(#220) Switch to lookup tables in hotspots isNameChar()/isNameStartChar() (#221)

Author: winfriedgerlach

release-notes/VERSION

@@ -9,6 +9,8 @@ Project: woodstox
 #213: SAX: `Locator#getSystemId` and `Locator#getPublicId` are not
   available during `startDocument` event
  (fix contributed by Philipp N)
+#221: Switch to lookup tables in hotspots `isNameChar()`/`isNameStartChar()`
+ (contributed by @winfriedgerlach)
 
 7.1.0 (22-Oct-2024)
 

src/main/java/com/ctc/wstx/io/WstxInputData.java

@@ -17,6 +17,8 @@
 
 import com.ctc.wstx.util.XmlChars;
 
+import java.util.stream.IntStream;
+
 /**
  * Base class used by readers (specifically, by
  * {@link com.ctc.wstx.sr.StreamScanner}, and its sub-classes)
@@ -50,6 +52,23 @@ public class WstxInputData
      */
     public final static int MAX_UNICODE_CHAR = 0x10FFFF;
 
+    private static final boolean[] asciiNameStartChars = new boolean[128];
+    static {
+        IntStream.rangeClosed('a', 'z').forEach(i -> asciiNameStartChars[i] = true);
+        IntStream.rangeClosed('A', 'Z').forEach(i -> asciiNameStartChars[i] = true);
+        asciiNameStartChars['_'] = true;
+    }
+
+    private static final boolean[] asciiNameChars = new boolean[128];
+    static {
+        IntStream.rangeClosed('a', 'z').forEach(i -> asciiNameChars[i] = true);
+        IntStream.rangeClosed('A', 'Z').forEach(i -> asciiNameChars[i] = true);
+        IntStream.rangeClosed('0', '9').forEach(i -> asciiNameChars[i] = true);
+        asciiNameChars['.'] = true;
+        asciiNameChars['-'] = true;
+        asciiNameChars['_'] = true;
+    }
+
     /*
     ////////////////////////////////////////////////////
     // Configuration
@@ -153,14 +172,9 @@ protected final boolean isNameStartChar(char c)
         /* First, let's handle 7-bit ascii range (identical between xml
          * 1.0 and 1.1)
          */
-        if (c <= 0x7A) { // 'z' or earlier
-            if (c >= 0x61) { // 'a' - 'z' are ok
-                return true;
-            }
-            if (c < 0x41) { // before 'A' just white space
-                return false;
-            }
-            return (c <= 0x5A) || (c == '_'); // 'A' - 'Z' and '_' are ok
+        if (c < 128) {
+            // this is performance critical, so we use a lookup table instead of if-branches
+            return asciiNameStartChars[c];
         }
         /* Ok, otherwise need to use a big honking bit sets... which
          * differ between 1.0 and 1.1
@@ -178,18 +192,9 @@ protected final boolean isNameStartChar(char c)
     protected final boolean isNameChar(char c)
     {
         // First, let's handle 7-bit ascii range
-        if (c <= 0x7A) { // 'z' or earlier
-            if (c >= 0x61) { // 'a' - 'z' are ok
-                return true;
-            }
-            if (c <= 0x5A) {
-                if (c >= 0x41) { // 'A' - 'Z' ok too
-                    return true;
-                }
-                // As are 0-9, '.' and '-'
-                return (c >= 0x30 && c <= 0x39) || (c == '.') || (c == '-');
-            }
-            return (c == 0x5F); // '_' is ok too
+        if (c < 128) {
+            // this is performance critical, so we use a lookup table instead of if-branches
+            return asciiNameChars[c];
         }
         return mXml11 ? XmlChars.is11NameChar(c) : XmlChars.is10NameChar(c);
     }

src/test/java/com/ctc/wstx/io/WstxInputDataTest.java

@@ -0,0 +1,77 @@
+package com.ctc.wstx.io;
+
+import com.ctc.wstx.util.XmlChars;
+import junit.framework.TestCase;
+import org.junit.Test;
+
+import java.util.stream.IntStream;
+
+public class WstxInputDataTest extends TestCase {
+
+    @Test
+    public void testIsNameStartCharBehavesSameAsBranchyVersion() {
+        WstxInputData wstxInputDataXml10 = new WstxInputData();
+        WstxInputData wstxInputDataXml11 = new WstxInputData();
+        wstxInputDataXml11.mXml11 = true;
+
+        // include all 7-bit ASCII characters plus some left and right
+        IntStream.range(-10, 138).forEach(i -> {
+            char c = (char) i;
+            assertEquals(isNameStartCharBranchy(c, false), wstxInputDataXml10.isNameStartChar(c));
+            assertEquals(isNameStartCharBranchy(c, true), wstxInputDataXml11.isNameStartChar(c));
+        });
+    }
+
+    // previous implementation with branches
+    private final boolean isNameStartCharBranchy(char c, boolean mXml11) {
+        /* First, let's handle 7-bit ascii range (identical between xml
+         * 1.0 and 1.1)
+         */
+        if (c <= 0x7A) { // 'z' or earlier
+            if (c >= 0x61) { // 'a' - 'z' are ok
+                return true;
+            }
+            if (c < 0x41) { // before 'A' just white space
+                return false;
+            }
+            return (c <= 0x5A) || (c == '_'); // 'A' - 'Z' and '_' are ok
+        }
+        /* Ok, otherwise need to use a big honking bit sets... which
+         * differ between 1.0 and 1.1
+         */
+        return mXml11 ? XmlChars.is11NameStartChar(c) : XmlChars.is10NameStartChar(c);
+    }
+
+    @Test
+    public void testIsNameCharBehavesSameAsBranchyVersion() {
+        WstxInputData wstxInputDataXml10 = new WstxInputData();
+        WstxInputData wstxInputDataXml11 = new WstxInputData();
+        wstxInputDataXml11.mXml11 = true;
+
+        // include all 7-bit ASCII characters plus some left and right
+        IntStream.range(-10, 138).forEach(i -> {
+            char c = (char) i;
+            assertEquals(isNameCharBranchy(c, false), wstxInputDataXml10.isNameChar(c));
+            assertEquals(isNameCharBranchy(c, true), wstxInputDataXml11.isNameChar(c));
+        });
+    }
+
+    // previous implementation with branches
+    private final boolean isNameCharBranchy(char c, boolean mXml11) {
+        // First, let's handle 7-bit ascii range
+        if (c <= 0x7A) { // 'z' or earlier
+            if (c >= 0x61) { // 'a' - 'z' are ok
+                return true;
+            }
+            if (c <= 0x5A) {
+                if (c >= 0x41) { // 'A' - 'Z' ok too
+                    return true;
+                }
+                // As are 0-9, '.' and '-'
+                return (c >= 0x30 && c <= 0x39) || (c == '.') || (c == '-');
+            }
+            return (c == 0x5F); // '_' is ok too
+        }
+        return mXml11 ? XmlChars.is11NameChar(c) : XmlChars.is10NameChar(c);
+    }
+}