HBASE-22833 MultiRowRangeFilter should provide a method for creating… (#493)

* HBASE-22833: MultiRowRangeFilter should provide a method for creating a filter which is functionally equivalent to multiple prefix filters

* Delete superfluous comments

* Add description for MultiRowRangeFilter constructor

* Add null check for rowKeyPrefixes

* Fix checkstyle

Signed-off-by: huzheng <openinx@gmail.com>

Author: titsuki

hbase-client/src/main/java/org/apache/hadoop/hbase/client/ClientUtil.java

@@ -17,8 +17,10 @@
  */
 package org.apache.hadoop.hbase.client;
 
-import org.apache.yetus.audience.InterfaceAudience;
+import java.util.Arrays;
+import org.apache.hadoop.hbase.HConstants;
 import org.apache.hadoop.hbase.util.Bytes;
+import org.apache.yetus.audience.InterfaceAudience;
 
 @InterfaceAudience.Private
 public class ClientUtil {
@@ -31,4 +33,46 @@ public static boolean areScanStartRowAndStopRowEqual(byte[] startRow, byte[] sto
   public static Cursor createCursor(byte[] row) {
     return new Cursor(row);
   }
+
+  /**
+   * <p>When scanning for a prefix the scan should stop immediately after the the last row that
+   * has the specified prefix. This method calculates the closest next rowKey immediately following
+   * the given rowKeyPrefix.</p>
+   * <p><b>IMPORTANT: This converts a rowKey<u>Prefix</u> into a rowKey</b>.</p>
+   * <p>If the prefix is an 'ASCII' string put into a byte[] then this is easy because you can
+   * simply increment the last byte of the array.
+   * But if your application uses real binary rowids you may run into the scenario that your
+   * prefix is something like:</p>
+   * &nbsp;&nbsp;&nbsp;<b>{ 0x12, 0x23, 0xFF, 0xFF }</b><br/>
+   * Then this stopRow needs to be fed into the actual scan<br/>
+   * &nbsp;&nbsp;&nbsp;<b>{ 0x12, 0x24 }</b> (Notice that it is shorter now)<br/>
+   * This method calculates the correct stop row value for this usecase.
+   *
+   * @param rowKeyPrefix the rowKey<u>Prefix</u>.
+   * @return the closest next rowKey immediately following the given rowKeyPrefix.
+   */
+  public static byte[] calculateTheClosestNextRowKeyForPrefix(byte[] rowKeyPrefix) {
+    // Essentially we are treating it like an 'unsigned very very long' and doing +1 manually.
+    // Search for the place where the trailing 0xFFs start
+    int offset = rowKeyPrefix.length;
+    while (offset > 0) {
+      if (rowKeyPrefix[offset - 1] != (byte) 0xFF) {
+        break;
+      }
+      offset--;
+    }
+
+    if (offset == 0) {
+      // We got an 0xFFFF... (only FFs) stopRow value which is
+      // the last possible prefix before the end of the table.
+      // So set it to stop at the 'end of the table'
+      return HConstants.EMPTY_END_ROW;
+    }
+
+    // Copy the right length of the original
+    byte[] newStopRow = Arrays.copyOfRange(rowKeyPrefix, 0, offset);
+    // And increment the last one
+    newStopRow[newStopRow.length - 1]++;
+    return newStopRow;
+  }
 }

hbase-client/src/main/java/org/apache/hadoop/hbase/client/Scan.java

@@ -21,7 +21,6 @@
 
 import java.io.IOException;
 import java.util.ArrayList;
-import java.util.Arrays;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
@@ -563,53 +562,11 @@ public Scan setRowPrefixFilter(byte[] rowPrefix) {
       setStopRow(HConstants.EMPTY_END_ROW);
     } else {
       this.setStartRow(rowPrefix);
-      this.setStopRow(calculateTheClosestNextRowKeyForPrefix(rowPrefix));
+      this.setStopRow(ClientUtil.calculateTheClosestNextRowKeyForPrefix(rowPrefix));
     }
     return this;
   }
 
-  /**
-   * <p>When scanning for a prefix the scan should stop immediately after the the last row that
-   * has the specified prefix. This method calculates the closest next rowKey immediately following
-   * the given rowKeyPrefix.</p>
-   * <p><b>IMPORTANT: This converts a rowKey<u>Prefix</u> into a rowKey</b>.</p>
-   * <p>If the prefix is an 'ASCII' string put into a byte[] then this is easy because you can
-   * simply increment the last byte of the array.
-   * But if your application uses real binary rowids you may run into the scenario that your
-   * prefix is something like:</p>
-   * &nbsp;&nbsp;&nbsp;<b>{ 0x12, 0x23, 0xFF, 0xFF }</b><br/>
-   * Then this stopRow needs to be fed into the actual scan<br/>
-   * &nbsp;&nbsp;&nbsp;<b>{ 0x12, 0x24 }</b> (Notice that it is shorter now)<br/>
-   * This method calculates the correct stop row value for this usecase.
-   *
-   * @param rowKeyPrefix the rowKey<u>Prefix</u>.
-   * @return the closest next rowKey immediately following the given rowKeyPrefix.
-   */
-  private byte[] calculateTheClosestNextRowKeyForPrefix(byte[] rowKeyPrefix) {
-    // Essentially we are treating it like an 'unsigned very very long' and doing +1 manually.
-    // Search for the place where the trailing 0xFFs start
-    int offset = rowKeyPrefix.length;
-    while (offset > 0) {
-      if (rowKeyPrefix[offset - 1] != (byte) 0xFF) {
-        break;
-      }
-      offset--;
-    }
-
-    if (offset == 0) {
-      // We got an 0xFFFF... (only FFs) stopRow value which is
-      // the last possible prefix before the end of the table.
-      // So set it to stop at the 'end of the table'
-      return HConstants.EMPTY_END_ROW;
-    }
-
-    // Copy the right length of the original
-    byte[] newStopRow = Arrays.copyOfRange(rowKeyPrefix, 0, offset);
-    // And increment the last one
-    newStopRow[newStopRow.length - 1]++;
-    return newStopRow;
-  }
-
   /**
    * Get all available versions.
    * @return this

hbase-client/src/main/java/org/apache/hadoop/hbase/filter/MultiRowRangeFilter.java

@@ -26,12 +26,13 @@
 import org.apache.hadoop.hbase.CellUtil;
 import org.apache.hadoop.hbase.HConstants;
 import org.apache.hadoop.hbase.PrivateCellUtil;
-import org.apache.yetus.audience.InterfaceAudience;
+import org.apache.hadoop.hbase.client.ClientUtil;
 import org.apache.hadoop.hbase.exceptions.DeserializationException;
+import org.apache.hadoop.hbase.util.Bytes;
+import org.apache.yetus.audience.InterfaceAudience;
 import org.apache.hbase.thirdparty.com.google.protobuf.InvalidProtocolBufferException;
 import org.apache.hbase.thirdparty.com.google.protobuf.UnsafeByteOperations;
 import org.apache.hadoop.hbase.shaded.protobuf.generated.FilterProtos;
-import org.apache.hadoop.hbase.util.Bytes;
 
 /**
  * Filter to support scan multiple row key ranges. It can construct the row key ranges from the
@@ -71,6 +72,33 @@ public MultiRowRangeFilter(List<RowRange> list) {
     this.ranges = new RangeIteration(rangeList);
   }
 
+  /**
+   * Constructor for creating a <code>MultiRowRangeFilter</code> from multiple rowkey prefixes.
+   *
+   * As <code>MultiRowRangeFilter</code> javadoc says (See the solution 1 of the first statement),
+   * if you try to create a filter list that scans row keys corresponding to given prefixes (e.g.,
+   * <code>FilterList</code> composed of multiple <code>PrefixFilter</code>s), this constructor
+   * provides a way to avoid creating an inefficient one.
+   *
+   * @param rowKeyPrefixes the array of byte array
+   */
+  public MultiRowRangeFilter(byte[][] rowKeyPrefixes) {
+    this(createRangeListFromRowKeyPrefixes(rowKeyPrefixes));
+  }
+
+  private static List<RowRange> createRangeListFromRowKeyPrefixes(byte[][] rowKeyPrefixes) {
+    if (rowKeyPrefixes == null) {
+      throw new IllegalArgumentException("Invalid rowkey prefixes");
+    }
+
+    List<RowRange> list = new ArrayList<>();
+    for (byte[] rowKeyPrefix: rowKeyPrefixes) {
+      byte[] stopRow = ClientUtil.calculateTheClosestNextRowKeyForPrefix(rowKeyPrefix);
+      list.add(new RowRange(rowKeyPrefix, true, stopRow, false));
+    }
+    return list;
+  }
+
   public List<RowRange> getRowRanges() {
     // Used by hbase-rest
     return this.rangeList;

hbase-server/src/test/java/org/apache/hadoop/hbase/filter/TestMultiRowRangeFilter.java

@@ -82,6 +82,61 @@ public static void tearDownAfterClass() throws Exception {
     TEST_UTIL.shutdownMiniCluster();
   }
 
+  @Test
+  public void testRowKeyPrefixWithEmptyPrefix() throws IOException {
+    byte[] prefix = {};
+    byte[][] rowKeyPrefixes = new byte[1][];
+    rowKeyPrefixes[0] = prefix;
+    MultiRowRangeFilter filter = new MultiRowRangeFilter(rowKeyPrefixes);
+    List<RowRange> actualRanges = filter.getRowRanges();
+    List<RowRange> expectedRanges = new ArrayList<>();
+    expectedRanges.add(
+      new RowRange(HConstants.EMPTY_START_ROW, true, HConstants.EMPTY_END_ROW, false)
+    );
+    assertRangesEqual(expectedRanges, actualRanges);
+  }
+
+  @Test
+  public void testRowKeyPrefixWithLastIncrementablePrefix() throws IOException {
+    byte[] prefix = {(byte) 0x12, (byte) 0x23, (byte) 0xFF, (byte) 0xFE};
+    byte[][] rowKeyPrefixes = new byte[1][];
+    rowKeyPrefixes[0] = prefix;
+    MultiRowRangeFilter filter = new MultiRowRangeFilter(rowKeyPrefixes);
+    List<RowRange> actualRanges = filter.getRowRanges();
+    List<RowRange> expectedRanges = new ArrayList<>();
+    final byte[] expectedStop = {(byte) 0x12, (byte) 0x23, (byte) 0xFF, (byte) 0xFF};
+    expectedRanges.add(new RowRange(prefix, true, expectedStop , false));
+    assertRangesEqual(expectedRanges, actualRanges);
+  }
+
+  @Test
+  public void testRowKeyPrefixWithoutLastIncrementablePrefix() throws IOException {
+    byte[] prefix = {(byte) 0x12, (byte) 0x23, (byte) 0xFF, (byte) 0xFF};
+    byte[][] rowKeyPrefixes = new byte[1][];
+    rowKeyPrefixes[0] = prefix;
+    MultiRowRangeFilter filter = new MultiRowRangeFilter(rowKeyPrefixes);
+    List<RowRange> actualRanges = filter.getRowRanges();
+    List<RowRange> expectedRanges = new ArrayList<>();
+    final byte[] expectedStop = {(byte) 0x12, (byte) 0x24};
+    expectedRanges.add(new RowRange(prefix, true, expectedStop , false));
+    assertRangesEqual(expectedRanges, actualRanges);
+  }
+
+  @Test
+  public void testRowKeyPrefixWithMergablePrefix() throws IOException {
+    byte[] prefix1 = {(byte) 0x12, (byte) 0x23, (byte) 0xFF, (byte) 0xFE};
+    byte[] prefix2 = {(byte) 0x12, (byte) 0x23, (byte) 0xFF, (byte) 0xFF};
+    byte[][] rowKeyPrefixes = new byte[2][];
+    rowKeyPrefixes[0] = prefix1;
+    rowKeyPrefixes[1] = prefix2;
+    MultiRowRangeFilter filter = new MultiRowRangeFilter(rowKeyPrefixes);
+    List<RowRange> actualRanges = filter.getRowRanges();
+    List<RowRange> expectedRanges = new ArrayList<>();
+    final byte[] expectedStop = {(byte) 0x12, (byte) 0x24};
+    expectedRanges.add(new RowRange(prefix1, true, expectedStop , false));
+    assertRangesEqual(expectedRanges, actualRanges);
+  }
+
   @Test
   public void testRanges() throws IOException {
     byte[] key1Start = new byte[] {-3};