Changeset 159764 in webkit

Timestamp:

Nov 25, 2013 1:53:32 PM (10 years ago)

Author:

akling@apple.com

Message:

Deduplicate shortish Text node strings during tree construction.
<​https://webkit.org/b/124855>

Let HTMLConstructionSite keep a hash set of already seen strings over
its lifetime. Use this to deduplicate the strings inside Text nodes
for any string up to 64 characters of length.

This optimization already sort-of existed for whitespace-only Texts,
but those are laundered in the AtomicString table which we definitely
don't want to pollute with every single Text. It might be a good idea
to stop using the AtomicString table for all-whitespace Text too.

3.82 MB progression on HTML5-8266 locally.

Reviewed by Anders Carlsson.

Location:

trunk/Source/WebCore

Files:

• ChangeLog (modified) (1 diff)
• html/parser/HTMLConstructionSite.cpp (modified) (2 diffs)
• html/parser/HTMLConstructionSite.h (modified) (2 diffs)

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2013-11-25  Andreas Kling  <akling@apple.com>
+
+        Deduplicate shortish Text node strings during tree construction.
+        <https://webkit.org/b/124855>
+
+        Let HTMLConstructionSite keep a hash set of already seen strings over
+        its lifetime. Use this to deduplicate the strings inside Text nodes
+        for any string up to 64 characters of length.
+
+        This optimization already sort-of existed for whitespace-only Texts,
+        but those are laundered in the AtomicString table which we definitely
+        don't want to pollute with every single Text. It might be a good idea
+        to stop using the AtomicString table for all-whitespace Text too.
+
+        3.82 MB progression on HTML5-8266 locally.
+
+        Reviewed by Anders Carlsson.
+
 2013-11-25  Nick Diego Yamane  <nick.yamane@openbossa.org>
 

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

@@ -506 +506 @@
 
     while (currentPosition < characters.length()) {
-        RefPtr<Text> textNode = Text::createWithLengthLimit(task.parent->document(), shouldUseAtomicString ? AtomicString(characters).string() : characters, currentPosition, lengthLimit);
+        RefPtr<Text> textNode = Text::createWithLengthLimit(task.parent->document(), stringForTextNode(characters, shouldUseAtomicString), currentPosition, lengthLimit);
         // If we have a whole string of unbreakable characters the above could lead to an infinite loop. Exceeding the length limit is the lesser evil.
         if (!textNode->length()) {
             String substring = characters.substring(currentPosition);
-            textNode = Text::create(task.parent->document(), shouldUseAtomicString ? AtomicString(substring).string() : substring);
+            textNode = Text::create(task.parent->document(), stringForTextNode(substring, shouldUseAtomicString));
         }
 
@@ -667 +667 @@
 }
 
-}
+String HTMLConstructionSite::stringForTextNode(const String& string, bool shouldUseAtomicString)
+{
+    static const unsigned maximumLengthForDeduplication = 64;
+    if (shouldUseAtomicString)
+        return AtomicString(string).string();
+    if (string.length() > maximumLengthForDeduplication)
+        return string;
+    return *m_stringsForDeduplication.add(string).iterator;
+}
+
+}

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

@@ -171 +171 @@
     void dispatchDocumentElementAvailableIfNeeded();
 
+    String stringForTextNode(const String&, bool shouldUseAtomicString);
+
     Document* m_document;
     
@@ -197 +199 @@
 
     bool m_inQuirksMode;
+
+    HashSet<String> m_stringsForDeduplication;
 };