Changeset 295167 in webkit

Timestamp:

Jun 2, 2022, 7:08:02 PM (4 years ago)

Author:

Chris Dumez

Message:

Optimize HTMLTokenizer::processEntity()
​https://bugs.webkit.org/show_bug.cgi?id=241239

Reviewed by Darin Adler.

Optimize HTMLTokenizer::processEntity() by appending all characters in the
StringBuffer at once, instead of one by one.

• Source/WebCore/html/parser/HTMLToken.h:

(WebCore::HTMLToken::appendToCharacters):

• Source/WebCore/html/parser/HTMLTokenizer.cpp:

(WebCore::HTMLTokenizer::bufferCharacters):
(WebCore::HTMLTokenizer::processEntity):

• Source/WebCore/html/parser/HTMLTokenizer.h:

Canonical link: ​https://commits.webkit.org/251251@main

Location:

trunk/Source

Files:

• WTF/wtf/text/StringBuilder.h (modified) (1 diff)
• WebCore/html/parser/HTMLToken.h (modified) (2 diffs)
• WebCore/html/parser/HTMLTokenizer.cpp (modified) (2 diffs)
• WebCore/html/parser/HTMLTokenizer.h (modified) (1 diff)

trunk/Source/WTF/wtf/text/StringBuilder.h

@@ -94 +94 @@
     const LChar* characters8() const { return characters<LChar>(); }
     const UChar* characters16() const { return characters<UChar>(); }
+    template<typename CharacterType> Span<const CharacterType> span() const { return Span { characters<CharacterType>(), length() }; }
     
     unsigned capacity() const;

trunk/Source/WebCore/html/parser/HTMLToken.h

@@ -123 +123 @@
     void appendToCharacter(UChar);
     void appendToCharacter(const Vector<LChar, 32>&);
+    template<typename CharacterType> void appendToCharacter(Span<const CharacterType>);
 
     // Comment.
@@ -374 +375 @@
     m_type = Character;
     m_data.appendVector(characters);
+}
+
+template<typename CharacterType>
+inline void HTMLToken::appendToCharacter(Span<const CharacterType> characters)
+{
+    m_type = Character;
+    m_data.append(characters);
+    if constexpr (std::is_same_v<CharacterType, UChar>) {
+        if (!charactersIsAll8BitData())
+            return;
+        for (auto character : characters)
+            m_data8BitCheck |= character;
+    }
 }
 

trunk/Source/WebCore/html/parser/HTMLTokenizer.cpp

@@ -82 +82 @@
 }
 
+template<typename CharacterType>
+inline void HTMLTokenizer::bufferCharacters(Span<const CharacterType> characters)
+{
+#if ASSERT_ENABLED
+    for (auto character : characters)
+        ASSERT(character != kEndOfFileMarker);
+#endif
+    m_token.appendToCharacter(characters);
+}
+
 inline bool HTMLTokenizer::emitAndResumeInDataState(SegmentedString& source)
 {
@@ -131 +141 @@
         bufferASCIICharacter('&');
     } else {
-        for (unsigned i = 0; i < decodedEntity.length(); ++i)
-            bufferCharacter(decodedEntity[i]);
+        if (decodedEntity.is8Bit())
+            bufferCharacters(decodedEntity.span<LChar>());
+        else
+            bufferCharacters(decodedEntity.span<UChar>());
     }
     return true;

trunk/Source/WebCore/html/parser/HTMLTokenizer.h

@@ -160 +160 @@
     void bufferASCIICharacter(UChar);
     void bufferCharacter(UChar);
+    template<typename CharacterType> void bufferCharacters(Span<const CharacterType>);
 
     bool emitAndResumeInDataState(SegmentedString&);