调函数？如下调用

——>m_frame->loader()->client()->dispatchDecidePolicyForNavigationAction(Source/WebCore/loader/PolicyChecker.cpp:87)

84 m_callback.set(request, formState.get(), function, argument);//callContinueLoadAfterNavigationPolicy

一旦客户端指示FrameLoader将本次加载视为一个导航，FrameLoader就推动DocumentLoader进入”provisional”状态，在该状态，DocumentLoader会发起一个网络请求，并等待以确定网络请求将发起一个下载还是一个新的document。
——>page->d->acceptNavigationRequest（Source/WebKit/qt/WebCoreSupport/FrameLoaderClientQt.cpp:1271）
        callPolicyFunction(Source/WebKit/qt/WebCoreSupport/FrameLoaderClientQt.cpp:1283)
        回调函数？如上
         //——>(m_frame->loader()->policyChecker()->*function)(action) (245)
——>callback.call(Source/WebCore/loader/PolicyChecker.cpp:160)
——>m_navigationFunction(Source/WebCore/loader/PolicyCallback.cpp:103)
——>loader->continueLoadAfterNavigationPolicy(Source/WebCore/loader/FrameLoader.cpp:2977)

2974 void FrameLoader::callContinueLoadAfterNavigationPolicy(void* argument,

2975 const ResourceRequest& request, PassRefPtr<FormState> formState, bool shouldContinue)

2976 {

2977 FrameLoader* loader = static_cast<FrameLoader*>(argument);

2978 loader->continueLoadAfterNavigationPolicy(request, formState, shouldContinue);

2979 }

——>continueLoadAfterWillSubmitForm(Source/WebCore/loader/FrameLoader.cpp:3105)
——>m_provisionalDocumentLoader->startLoadingMainResource(Source/WebCore/loader/FrameLoader.cpp:2572)

接下去，DocumentLoader会创建一个MainResourceLoader对象，这个对象主要用来通过ResourceHandle接口同平台网络库进行交互。将MainResourceLoader和DocumentLoader分开来主要有两个目的：（1）MainResourceLoader让DocumentLoader从处理ResourceHandle回调的细节中抽身出来（2）降低MainResourceLoader的生命周期和DocumentLoader的生命周期（同Document绑定）的耦合度。

——>m_mainResourceLoader->load(Source/WebCore/loader/DocumentLoader.cpp:798)
——>loadNow(Source/WebCore/loader/MainResourceLoader.cpp:612)
——>ResourceHandle::create(Source/WebCore/loader/MainResourceLoader.cpp:585)
——>newHandle->start(Source/WebCore/platform/network/ResourceHandle.cpp:71)
——>new QNetworkReplyHandler(Source/WebCore/platform/network/qt/ResourceHandleQt.cpp:100)
       ——>r.toNetworkRequest(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:399) 设置http头
——>m_queue.push 提交事件？ (Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:401)
——>flush();(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:164)
——>(m_replyHandler->*(m_enqueuedCalls.takeFirst()))();（Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:195）
——>QNetworkReplyHandler::start()(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:659)
        建立连接(ESTABLISHED)、发送请求 665： QNetworkReply* reply = sendNetworkRequest(d->m_context->networkAccessManager(), d->m_firstRequest);
         ——>601 WebCore::QNetworkReplyHandler::sendNetworkRequest
               ——>626 manager->get(m_request)
      665     QNetworkReply* reply = sendNetworkRequest(d->m_context->networkAccessManager(), d->m_firstRequest);
      669     m_replyWrapper = new QNetworkReplyWrapper(&m_queue, reply, m_resourceHandle->shouldContentSniff() && d->m_context-
    >mimeSniffingEnabled(), this);
               ——>208 QNetworkReplyWrapper::QNetworkReplyWrapper(QNetworkReplyHandlerCallQueue* queue, QNetworkReply* reply, bool
    sniffMIMETypes, QObject* parent)
                  209     : QObject(parent)
                  210     , m_reply(reply)
                  211     , m_queue(queue)
                  212     , m_responseContainsData(false)
                  213     , m_sniffMIMETypes(sniffMIMETypes)
                  214 {
                   215     Q_ASSERT(m_reply);
                  216
                  217     // setFinished() must be the first that we connect, so isFinished() is updated when running other slots.
                   //QT通过此种方式异步回调??(yes, 信号、槽机制，http://www.ibm.com/developerworks/cn/linux/guitoolkit/qt/signal-slot/ )
                  218     connect(m_reply, SIGNAL(finished()), this, SLOT(setFinished()));
                  219     connect(m_reply, SIGNAL(finished()), this, SLOT(receiveMetaData()));
                  220     connect(m_reply, SIGNAL(readyRead()), this, SLOT(receiveMetaData()));
                  221 }

一旦加载系统接收到足够的信息可以确定资源确实代表了document，FrameLoader就将DocumentLoader推向”committed”状态，在该状态中，frame将显示document。

下载完了之后
——>QtMIMETypeSniffer::qt_metacall()(WebKitBuild/Debug/WebCore/moc_QtMIMETypeSniffer.cpp:77)
——>trySniffing()(Source/WebCore/platform/network/qt/QtMIMETypeSniffer.cpp:59)
——>m_sniffer.sniff(data.constData(), data.size())(Source/WebCore/platform/network/qt/QtMIMETypeSniffer.cpp:51)

50 QByteArray data = m_reply->peek(m_sniffer.dataSize());

51 const char* sniffedMimeType = m_sniffer.sniff(data.constData(), data.size());

       data.constData()保存着网页内容，涉及的变量包括m_reply、m_sniffer

      ——>emit finished (Source/WebCore/platform/network/qt/QtMIMETypeSniffer.cpp:65)
      ——>QMetaObject::activate(WebKitBuild/Debug/WebCore/moc_QtMIMETypeSniffer.cpp:88)
      ……
      WebCore::QNetworkReplyWrapper::qt_metacall (WebKitBuild/Debug/WebCore/moc_QNetworkReplyHandler.cpp)

77 if (_c == QMetaObject::InvokeMetaMethod) {

78 switch (_id) {

79 case 0: receiveMetaData(); break;

80 case 1: didReceiveFinished(); break;

81 case 2: didReceiveReadyRead(); break;

82 case 3: receiveSniffedMIMEType(); break;

83 case 4: setFinished(); break;

84 default: ;

85 }

86 _id -= 5;

87 }

      ——>receiveSniffedMIMEType()(WebKitBuild/Debug/WebCore/moc_QNetworkReplyHandler.cpp:82)
      ——>emitMetaDataChanged()(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:305)
         ——>QueueLocker lock(m_queue)(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp)
         ——>~QueueLocker() { m_queue->unlock(); } (203)
         ——>flush();(178)
         ——>(m_replyHandler->*(m_enqueuedCalls.takeFirst()))(); （195）

         ——>1、QNetworkReplyHandler::forwardData (572)
         ——>client->didReceiveData(m_resourceHandle, data.constData(), data.length(), -1); (586)
      ——>ResourceLoader::didReceiveData(ResourceHandle*, const char* data, int length, int encodedDataLength) (Source/WebCore/loader/ResourceLoader.cpp:427)
               data指向保存页面内容
      ——>MainResourceLoader::didReceiveData(Source/WebCore/loader/MainResourceLoader.cpp:431)
      ——>ResourceLoader::didReceiveData (Source/WebCore/loader/ResourceLoader.cpp:267)

279 addData(data, length, allAtOnce);

280 // FIXME: If we get a resource with more than 2B bytes, this code won't do the right thing.

281 // However, with today's computers and networking speeds, this won't happen in practice.

282 // Could be an issue with a giant local file.

283 if (m_sendResourceLoadCallbacks && m_frame)

284 frameLoader()->notifier()->didReceiveData(this, data, length, static_cast<int>(encodedDataLength));

            保存页面到cache 函数： addData()
      ——>ResourceLoadNotifier::didReceiveData(Source/WebCore/loader/ResourceLoadNotifier.cpp:77)
         ——>ResourceLoadNotifier::dispatchDidReceiveContentLength (133)
      ——>InspectorInstrumentation::didReceiveContentLength（Source/WebCore/inspector/InspectorInstrumentation.h：713）
      ——>InspectorInstrumentation::didReceiveContentLengthImpl（Source/WebCore/inspector/InspectorInstrumentation.cpp：487）

html解析
      ——>2、QNetworkReplyHandler::finish(Source/WebCore/platform/network/qt/QNetworkReplyHandler.cpp:437)
      ——>ResourceLoader::didFinishLoading (Source/WebCore/loader/ResourceLoader.cpp:434)
      ——>MainResourceLoader::didFinishLoading(Source/WebCore/loader/MainResourceLoader.cpp:466)
      ——>FrameLoader::finishedLoading (Source/WebCore/loader/FrameLoader.cpp：2287)
      ——>DocumentLoader::finishedLoading (Source/WebCore/loader/DocumentLoader.cpp:282)
         ——>FrameLoader::finishedLoadingDocument(DocumentLoader* loader) (Source/WebCore/loader/FrameLoader.cpp:2333)
            ——>FrameLoaderClientQt::finishedLoading(DocumentLoader* loader)(Source/WebKit/qt/WebCoreSupport/FrameLoaderClientQt.cpp:626)
         ——>DocumentWriter::end() (Source/WebCore/loader/DocumentWriter.cpp:211)

211 void DocumentWriter::end()

212 {

213 m_frame->loader()->didEndDocument();

214 endIfNotLoadingMainResource();

215 }

213 ——>Document::finishParsing() (Source/WebCore/dom/Document.cpp:2246)
214： DocumentWriter::endIfNotLoadingMainResource() （Source/WebCore/loader/DocumentWriter.cpp 228）

228 addData(0, 0, true);

229 m_frame->document()->finishParsing();

               ——>DocumentWriter::addData() (208)
                  ——>DecodedDataDocumentParser::appendBytes(DocumentWriter* writer , const char* data, int length, bool shouldFlush)（Source/WebCore/dom/DecodedDataDocumentParser.cpp：40）
                  ——> HTMLDocumentParser::append(const SegmentedString& source) (Source/WebCore/html/parser/HTMLDocumentParser.cpp:367)
                  ——> HTMLDocumentParser::pumpTokenizerIfPossible(SynchronousMode mode) (175)
                  ——> HTMLDocumentParser::pumpTokenizer(SynchronousMode mode) (261)
                  解析、执行script
                     ——> HTMLDocumentParser::canTakeNextToken(SynchronousMode mode, PumpSession& session) (223)
                     ——> HTMLDocumentParser::runScriptsForPausedTreeBuilder() (205)
                     ——> HTMLScriptRunner::execute(PassRefPtr<Element> scriptElement, const TextPosition1& scriptStartPosition) Source/WebCore/html/parser/HTMLScriptRunner.cpp:167)
                     ——> WebCore::HTMLScriptRunner::runScript(this=0x81e2a58, script=0x832c040, scriptStartPosition=...) (296)
                     ——> WebCore::ScriptElement::prepareScript (this=0x832c084, scriptStartPosition=...,     supportLegacyTypes=WebCore::ScriptElement::DisallowLegacyTypeInTypeAttribute)(Source/WebCore/dom/ScriptElement.cpp:240)
                     ——> WebCore::ScriptElement::executeScript (this=0x832c084, sourceCode=...)(283)
                     ——> WebCore::ScriptController::evaluate (this=0x828d1a4, sourceCode=...)(Source/WebCore/bindings/js/ScriptController.cpp:166)
                     ——> WebCore::ScriptController::evaluateInWorld (143)
                     ——> WebCore::JSMainThreadExecState::evaluate(Source/WebCore/bindings/js/JSMainThreadExecState.h:54)
                     ——> JSC::evaluate(exec=0xae3f499c, scopeChain=0xae3ee5c8, source=..., thisValue=...) (Source/JavaScriptCore/runtime/Completion.cpp:64)
                     ——> JSC::Interpreter::execute(Source/JavaScriptCore/interpreter/Interpreter.cpp:767)
                     ——> JSC::JITCode::execute (Source/JavaScriptCore/jit/JITCode.h:77)
                     ——> ?? ()
                     ——>WebCore::jsHTMLDocumentPrototypeFunctionWrite(WebCore/generated/JSHTMLDocument.cpp:445)
                     ——> WebCore::JSHTMLDocument::write(Source/WebCore/bindings/js/JSHTMLDocumentCustom.cpp:161)
                     ——> WebCore::documentWrite (156)
                     ——> WebCore::Document::write (Source/WebCore/dom/Document.cpp:2227)
                           ——> WebCore::HTMLDocumentParser::insert(Source/WebCore/html/parser/HTMLDocumentParser.cpp:324)
                              回到了html的解析
                           ——> WebCore::HTMLDocumentParser::pumpTokenizerIfPossible (175)
                           ——> WebCore::HTMLDocumentParser::pumpTokenizer (299)
                                    查看是否有script，1、如果有，转到“ 解析、执行script”；2、如果没有转到下面“”
                  完成解析html
                  InspectorInstrumentation::didWriteHTML(cookie, m_tokenizer->lineNumber());
                  ——> InspectorInstrumentation::didWriteHTML (Source/WebCore/inspector/InspectorInstrumentation.h:803)
                  ——>
            ——> Document::finishParsing() (Source/WebCore/dom/Document.cpp:2259)
                  ——>HTMLDocumentParser::finish() (Source/WebCore/html/parser/HTMLDocumentParser.cpp:427)
                  ——>HTMLDocumentParser::attemptToEnd() (399)
                  ——>

生成dom树


附件(js)下载、解析



二、下载
三、加载

参考文档
1.官方文档：http://www.webkit.org/blog/1188/how-webkit-loads-a-web-page/
2.dlmu2001的译文：http://blog.csdn.net/dlmu2001/article/details/5941432
3.dlmu200：WebCore中的http请求信息维护 http://blog.csdn.net/dlmu2001/article/details/59361221
4.其他技术性文章：Introduction to WebKit Objective-C Programming Guide http://developer.apple.com/library/mac/#documentation/Cocoa/Conceptual/DisplayWebContent/DisplayWebContent.html
5.其他技术性文章：Introduction to WebKit DOM Programming Topicshttp://developer.apple.com/library/mac/#documentation/AppleApplications/Conceptual/SafariJSProgTopics/WebKitJavaScript.html
6. http://www.webkit.org/coding/technical-articles.html

相关技巧
1.cscope查找类的定义：cs find e class.*QString[^;]*($|{)

0 0