Fix countCodePoints range scan for non-zero offsets

countCodePoints(Slice, offset, length) mixed absolute and relative
bounds when chunk-scanning, so non-zero offsets could scan the wrong
byte window and return incorrect code point counts.

Author: dain

src/main/java/io/airlift/slice/SliceUtf8.java

@@ -120,22 +120,22 @@ public static int countCodePoints(Slice utf8, int offset, int length)
             return 0;
         }
 
+        int end = offset + length;
         int continuationBytesCount = 0;
-        // Length rounded to 8 bytes
-        int length8 = length & 0x7FFF_FFF8;
-        for (; offset < length8; offset += 8) {
+        int lastLongStart = end - 8;
+        for (; offset <= lastLongStart; offset += 8) {
             // Count bytes which are NOT the start of a code point
             continuationBytesCount += countContinuationBytes(utf8.getLongUnchecked(offset));
         }
         // Enough bytes left for 32 bits?
-        if (offset + 4 < length) {
+        if (offset <= end - 4) {
             // Count bytes which are NOT the start of a code point
             continuationBytesCount += countContinuationBytes(utf8.getIntUnchecked(offset));
 
             offset += 4;
         }
         // Do the rest one by one
-        for (; offset < length; offset++) {
+        for (; offset < end; offset++) {
             // Count bytes which are NOT the start of a code point
             continuationBytesCount += countContinuationBytes(utf8.getByteUnchecked(offset));
         }

src/test/java/io/airlift/slice/TestSliceUtf8.java

@@ -192,6 +192,18 @@ public void testCodePointCount()
         assertThat(countCodePoints(wrappedBuffer(CONTINUATION_BYTE))).isEqualTo(0);
     }
 
+    @Test
+    public void testCodePointCountRange()
+    {
+        Slice utf8 = utf8Slice("€é€😀😀😀");
+        for (int offset = 0; offset <= utf8.length(); offset++) {
+            for (int length = 0; length <= utf8.length() - offset; length++) {
+                assertThat(countCodePoints(utf8, offset, length))
+                        .isEqualTo(countCodePoints(utf8.slice(offset, length)));
+            }
+        }
+    }
+
     private static void assertCodePointCount(String string)
     {
         assertThat(countCodePoints(utf8Slice(string))).isEqualTo(string.codePoints().count());