https://bugs.webkit.org/show_bug.cgi?id=57073

Rework the AtomicHTMLConstructor to reserve space for attributes based on the size of the attribute list.
Saves 1.5 MB on Membuster.

Reviewed by Geoff Garen.

No new tests because there was no change in behavior.  Current tests pass.

* html/parser/HTMLToken.h:
(WebCore::AtomicHTMLToken::AtomicHTMLToken):
(WebCore::AtomicHTMLToken::initializeAttributes):



git-svn-id: svn://svn.chromium.org/blink/trunk@81937 bbb929c8-8fbe-4397-9dbb-9b2b20218538

third_party/WebKit/Source/WebCore/ChangeLog

+2011-03-24  Stephanie Lewis  <slewis@apple.com>
+
+        Reviewed by Geoff Garen.
+
+        https://bugs.webkit.org/show_bug.cgi?id=57073
+        Rework the AtomicHTMLConstructor to reserve space for attributes based on the size of the attribute list.
+        Saves 1.5 MB on Membuster.
+
+        No new tests because there was no change in behavior.  Current tests pass.
+
+        * html/parser/HTMLToken.h:
+        (WebCore::AtomicHTMLToken::AtomicHTMLToken):
+        (WebCore::AtomicHTMLToken::initializeAttributes):
+
 2011-03-24  Sheriff Bot  <webkit.review.bot@gmail.com>
 
         Unreviewed, rolling out r81916 and r81917.

third_party/WebKit/Source/WebCore/html/parser/HTMLToken.h

@@ -398,27 +398,7 @@ public:
         case HTMLToken::EndTag: {
             m_selfClosing = token.selfClosing();
             m_name = AtomicString(token.name().data(), token.name().size());
-            const HTMLToken::AttributeList& attributes = token.attributes();
-            for (HTMLToken::AttributeList::const_iterator iter = attributes.begin();
-                 iter != attributes.end(); ++iter) {
-                if (!iter->m_name.isEmpty()) {
-                    String name(iter->m_name.data(), iter->m_name.size());
-                    String value(iter->m_value.data(), iter->m_value.size());
-                    ASSERT(iter->m_nameRange.m_start);
-                    ASSERT(iter->m_nameRange.m_end);
-                    ASSERT(iter->m_valueRange.m_start);
-                    ASSERT(iter->m_valueRange.m_end);
-                    RefPtr<Attribute> mappedAttribute = Attribute::createMapped(name, value);
-                    if (!m_attributes) {
-                        m_attributes = NamedNodeMap::create();
-                        // Reserving capacity here improves the parser
-                        // benchmark.  It might be worth experimenting with
-                        // the constant to see where the optimal point is.
-                        m_attributes->reserveInitialCapacity(10);
-                    }
-                    m_attributes->insertAttribute(mappedAttribute.release(), false);
-                }
-            }
+            initializeAttributes(token.attributes());
             break;
         }
         case HTMLToken::Comment:
@@ -513,6 +493,8 @@ public:
 private:
     HTMLToken::Type m_type;
 
+    void initializeAttributes(const HTMLToken::AttributeList& attributes);
+    
     bool usesName() const
     {
         return m_type == HTMLToken::StartTag || m_type == HTMLToken::EndTag || m_type == HTMLToken::DOCTYPE;
@@ -548,6 +530,30 @@ private:
     RefPtr<NamedNodeMap> m_attributes;
 };
 
+inline void AtomicHTMLToken::initializeAttributes(const HTMLToken::AttributeList& attributes)
+{
+    size_t size = attributes.size();
+    if (!size)
+        return;
+
+    m_attributes = NamedNodeMap::create();
+    m_attributes->reserveInitialCapacity(size);
+    for (size_t i = 0; i < size; ++i) {
+        const HTMLToken::Attribute& attribute = attributes[i];
+        if (attribute.m_name.isEmpty())
+            continue;
+
+        ASSERT(attribute.m_nameRange.m_start);
+        ASSERT(attribute.m_nameRange.m_end);
+        ASSERT(attribute.m_valueRange.m_start);
+        ASSERT(attribute.m_valueRange.m_end);
+
+        String name(attribute.m_name.data(), attribute.m_name.size());
+        String value(attribute.m_value.data(), attribute.m_value.size());
+        m_attributes->insertAttribute(Attribute::createMapped(name, value), false);
+    }
+}
+
 }
 
 #endif