From d09a1069790832e53171bb7d1c8a18f8ddcbfeab Mon Sep 17 00:00:00 2001
From: Benoit Tellier <btellier@linagora.com>
Date: Fri, 1 Jul 2022 11:16:36 +0700
Subject: [PATCH] ISSUE-1773 Tokeniser: recycle string builders too

---
 .../jsoup/helper/StringBuilderRecycler.java   | 23 ++++++++++++++++
 src/main/java/org/jsoup/parser/Tokeniser.java | 27 +++++++++++++++++--
 .../java/org/jsoup/parser/TreeBuilder.java    |  1 +
 3 files changed, 49 insertions(+), 2 deletions(-)
 create mode 100644 src/main/java/org/jsoup/helper/StringBuilderRecycler.java

diff --git a/src/main/java/org/jsoup/helper/StringBuilderRecycler.java b/src/main/java/org/jsoup/helper/StringBuilderRecycler.java
new file mode 100644
index 0000000000..35f6adbb0c
--- /dev/null
+++ b/src/main/java/org/jsoup/helper/StringBuilderRecycler.java
@@ -0,0 +1,23 @@
+package org.jsoup.helper;
+
+import java.util.ArrayList;
+
+public class StringBuilderRecycler {
+    protected final ArrayList<StringBuilder> stringBuilders = new ArrayList<>();
+
+    public StringBuilder get(int minSize) {
+        if (!stringBuilders.isEmpty()) {
+            StringBuilder stringBuilder = stringBuilders.remove(stringBuilders.size() - 1);
+            // Too small string builders are thrown away
+            if (stringBuilder.capacity() >= minSize) {
+                stringBuilder.setLength(0);
+                return stringBuilder;
+            }
+        }
+        return new StringBuilder(minSize);
+    }
+
+    public void releaseByteBuffer(StringBuilder stringBuilder) {
+        stringBuilders.add(stringBuilder);
+    }
+}
diff --git a/src/main/java/org/jsoup/parser/Tokeniser.java b/src/main/java/org/jsoup/parser/Tokeniser.java
index be00b7f63c..3d0771b73d 100644
--- a/src/main/java/org/jsoup/parser/Tokeniser.java
+++ b/src/main/java/org/jsoup/parser/Tokeniser.java
@@ -1,16 +1,33 @@
 package org.jsoup.parser;
 
+import org.jsoup.helper.StringBuilderRecycler;
 import org.jsoup.helper.Validate;
 import org.jsoup.internal.StringUtil;
 import org.jsoup.nodes.Entities;
 
 import javax.annotation.Nullable;
+
+import java.lang.ref.SoftReference;
 import java.util.Arrays;
 
 /**
  * Readers the input stream into tokens.
  */
 final class Tokeniser {
+    protected static final ThreadLocal<SoftReference<StringBuilderRecycler>> stringCacheRef = new ThreadLocal<>();
+
+    public static StringBuilderRecycler getStringBuilderCache() {
+        SoftReference<StringBuilderRecycler> ref = stringCacheRef.get();
+        StringBuilderRecycler br = (ref == null) ? null : ref.get();
+
+        if (br == null) {
+            br = new StringBuilderRecycler();
+            ref = new SoftReference<>(br);
+            stringCacheRef.set(ref);
+        }
+        return br;
+    }
+
     static final char replacementChar = '\uFFFD'; // replaces null character
     private static final char[] notCharRefCharsSorted = new char[]{'\t', '\n', '\r', '\f', ' ', '<', '&'};
 
@@ -37,8 +54,8 @@ final class Tokeniser {
     @Nullable private Token emitPending = null; // the token we are about to emit on next read
     private boolean isEmitPending = false;
     @Nullable private String charsString = null; // characters pending an emit. Will fall to charsBuilder if more than one
-    private final StringBuilder charsBuilder = new StringBuilder(1024); // buffers characters to output as one token, if more than one emit per read
-    StringBuilder dataBuffer = new StringBuilder(1024); // buffers data looking for </script>
+    private StringBuilder charsBuilder = getStringBuilderCache().get(1024); // buffers characters to output as one token, if more than one emit per read
+    StringBuilder dataBuffer = getStringBuilderCache().get(1024);  // buffers data looking for </script>
 
     Token.StartTag startPending = new Token.StartTag();
     Token.EndTag endPending = new Token.EndTag();
@@ -360,4 +377,10 @@ String unescapeEntities(boolean inAttribute) {
         }
         return StringUtil.releaseBuilder(builder);
     }
+
+    void release() {
+        StringBuilderRecycler stringBuilderCache = getStringBuilderCache();
+        stringBuilderCache.releaseByteBuffer(charsBuilder);
+        stringBuilderCache.releaseByteBuffer(dataBuffer);
+    }
 }
diff --git a/src/main/java/org/jsoup/parser/TreeBuilder.java b/src/main/java/org/jsoup/parser/TreeBuilder.java
index 902cbdf0ae..877cce82d8 100644
--- a/src/main/java/org/jsoup/parser/TreeBuilder.java
+++ b/src/main/java/org/jsoup/parser/TreeBuilder.java
@@ -63,6 +63,7 @@ Document parse(Reader input, String baseUri, Parser parser) {
         // tidy up - as the Parser and Treebuilder are retained in document for settings / fragments
         reader.close();
         reader = null;
+        tokeniser.release();
         tokeniser = null;
         stack = null;
         seenTags = null;