Add innerHTML parsing with HTML5 parser

content/base/src/Makefile.in

@@ -84,6 +84,7 @@ REQUIRES	= xpcom \
 		  shistory \
 		  editor \
 		  windowwatcher \
+		  html5 \
 		  $(NULL)
 
 ifdef ACCESSIBILITY

content/base/src/nsContentUtils.cpp

@@ -161,6 +161,7 @@ static NS_DEFINE_CID(kXTFServiceCID, NS_XTFSERVICE_CID);
 #include "nsIOfflineCacheUpdate.h"
 #include "nsCPrefetchService.h"
 #include "nsIChromeRegistry.h"
+#include "nsHtml5Module.h"
 
 #ifdef IBMBIDI
 #include "nsIBidiKeyboard.h"
@@ -3459,6 +3460,44 @@ nsContentUtils::CreateContextualFragment(nsIDOMNode* aContextNode,
   // for compiling event handlers... so just bail out.
   nsCOMPtr<nsIDocument> document = node->GetOwnerDoc();
   NS_ENSURE_TRUE(document, NS_ERROR_NOT_AVAILABLE);
+  
+  PRBool bCaseSensitive = document->IsCaseSensitive();
+
+  nsCOMPtr<nsIHTMLDocument> htmlDoc(do_QueryInterface(document));
+  PRBool bHTML = htmlDoc && !bCaseSensitive;
+
+  if (bHTML && nsContentUtils::GetBoolPref("html5.enable", PR_TRUE)) {
+    // See if the document has a cached fragment parser. nsHTMLDocument is the
+    // only one that should really have one at the moment.
+    nsCOMPtr<nsIParser> parser = document->GetFragmentParser();
+    if (parser) {
+      // Get the parser ready to use.
+      parser->Reset();
+    }
+    else {
+      // Create a new parser for this operation.
+      parser = nsHtml5Module::NewHtml5Parser();
+      if (!parser) {
+        return NS_ERROR_OUT_OF_MEMORY;
+      }
+      document->SetFragmentParser(parser);
+    }
+    nsCOMPtr<nsIDOMDocumentFragment> frag;
+    rv = NS_NewDocumentFragment(getter_AddRefs(frag), document->NodeInfoManager());
+    NS_ENSURE_SUCCESS(rv, rv);
+    // XXX how does ownership of the fragment work here?
+    
+    nsCOMPtr<nsIContent> contextAsContent = do_QueryInterface(aContextNode);
+    NS_ASSERTION(contextAsContent, "Context node did not QI to nsIContent");
+    
+    parser->ParseFragment(aFragment, frag, contextAsContent->Tag(), contextAsContent->GetNameSpaceID());
+    
+    if (!aWillOwnFragment) {
+      NS_ADDREF(frag);
+    }
+    *aReturn = frag;
+    return NS_OK;
+  }
 
   nsAutoTArray<nsString, 32> tagStack;
   nsAutoString uriStr, nameStr;
@@ -3516,14 +3555,9 @@ nsContentUtils::CreateContextualFragment(nsIDOMNode* aContextNode,
   }
 
   nsCAutoString contentType;
-  PRBool bCaseSensitive = PR_TRUE;
   nsAutoString buf;
   document->GetContentType(buf);
   LossyCopyUTF16toASCII(buf, contentType);
-  bCaseSensitive = document->IsCaseSensitive();
-
-  nsCOMPtr<nsIHTMLDocument> htmlDoc(do_QueryInterface(document));
-  PRBool bHTML = htmlDoc && !bCaseSensitive;
 
   // See if the document has a cached fragment parser. nsHTMLDocument is the
   // only one that should really have one at the moment.

content/html/parser/src/nsHtml5Parser.cpp

@@ -49,6 +49,7 @@
 #include "nsEncoderDecoderUtils.h"
 #include "nsContentUtils.h"
 #include "nsICharsetDetector.h"
+#include "nsIScriptElement.h"
 
 #include "nsHtml5DocumentMode.h"
 #include "nsHtml5Tokenizer.h"
@@ -435,6 +436,50 @@ nsHtml5Parser::ParseFragment(const nsAString& aSourceBuffer,
   return NS_ERROR_NOT_IMPLEMENTED;
 }
 
+NS_IMETHODIMP
+nsHtml5Parser::ParseFragment(const nsAString& aSourceBuffer,
+                             nsISupports* aTargetNode,
+                             nsIAtom* aContextLocalName,
+                             PRInt32 aContextNamespace)
+{
+  nsCOMPtr<nsIContent> target = do_QueryInterface(aTargetNode);
+  NS_ASSERTION(target, "Target did not QI to nsIContent");
+  mTreeBuilder->setFragmentContext(aContextLocalName, aContextNamespace, target);
+  mFragmentMode = PR_TRUE;
+  NS_ASSERTION((mLifeCycle == NOT_STARTED), "Tried to start parse without initializing the parser properly.");
+  mTokenizer->start();
+  mLifeCycle = PARSING;
+  mParser = this;
+  mNodeInfoManager = target->GetOwnerDoc()->NodeInfoManager();
+
+  if (!aSourceBuffer.IsEmpty()) {
+    PRBool lastWasCR = PR_FALSE;
+    nsHtml5UTF16Buffer buffer(aSourceBuffer.Length());
+    memcpy(buffer.getBuffer(), aSourceBuffer.BeginReading(), aSourceBuffer.Length() * sizeof(PRUnichar));
+    buffer.setEnd(aSourceBuffer.Length());
+    while (buffer.hasMore()) {
+      buffer.adjust(lastWasCR);
+      lastWasCR = PR_FALSE;
+      if (buffer.hasMore()) {
+        lastWasCR = mTokenizer->tokenizeBuffer(&buffer);
+        if (mScriptElement) {
+          nsCOMPtr<nsIScriptElement> script = do_QueryInterface(mScriptElement);
+          NS_ASSERTION(script, "mScriptElement didn't QI to nsIScriptElement!");
+          script->PreventExecution();
+          mScriptElement = nsnull;
+        }
+      }
+    }
+  }
+
+  NS_ASSERTION((mLifeCycle == STREAM_ENDING), "Bad life cycle.");
+  mTokenizer->eof();
+  mTokenizer->end();
+  mLifeCycle = TERMINATED;
+  DropParserAndPerfHint();
+  return NS_OK;
+}
+
 NS_IMETHODIMP
 nsHtml5Parser::BuildModel(void)
 {
@@ -463,7 +508,46 @@ nsHtml5Parser::CancelParsingEvents()
 void
 nsHtml5Parser::Reset()
 {
-  NS_NOTREACHED("Can't reset.");
+    mNeedsCharsetSwitch = PR_FALSE;
+    mLastWasCR = PR_FALSE;
+    mTerminated = PR_FALSE;
+    mLayoutStarted = PR_FALSE;
+    mFragmentMode = PR_FALSE;
+    mBlocked = PR_FALSE;
+    mSuspending = PR_FALSE;
+    mLifeCycle = NOT_STARTED;
+    mStreamListenerState = eNone;
+
+    mScriptElement = nsnull;
+    mScriptsExecuting = 0;
+     
+    mRootContextKey = nsnull;
+    mRequest = nsnull; 
+    mObserver = nsnull;
+    mContinueEvent = nsnull;  // weak ref
+ 
+    // tree-related stuff
+    mDocElement = nsnull; // weak ref 
+  
+    // encoding-related stuff
+    mCharsetSource = kCharsetUninitialized;
+    mCharset.Assign("");
+    mPendingCharset.Assign("");
+    mUnicodeDecoder = nsnull;
+    mSniffingBuffer = nsnull;
+    mSniffingLength = 0;
+    mBomState = BOM_SNIFFING_NOT_STARTED;
+
+    delete mMetaScanner;
+    mMetaScanner = nsnull;
+        
+    // Portable parser objects    
+    nsHtml5UTF16Buffer*          mFirstBuffer; // manually managed strong ref
+    nsHtml5UTF16Buffer*          mLastBuffer; // weak ref; always points to 
+                      // a buffer of the size NS_HTML5_PARSER_READ_BUFFER_SIZE
+    nsHtml5TreeBuilder*          mTreeBuilder; // manually managed strong ref
+    nsHtml5Tokenizer*            mTokenizer; // manually managed strong ref
+
 }
     
 /* End nsIParser  */

content/html/parser/src/nsHtml5Parser.h

@@ -216,6 +216,10 @@ class nsHtml5Parser : public nsIParser,
                              const nsACString& aContentType,
                              nsDTDMode aMode = eDTDMode_autodetect);
 
+    NS_IMETHOD ParseFragment(const nsAString& aSourceBuffer,
+                             nsISupports* aTargetNode,
+                             nsIAtom* aContextLocalName,
+                             PRInt32 aContextNamespace);
 
     /**
      * This method gets called when the tokens have been consumed, and it's time

content/html/parser/src/nsHtml5TreeBuilder.cpp

@@ -74,7 +74,12 @@ nsHtml5TreeBuilder::startTokenization(nsHtml5Tokenizer* self)
   start(fragment);
   startCoalescing();
   if (fragment) {
-    nsIContent* elt = createHtmlElementSetAsRoot(tokenizer->emptyAttributes());
+    nsIContent* elt;
+    if (!!contextNode) {
+      elt = contextNode;
+    } else {
+      elt = createHtmlElementSetAsRoot(tokenizer->emptyAttributes());
+    }
     nsHtml5StackNode* node = new nsHtml5StackNode(kNameSpaceID_XHTML, nsHtml5ElementName::HTML, elt);
     currentPtr++;
     stack[currentPtr] = node;
@@ -88,6 +93,10 @@ nsHtml5TreeBuilder::startTokenization(nsHtml5Tokenizer* self)
     } else {
       tokenizer->setContentModelFlag(NS_HTML5TOKENIZER_DATA, contextName);
     }
+    nsHtml5Portability::releaseLocal(contextName);
+    contextName = nsnull;
+    nsHtml5Portability::releaseElement(contextNode);
+    contextNode = nsnull;
     nsHtml5Portability::releaseElement(elt);
   } else {
     mode = NS_HTML5TREE_BUILDER_INITIAL;
@@ -2717,12 +2726,15 @@ nsHtml5TreeBuilder::resetTheInsertionMode()
   foreignFlag = NS_HTML5TREE_BUILDER_NOT_IN_FOREIGN;
   nsHtml5StackNode* node;
   nsIAtom* name;
+  PRInt32 ns;
   for (PRInt32 i = currentPtr; i >= 0; i--) {
     node = stack[i];
     name = node->name;
+    ns = node->ns;
     if (!i) {
-      if (!(contextName == nsHtml5Atoms::td || contextName == nsHtml5Atoms::th)) {
+      if (!(contextNamespace == kNameSpaceID_XHTML && (contextName == nsHtml5Atoms::td || contextName == nsHtml5Atoms::th))) {
         name = contextName;
+        ns = contextNamespace;
       } else {
         mode = NS_HTML5TREE_BUILDER_IN_BODY;
         return;
@@ -3468,6 +3480,17 @@ nsHtml5TreeBuilder::requestSuspension()
   tokenizer->requestSuspension();
 }
 
+void 
+nsHtml5TreeBuilder::setFragmentContext(nsIAtom* context, PRInt32 ns, nsIContent* node)
+{
+  this->contextName = context;
+  nsHtml5Portability::retainLocal(context);
+  this->contextNamespace = ns;
+  this->contextNode = node;
+  nsHtml5Portability::retainElement(node);
+  this->fragment = (!!contextName);
+}
+
 nsIContent* 
 nsHtml5TreeBuilder::currentNode()
 {

content/html/parser/src/nsHtml5TreeBuilder.h

@@ -77,6 +77,7 @@ class nsHtml5TreeBuilder
     PRBool fragment;
     nsIAtom* contextName;
     PRInt32 contextNamespace;
+    nsIContent* contextNode;
     jArray<nsHtml5StackNode*,PRInt32> stack;
     PRInt32 currentPtr;
     jArray<nsHtml5StackNode*,PRInt32> listOfActiveFormattingElements;
@@ -187,6 +188,9 @@ class nsHtml5TreeBuilder
     void appendDoctypeToDocument(nsIAtom* name, nsString* publicIdentifier, nsString* systemIdentifier);
     void elementPushed(PRInt32 ns, nsIAtom* name, nsIContent* node);
     void elementPopped(PRInt32 ns, nsIAtom* name, nsIContent* node);
+  public:
+    void setFragmentContext(nsIAtom* context, PRInt32 ns, nsIContent* node);
+  protected:
     nsIContent* currentNode();
   public:
     PRBool isScriptingEnabled();

content/html/parser/src/nsHtml5TreeBuilderCppSupplement.h

@@ -291,9 +291,13 @@ nsHtml5TreeBuilder::endCoalescing()
 void
 nsHtml5TreeBuilder::start(PRBool fragment)
 {
-  mHasProcessedBase = PR_FALSE;
-  mParser->WillBuildModelImpl();
-  mParser->GetDocument()->BeginLoad(); // XXX fragment?
+  if (fragment) {
+    mHasProcessedBase = PR_TRUE;  
+  } else {
+    mHasProcessedBase = PR_FALSE;
+    mParser->WillBuildModelImpl();
+    mParser->GetDocument()->BeginLoad(); // XXX fragment?
+  }
 }
 
 void

parser/htmlparser/public/nsIParser.h

@@ -52,6 +52,7 @@
 #include "nsIDTD.h"
 #include "nsStringGlue.h"
 #include "nsTArray.h"
+#include "nsIAtom.h"
 
 // 3007e9c0-4d3e-4c80-8cae-fbb4723d88f2
 #define NS_IPARSER_IID \
@@ -258,6 +259,11 @@ class nsIParser : public nsISupports {
                              const nsACString& aContentType,
                              nsDTDMode aMode = eDTDMode_autodetect) = 0;
 
+    NS_IMETHOD ParseFragment(const nsAString& aSourceBuffer,
+                             nsISupports* aTargetNode,
+                             nsIAtom* aContextLocalName,
+                             PRInt32 aContextNamespace) = 0;
+
     /**
      * This method gets called when the tokens have been consumed, and it's time
      * to build the model via the content sink.

parser/htmlparser/src/nsParser.cpp

@@ -2047,6 +2047,15 @@ nsParser::Parse(const nsAString& aSourceBuffer,
   return result;
 }
 
+NS_IMETHODIMP
+nsParser::ParseFragment(const nsAString& aSourceBuffer,
+                        nsISupports* aTargetNode,
+                        nsIAtom* aContextLocalName,
+                        PRInt32 aContextNamespace)
+{
+  return NS_ERROR_NOT_IMPLEMENTED;
+}
+
 NS_IMETHODIMP
 nsParser::ParseFragment(const nsAString& aSourceBuffer,
                         void* aKey,

parser/htmlparser/src/nsParser.h

@@ -224,7 +224,11 @@ class nsParser : public nsIParser,
                              const nsACString& aContentType,
                              nsDTDMode aMode = eDTDMode_autodetect);
 
-
+    NS_IMETHOD ParseFragment(const nsAString& aSourceBuffer,
+                             nsISupports* aTargetNode,
+                             nsIAtom* aContextLocalName,
+                             PRInt32 aContextNamespace);
+                             
     /**
      * This method gets called when the tokens have been consumed, and it's time
      * to build the model via the content sink.