Speed and GC optimizations in html()

Attribute: removed regex matchers to check for key validity, using a simple scan now.

Entities: use a charBuf to append supplemental characters

Improves on #2183

CHANGES.md

@@ -10,7 +10,7 @@
   root `Document` node. This enables source position tracking on the Document node (which was previously unset). And
   it also enables the node traversor to see the outer Document node. [2182](https://github.com/jhy/jsoup/pull/2182)
 * Speed optimized `html()` and `Entities.escape()` when the input contains UTF characters in a supplementary plane, by
-  around 22%. [2183](https://github.com/jhy/jsoup/pull/2183)
+  around 49%. [2183](https://github.com/jhy/jsoup/pull/2183)
 
 ### Bug Fixes
 

src/main/java/org/jsoup/nodes/Attribute.java

@@ -174,29 +174,55 @@ static void htmlNoValidate(String key, @Nullable String val, Appendable accum, D
         }
     }
 
-    private static final Pattern xmlKeyValid = Pattern.compile("[a-zA-Z_:][-a-zA-Z0-9_:.]*");
     private static final Pattern xmlKeyReplace = Pattern.compile("[^-a-zA-Z0-9_:.]+");
-    private static final Pattern htmlKeyValid = Pattern.compile("[^\\x00-\\x1f\\x7f-\\x9f \"'/=]+");
     private static final Pattern htmlKeyReplace = Pattern.compile("[\\x00-\\x1f\\x7f-\\x9f \"'/=]+");
-
     /**
      * Get a valid attribute key for the given syntax. If the key is not valid, it will be coerced into a valid key.
      * @param key the original attribute key
      * @param syntax HTML or XML
      * @return the original key if it's valid; a key with invalid characters replaced with "_" otherwise; or null if a valid key could not be created.
      */
     @Nullable public static String getValidKey(String key, Syntax syntax) {
-        if (syntax == Syntax.xml && !xmlKeyValid.matcher(key).matches()) {
+        if (syntax == Syntax.xml && !isValidXmlKey(key)) {
             key = xmlKeyReplace.matcher(key).replaceAll("_");
-            return xmlKeyValid.matcher(key).matches() ? key : null; // null if could not be coerced
+            return isValidXmlKey(key) ? key : null; // null if could not be coerced
         }
-        else if (syntax == Syntax.html && !htmlKeyValid.matcher(key).matches()) {
+        else if (syntax == Syntax.html && !isValidHtmlKey(key)) {
             key = htmlKeyReplace.matcher(key).replaceAll("_");
-            return htmlKeyValid.matcher(key).matches() ? key : null; // null if could not be coerced
+            return isValidHtmlKey(key) ? key : null; // null if could not be coerced
         }
         return key;
     }
 
+    // perf critical in html() so using manual scan vs regex:
+    // note that we aren't using anything in supplemental space, so OK to iter charAt
+    private static boolean isValidXmlKey(String key) {
+        // =~ [a-zA-Z_:][-a-zA-Z0-9_:.]*
+        final int length = key.length();
+        if (length ==0) return false;
+        char c = key.charAt(0);
+        if (!((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_' || c == ':'))
+            return false;
+        for (int i = 1; i < length; i++) {
+            c = key.charAt(i);
+            if (!((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_' || c == ':'))
+                return false;
+        }
+        return true;
+    }
+
+    private static boolean isValidHtmlKey(String key) {
+        // =~ [\x00-\x1f\x7f-\x9f "'/=]+
+        final int length = key.length();
+        if (length ==0) return false;
+        for (int i = 0; i < length; i++) {
+            char c = key.charAt(i);
+            if (c <= 0x1f || c >= 0x7f && c <= 0x9f || c == ' ' || c == '"' || c == '\'' || c == '/' || c == '=')
+                return false;
+        }
+        return true;
+    }
+
     /**
      Get the string representation of this attribute, implemented as {@link #html()}.
      @return string

src/main/java/org/jsoup/nodes/Entities.java

@@ -248,12 +248,21 @@ private static void appendEscaped(Appendable accum, OutputSettings out, int opti
             }
         } else {
             if (canEncode(coreCharset, c, encoder)) {
-                String s = new String(Character.toChars(codePoint));
-                accum.append(s);
-            } else appendEncoded(accum, escapeMode, codePoint);
+                // reads into charBuf - we go through these steps to avoid GC objects as much as possible (would be a new String and a new char[2] for each character)
+                char[] chars = charBuf.get();
+                int len = Character.toChars(codePoint, chars, 0);
+                if (accum instanceof StringBuilder) // true unless the user supplied their own
+                    ((StringBuilder) accum).append(chars, 0, len);
+                else
+                    accum.append(new String(chars, 0, len));
+            } else {
+                appendEncoded(accum, escapeMode, codePoint);
+            }
         }
     }
 
+    private static final ThreadLocal<char[]> charBuf = ThreadLocal.withInitial(() -> new char[2]);
+
     private static void appendNbsp(Appendable accum, EscapeMode escapeMode) throws IOException {
         if (escapeMode != EscapeMode.xhtml) accum.append("&nbsp;");
         else accum.append("&#xa0;");