Fixed a performance regression in 1.12

Was building useless stringbuilders on every token read!
jhy · Feb 23, 2020 · 9d9e53c · 9d9e53c
1 parent a657ae0
commit 9d9e53c
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 11 deletions.
diff --git a/src/main/java/org/jsoup/parser/Tokeniser.java b/src/main/java/org/jsoup/parser/Tokeniser.java
@@ -53,21 +53,20 @@ final class Tokeniser {
     }
 
     Token read() {
-        int pos = reader.pos(); // count how many reads we do in a row without making progress, and bail if stuck in a loop
+        final CharacterReader r = this.reader;
+        final int pos = r.pos(); // count how many reads we do in a row without making progress, and bail if stuck in a loop
         int consecutiveReads = 0;
         while (!isEmitPending) {
-            state.read(this, reader);
-            if (reader.pos() <= pos) {
-                consecutiveReads++;
-            }
-            Validate.isTrue(consecutiveReads < 10,
-                "BUG: Not making progress from state: " + this.state.name() + " with current char=" + reader.current());
+            state.read(this, r);
+            if (++consecutiveReads > 10 && r.pos() <= pos)
+                Validate.wtf("BUG: Not making progress from state: " + this.state.name() + " with current char=" + r.current());
         }
 
         // if emit is pending, a non-character token was found: return any chars in buffer, and leave token for next read:
-        if (charsBuilder.length() > 0) {
-            String str = charsBuilder.toString();
-            charsBuilder.delete(0, charsBuilder.length());
+        final StringBuilder cb = this.charsBuilder;
+        if (cb.length() != 0) {
+            String str = cb.toString();
+            cb.delete(0, cb.length());
             charsString = null;
             return charPending.data(str);
         } else if (charsString != null) {

diff --git a/src/main/java/org/jsoup/parser/TreeBuilder.java b/src/main/java/org/jsoup/parser/TreeBuilder.java
@@ -58,26 +58,31 @@ Document parse(Reader input, String baseUri, Parser parser) {
     abstract List<Node> parseFragment(String inputFragment, Element context, String baseUri, Parser parser);
 
     protected void runParser() {
+        final Tokeniser tokeniser = this.tokeniser;
+        final Token.TokenType eof = Token.TokenType.EOF;
+
         while (true) {
             Token token = tokeniser.read();
             process(token);
             token.reset();
 
-            if (token.type == Token.TokenType.EOF)
+            if (token.type == eof)
                 break;
         }
     }
 
     protected abstract boolean process(Token token);
 
     protected boolean processStartTag(String name) {
+        final Token.StartTag start = this.start;
         if (currentToken == start) { // don't recycle an in-use token
             return process(new Token.StartTag().name(name));
         }
         return process(start.reset().name(name));
     }
 
     public boolean processStartTag(String name, Attributes attrs) {
+        final Token.StartTag start = this.start;
         if (currentToken == start) { // don't recycle an in-use token
             return process(new Token.StartTag().nameAttr(name, attrs));
         }