fix bug in comparing-to-empty-array; realize that we need compare-to-empty-array to specify start/finish of slices.

patch by jbellis; reviewed by Stu Hood for CASSANDRA-119

git-svn-id: https://svn.apache.org/repos/asf/incubator/cassandra/trunk@798273 13f79535-47bb-0310-9956-ffa450edef68

src/java/org/apache/cassandra/db/marshal/AbstractType.java

@@ -4,8 +4,15 @@ import java.util.Comparator;
 import java.util.Collection;
 
 import org.apache.cassandra.db.IColumn;
-import org.apache.cassandra.db.SuperColumn;
 
+/**
+ * Specifies a Comparator for a specific type of byte[].
+ *
+ * Note that empty byte[] are used to represent "start at the beginning"
+ * or "stop at the end" arguments to get_slice, so the Comparator
+ * should always handle those values even if they normally do not
+ * represent a valid byte[] for the type being compared.
+ */
 public abstract class AbstractType implements Comparator<byte[]>
 {
     /** get a string representation of the bytes suitable for log messages */

src/java/org/apache/cassandra/db/marshal/LongType.java

@@ -7,14 +7,13 @@ public class LongType extends AbstractType
 {
     public int compare(byte[] o1, byte[] o2)
     {
-        // TODO rm hack to support the "i'm going to pretend [] is an index entry if I didn't actually index anything" hack
         if (o1.length == 0)
         {
             return o2.length == 0 ? 0 : -1;
         }
         if (o2.length == 0)
         {
-            return -1;
+            return 1;
         }
 
         long L1 = ByteBuffer.wrap(o1).order(ByteOrder.LITTLE_ENDIAN).getLong();

src/java/org/apache/cassandra/db/marshal/UTF8Type.java

@@ -6,16 +6,6 @@ public class UTF8Type extends AbstractType
 {
     public int compare(byte[] o1, byte[] o2)
     {
-        // TODO rm hack to support the "i'm going to pretend [] is an index entry if I didn't actually index anything" hack
-        if (o1.length == 0)
-        {
-            return o2.length == 0 ? 0 : -1;
-        }
-        if (o2.length == 0)
-        {
-            return -1;
-        }
-
         try
         {
             return new String(o1, "UTF-8").compareTo(new String(o2, "UTF-8"));

src/java/org/apache/cassandra/db/marshal/UUIDType.java

@@ -13,14 +13,13 @@ public class UUIDType extends AbstractType
 
     public int compare(byte[] o1, byte[] o2)
     {
-        // TODO rm hack to support the "i'm going to pretend [] is an index entry if I didn't actually index anything" hack
         if (o1.length == 0)
         {
             return o2.length == 0 ? 0 : -1;
         }
         if (o2.length == 0)
         {
-            return -1;
+            return 1;
         }
 
         return getUUID(o1).compareTo(getUUID(o2));

src/java/org/apache/cassandra/io/SequenceFile.java

@@ -258,7 +258,6 @@ public class SequenceFile
             if (columnIndexList.size() == 0)
             {
                 /* if there is no column index, add an index entry that covers the full space. */
-                // TODO can we remove this?  it causes a lot of ugliness in everything that touches Marshal
                 return Arrays.asList(new IndexHelper.ColumnIndexInfo(ArrayUtils.EMPTY_BYTE_ARRAY, 0, totalNumCols, comparator_));
             }
 

test/unit/org/apache/cassandra/db/marshal/AsciiTypeTest.java

@@ -4,8 +4,6 @@ import org.apache.commons.lang.ArrayUtils;
 
 import org.junit.Test;
 
-import junit.framework.TestCase;
-
 public class AsciiTypeTest
 {
     @Test
@@ -13,7 +11,10 @@ public class AsciiTypeTest
     {
         AsciiType comparator = new AsciiType();
         assert comparator.compare(ArrayUtils.EMPTY_BYTE_ARRAY, "asdf".getBytes()) < 0;
+        assert comparator.compare("asdf".getBytes(), ArrayUtils.EMPTY_BYTE_ARRAY) > 0;
+        assert comparator.compare(ArrayUtils.EMPTY_BYTE_ARRAY, ArrayUtils.EMPTY_BYTE_ARRAY) == 0;
         assert comparator.compare("z".getBytes(), "a".getBytes()) > 0;
+        assert comparator.compare("a".getBytes(), "z".getBytes()) < 0;
         assert comparator.compare("asdf".getBytes(), "asdf".getBytes()) == 0;
         assert comparator.compare("asdz".getBytes(), "asdf".getBytes()) > 0;
     }

test/unit/org/apache/cassandra/db/marshal/UTF8TypeTest.java

@@ -0,0 +1,22 @@
+package org.apache.cassandra.db.marshal;
+
+import java.io.UnsupportedEncodingException;
+
+import org.apache.commons.lang.ArrayUtils;
+
+import org.junit.Test;
+
+public class UTF8TypeTest
+{
+    @Test
+    public void testCompare() throws UnsupportedEncodingException
+    {
+        UTF8Type comparator = new UTF8Type();
+        assert comparator.compare(ArrayUtils.EMPTY_BYTE_ARRAY, "asdf".getBytes()) < 0;
+        assert comparator.compare("asdf".getBytes(), ArrayUtils.EMPTY_BYTE_ARRAY) > 0;
+        assert comparator.compare(ArrayUtils.EMPTY_BYTE_ARRAY, ArrayUtils.EMPTY_BYTE_ARRAY) == 0;
+        assert comparator.compare("z".getBytes("UTF-8"), "a".getBytes("UTF-8")) > 0;
+        assert comparator.compare("z".getBytes("UTF-8"), "z".getBytes("UTF-8")) == 0;
+        assert comparator.compare("a".getBytes("UTF-8"), "z".getBytes("UTF-8")) < 0;
+    }
+}