/*
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.apache.hadoop.hbase.mapreduce;

import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION;
import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION_DEFAULT;
import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_DEFAULT;
import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY;
import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_ROW_LIMIT_PER_INPUTSPLIT;
import static org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormatImpl.SNAPSHOT_INPUTFORMAT_SCANNER_READTYPE;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertFalse;
import static org.junit.jupiter.api.Assertions.assertTrue;
import static org.mockito.Mockito.mock;
import static org.mockito.Mockito.when;

import java.io.IOException;
import java.util.Arrays;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseTestingUtil;
import org.apache.hadoop.hbase.HConstants;
import org.apache.hadoop.hbase.HDFSBlocksDistribution;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.client.Admin;
import org.apache.hadoop.hbase.client.Result;
import org.apache.hadoop.hbase.client.Scan;
import org.apache.hadoop.hbase.client.Scan.ReadType;
import org.apache.hadoop.hbase.client.Table;
import org.apache.hadoop.hbase.client.TestTableSnapshotScanner;
import org.apache.hadoop.hbase.io.HFileLink;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormat.TableSnapshotRegionRecordReader;
import org.apache.hadoop.hbase.mapreduce.TableSnapshotInputFormat.TableSnapshotRegionSplit;
import org.apache.hadoop.hbase.snapshot.RestoreSnapshotHelper;
import org.apache.hadoop.hbase.snapshot.SnapshotTestingUtils;
import org.apache.hadoop.hbase.testclassification.LargeTests;
import org.apache.hadoop.hbase.testclassification.VerySlowMapReduceTests;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.hbase.util.CommonFSUtils;
import org.apache.hadoop.hbase.util.RegionSplitter;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.InputSplit;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.RecordReader;
import org.apache.hadoop.mapreduce.Reducer;
import org.apache.hadoop.mapreduce.TaskAttemptContext;
import org.apache.hadoop.mapreduce.lib.output.NullOutputFormat;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Tag;
import org.junit.jupiter.api.Test;
import org.junit.jupiter.api.TestInfo;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import org.apache.hbase.thirdparty.com.google.common.collect.Lists;

@Tag(VerySlowMapReduceTests.TAG)
@Tag(LargeTests.TAG)
public class TestTableSnapshotInputFormat extends TableSnapshotInputFormatTestBase {

  private static final Logger LOG = LoggerFactory.getLogger(TestTableSnapshotInputFormat.class);

  private static final byte[] bbb = Bytes.toBytes("bbb");
  private static final byte[] yyy = Bytes.toBytes("yyy");
  private static final byte[] bbc = Bytes.toBytes("bbc");
  private static final byte[] yya = Bytes.toBytes("yya");

  private String name;

  @BeforeEach
  public void setUp(TestInfo testInfo) throws Exception {
    name = testInfo.getTestMethod().get().getName();
  }

  @Override
  protected byte[] getStartRow() {
    return bbb;
  }

  @Override
  protected byte[] getEndRow() {
    return yyy;
  }

  @Test
  public void testGetBestLocations() throws IOException {
    TableSnapshotInputFormatImpl tsif = new TableSnapshotInputFormatImpl();
    Configuration conf = UTIL.getConfiguration();

    HDFSBlocksDistribution blockDistribution = new HDFSBlocksDistribution();
    assertEquals(null, TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h1" }, 1);
    assertEquals(Lists.newArrayList("h1"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h1" }, 1);
    assertEquals(Lists.newArrayList("h1"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h2" }, 1);
    assertEquals(Lists.newArrayList("h1"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution = new HDFSBlocksDistribution();
    blockDistribution.addHostsAndBlockWeight(new String[] { "h1" }, 10);
    blockDistribution.addHostsAndBlockWeight(new String[] { "h2" }, 7);
    blockDistribution.addHostsAndBlockWeight(new String[] { "h3" }, 5);
    blockDistribution.addHostsAndBlockWeight(new String[] { "h4" }, 1);
    assertEquals(Lists.newArrayList("h1"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h2" }, 2);
    assertEquals(Lists.newArrayList("h1", "h2"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h2" }, 3);
    assertEquals(Lists.newArrayList("h2", "h1"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));

    blockDistribution.addHostsAndBlockWeight(new String[] { "h3" }, 6);
    blockDistribution.addHostsAndBlockWeight(new String[] { "h4" }, 9);

    assertEquals(Lists.newArrayList("h2", "h3", "h4"),
      TableSnapshotInputFormatImpl.getBestLocations(conf, blockDistribution));
  }

  @Test
  public void testTableSnapshotRegionRecordReaderGetFilesRead() throws Exception {
    final TableName tableName = TableName.valueOf(name);
    String snapshotName = name + "_snapshot";
    try {
      // Setup: create table, load data, snapshot, and configure job with restore dir
      createTableAndSnapshot(UTIL, tableName, snapshotName, getStartRow(), getEndRow(), 1);

      Configuration conf = UTIL.getConfiguration();
      Job job = new Job(conf);
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);
      Scan scan = new Scan().withStartRow(getStartRow()).withStopRow(getEndRow());
      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
        TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
        tmpTableDir);

      // Get splits (one per region) and extract delegate split for restore path and region info
      TableSnapshotInputFormat tsif = new TableSnapshotInputFormat();
      List<InputSplit> splits = tsif.getSplits(job);
      assertEquals(1, splits.size());

      InputSplit split = splits.get(0);
      assertTrue(split instanceof TableSnapshotRegionSplit);
      TableSnapshotRegionSplit snapshotRegionSplit = (TableSnapshotRegionSplit) split;
      TableSnapshotInputFormatImpl.InputSplit implSplit = snapshotRegionSplit.getDelegate();

      // Collect expected store file paths from the restored region directory
      Set<String> expectedFiles = new HashSet<>();
      Path restorePath = new Path(implSplit.getRestoreDir());
      FileSystem fs = restorePath.getFileSystem(conf);
      Path tableDir =
        CommonFSUtils.getTableDir(restorePath, implSplit.getTableDescriptor().getTableName());
      Path regionPath = new Path(tableDir, implSplit.getRegionInfo().getEncodedName());
      FileStatus[] familyDirs = fs.listStatus(regionPath);
      if (familyDirs != null) {
        for (FileStatus fam : familyDirs) {
          if (fam.isDirectory()) {
            FileStatus[] files = fs.listStatus(fam.getPath());
            if (files != null) {
              for (FileStatus f : files) {
                if (f.isFile()) {
                  String referenceFileName = f.getPath().getName();
                  expectedFiles.add(HFileLink.getReferencedHFileName(referenceFileName));
                }
              }
            }
          }
        }
      }
      assertFalse(expectedFiles.isEmpty(),
        "Should have at least one store file after snapshot restore");

      // Create record reader, initialize with split (opens underlying ClientSideRegionScanner)
      TaskAttemptContext taskAttemptContext = mock(TaskAttemptContext.class);
      when(taskAttemptContext.getConfiguration()).thenReturn(job.getConfiguration());

      RecordReader<ImmutableBytesWritable, Result> rr =
        tsif.createRecordReader(split, taskAttemptContext);
      assertTrue(rr instanceof TableSnapshotRegionRecordReader);
      TableSnapshotRegionRecordReader recordReader = (TableSnapshotRegionRecordReader) rr;
      recordReader.initialize(split, taskAttemptContext);

      // Before close: getFilesRead() must be empty
      Set<Path> filesReadBeforeClose = recordReader.getFilesRead();
      assertTrue(filesReadBeforeClose.isEmpty(), "Should return empty set before closing");

      // Read a few key-values; getFilesRead() must still be empty until close
      int count = 0;
      while (count < 3 && recordReader.nextKeyValue()) {
        count++;
      }

      filesReadBeforeClose = recordReader.getFilesRead();
      assertTrue(filesReadBeforeClose.isEmpty(),
        "Should return empty set before closing even after reading");

      // Close reader so underlying scanner reports files successfully read
      recordReader.close();

      // After close: getFilesRead() must match expected store file set
      Set<String> filesReadAfterClose =
        recordReader.getFilesRead().stream().map(Path::getName).collect(Collectors.toSet());

      assertEquals(expectedFiles, filesReadAfterClose, "Should contain all expected file paths");
    } finally {
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  public static enum TestTableSnapshotCounters {
    VALIDATION_ERROR
  }

  public static class TestTableSnapshotMapper
    extends TableMapper<ImmutableBytesWritable, NullWritable> {
    @Override
    protected void map(ImmutableBytesWritable key, Result value, Context context)
      throws IOException, InterruptedException {
      // Validate a single row coming from the snapshot, and emit the row key
      verifyRowFromMap(key, value);
      context.write(key, NullWritable.get());
    }
  }

  public static class TestTableSnapshotReducer
    extends Reducer<ImmutableBytesWritable, NullWritable, NullWritable, NullWritable> {
    HBaseTestingUtil.SeenRowTracker rowTracker = new HBaseTestingUtil.SeenRowTracker(bbb, yyy);

    @Override
    protected void reduce(ImmutableBytesWritable key, Iterable<NullWritable> values,
      Context context) throws IOException, InterruptedException {
      rowTracker.addRow(key.get());
    }

    @Override
    protected void cleanup(Context context) throws IOException, InterruptedException {
      rowTracker.validate();
    }
  }

  @Test
  public void testInitTableSnapshotMapperJobConfig() throws Exception {
    final TableName tableName = TableName.valueOf(name);
    String snapshotName = "foo";

    try {
      createTableAndSnapshot(UTIL, tableName, snapshotName, getStartRow(), getEndRow(), 1);
      Job job = new Job(UTIL.getConfiguration());
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);

      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, new Scan(),
        TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
        tmpTableDir);

      // TODO: would be better to examine directly the cache instance that results from this
      // config. Currently this is not possible because BlockCache initialization is static.
      assertEquals(HConstants.HFILE_BLOCK_CACHE_SIZE_DEFAULT,
        job.getConfiguration().getFloat(HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, -1), 0.01,
        "Snapshot job should be configured for default LruBlockCache.");
      assertEquals(0, job.getConfiguration().getFloat("hbase.bucketcache.size", -1), 0.01,
        "Snapshot job should not use BucketCache.");
    } finally {
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  @Test
  public void testWithMockedMapReduceSingleRegionByRegionLocation() throws Exception {
    Configuration conf = UTIL.getConfiguration();
    conf.setBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION, true);
    try {
      testWithMockedMapReduce(UTIL, name + "Snapshot", 1, 1, 1, true);
    } finally {
      conf.unset(SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION);
    }
  }

  @Override
  public void testRestoreSnapshotDoesNotCreateBackRefLinksInit(TableName tableName,
    String snapshotName, Path tmpTableDir) throws Exception {
    Job job = new Job(UTIL.getConfiguration());
    TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, new Scan(),
      TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
      tmpTableDir);
  }

  @Override
  public void testWithMockedMapReduce(HBaseTestingUtil util, String snapshotName, int numRegions,
    int numSplitsPerRegion, int expectedNumSplits, boolean setLocalityEnabledTo) throws Exception {
    final TableName tableName = TableName.valueOf(name);
    try {
      createTableAndSnapshot(util, tableName, snapshotName, getStartRow(), getEndRow(), numRegions);

      Configuration conf = util.getConfiguration();
      conf.setBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY, setLocalityEnabledTo);
      conf.setBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION,
        SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION_DEFAULT);
      Job job = new Job(conf);
      Path tmpTableDir = util.getDataTestDirOnTestFS(snapshotName);
      Scan scan = new Scan().withStartRow(getStartRow()).withStopRow(getEndRow()); // limit the scan

      if (numSplitsPerRegion > 1) {
        TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
          TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job,
          false, tmpTableDir, new RegionSplitter.UniformSplit(), numSplitsPerRegion);
      } else {
        TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
          TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job,
          false, tmpTableDir);
      }

      verifyWithMockedMapReduce(job, numRegions, expectedNumSplits, getStartRow(), getEndRow());

    } finally {
      util.getAdmin().deleteSnapshot(snapshotName);
      util.deleteTable(tableName);
    }
  }

  @Test
  public void testWithMockedMapReduceWithSplitsPerRegion() throws Exception {
    String snapshotName = "testWithMockedMapReduceMultiRegion";
    final TableName tableName = TableName.valueOf(name);
    try {
      createTableAndSnapshot(UTIL, tableName, snapshotName, getStartRow(), getEndRow(), 10);

      Configuration conf = UTIL.getConfiguration();
      conf.setBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY, false);
      Job job = new Job(conf);
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);
      // test scan with startRow and stopRow
      Scan scan = new Scan().withStartRow(bbc).withStopRow(yya);

      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
        TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
        tmpTableDir, new RegionSplitter.UniformSplit(), 5);

      verifyWithMockedMapReduce(job, 10, 40, bbc, yya);
    } finally {
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  @Test
  public void testWithMockedMapReduceWithNoStartRowStopRow() throws Exception {
    String snapshotName = "testWithMockedMapReduceMultiRegion";
    final TableName tableName = TableName.valueOf(name);
    try {
      createTableAndSnapshot(UTIL, tableName, snapshotName, getStartRow(), getEndRow(), 10);

      Configuration conf = UTIL.getConfiguration();
      conf.setBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY, false);
      Job job = new Job(conf);
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);
      // test scan without startRow and stopRow
      Scan scan2 = new Scan();

      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan2,
        TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
        tmpTableDir, new RegionSplitter.UniformSplit(), 5);

      verifyWithMockedMapReduce(job, 10, 50, HConstants.EMPTY_START_ROW,
        HConstants.EMPTY_START_ROW);

    } finally {
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  @Test
  public void testScanLimit() throws Exception {
    final TableName tableName = TableName.valueOf(name);
    final String snapshotName = tableName + "Snapshot";
    Table table = null;
    try {
      UTIL.getConfiguration().setInt(SNAPSHOT_INPUTFORMAT_ROW_LIMIT_PER_INPUTSPLIT, 10);
      if (UTIL.getAdmin().tableExists(tableName)) {
        UTIL.deleteTable(tableName);
      }

      UTIL.createTable(tableName, FAMILIES, new byte[][] { bbb, yyy });

      Admin admin = UTIL.getAdmin();

      int regionNum = admin.getRegions(tableName).size();
      // put some stuff in the table
      table = UTIL.getConnection().getTable(tableName);
      UTIL.loadTable(table, FAMILIES);

      Path rootDir = CommonFSUtils.getRootDir(UTIL.getConfiguration());
      FileSystem fs = rootDir.getFileSystem(UTIL.getConfiguration());

      SnapshotTestingUtils.createSnapshotAndValidate(admin, tableName, Arrays.asList(FAMILIES),
        null, snapshotName, rootDir, fs, true);

      Job job = new Job(UTIL.getConfiguration());
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);
      Scan scan = new Scan();
      TableMapReduceUtil.addDependencyJarsForClasses(job.getConfiguration(),
        TestTableSnapshotInputFormat.class);

      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
        RowCounter.RowCounterMapper.class, NullWritable.class, NullWritable.class, job, true,
        tmpTableDir);
      assertTrue(job.waitForCompletion(true));
      assertEquals(10 * regionNum,
        job.getCounters().findCounter(RowCounter.RowCounterMapper.Counters.ROWS).getValue());
    } finally {
      if (table != null) {
        table.close();
      }
      UTIL.getConfiguration().unset(SNAPSHOT_INPUTFORMAT_ROW_LIMIT_PER_INPUTSPLIT);
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  @Test
  public void testNoDuplicateResultsWhenSplitting() throws Exception {
    TableName tableName = TableName.valueOf("testNoDuplicateResultsWhenSplitting");
    String snapshotName = "testSnapshotBug";
    try {
      if (UTIL.getAdmin().tableExists(tableName)) {
        UTIL.deleteTable(tableName);
      }

      UTIL.createTable(tableName, FAMILIES);
      Admin admin = UTIL.getAdmin();

      // put some stuff in the table
      Table table = UTIL.getConnection().getTable(tableName);
      UTIL.loadTable(table, FAMILIES);

      // split to 2 regions
      admin.split(tableName, Bytes.toBytes("eee"));
      TestTableSnapshotScanner.blockUntilSplitFinished(UTIL, tableName, 2);

      Path rootDir = CommonFSUtils.getRootDir(UTIL.getConfiguration());
      FileSystem fs = rootDir.getFileSystem(UTIL.getConfiguration());

      SnapshotTestingUtils.createSnapshotAndValidate(admin, tableName, Arrays.asList(FAMILIES),
        null, snapshotName, rootDir, fs, true);

      // load different values
      byte[] value = Bytes.toBytes("after_snapshot_value");
      UTIL.loadTable(table, FAMILIES, value);

      // cause flush to create new files in the region
      admin.flush(tableName);
      table.close();

      Job job = new Job(UTIL.getConfiguration());
      Path tmpTableDir = UTIL.getDataTestDirOnTestFS(snapshotName);
      // limit the scan
      Scan scan = new Scan().withStartRow(getStartRow()).withStopRow(getEndRow());

      TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
        TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
        tmpTableDir);

      verifyWithMockedMapReduce(job, 2, 2, getStartRow(), getEndRow());
    } finally {
      UTIL.getAdmin().deleteSnapshot(snapshotName);
      UTIL.deleteTable(tableName);
    }
  }

  @Test
  public void testScannerReadTypeConfiguration() throws IOException {
    Configuration conf = new Configuration(false);
    // Explicitly set ReadTypes should persist
    for (ReadType readType : Arrays.asList(ReadType.PREAD, ReadType.STREAM)) {
      Scan scanWithReadType = new Scan();
      scanWithReadType.setReadType(readType);
      assertEquals(scanWithReadType.getReadType(),
        serializeAndReturn(conf, scanWithReadType).getReadType());
    }
    // We should only see the DEFAULT ReadType getting updated to STREAM.
    Scan scanWithoutReadType = new Scan();
    assertEquals(ReadType.DEFAULT, scanWithoutReadType.getReadType());
    assertEquals(ReadType.STREAM, serializeAndReturn(conf, scanWithoutReadType).getReadType());

    // We should still be able to force a certain ReadType when DEFAULT is given.
    conf.setEnum(SNAPSHOT_INPUTFORMAT_SCANNER_READTYPE, ReadType.PREAD);
    assertEquals(ReadType.DEFAULT, scanWithoutReadType.getReadType());
    assertEquals(ReadType.PREAD, serializeAndReturn(conf, scanWithoutReadType).getReadType());
  }

  /**
   * Serializes and deserializes the given scan in the same manner that TableSnapshotInputFormat
   * does.
   */
  private Scan serializeAndReturn(Configuration conf, Scan s) throws IOException {
    conf.set(TableInputFormat.SCAN, TableMapReduceUtil.convertScanToString(s));
    return TableSnapshotInputFormatImpl.extractScanFromConf(conf);
  }

  private void verifyWithMockedMapReduce(Job job, int numRegions, int expectedNumSplits,
    byte[] startRow, byte[] stopRow) throws IOException, InterruptedException {
    TableSnapshotInputFormat tsif = new TableSnapshotInputFormat();
    List<InputSplit> splits = tsif.getSplits(job);

    assertEquals(expectedNumSplits, splits.size());

    HBaseTestingUtil.SeenRowTracker rowTracker = new HBaseTestingUtil.SeenRowTracker(startRow,
      stopRow.length > 0 ? stopRow : Bytes.toBytes("\uffff"));

    boolean localityEnabled = job.getConfiguration().getBoolean(
      SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY, SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_DEFAULT);

    boolean byRegionLoc =
      job.getConfiguration().getBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION,
        SNAPSHOT_INPUTFORMAT_LOCALITY_BY_REGION_LOCATION_DEFAULT);
    for (int i = 0; i < splits.size(); i++) {
      // validate input split
      InputSplit split = splits.get(i);
      assertTrue(split instanceof TableSnapshotRegionSplit);
      TableSnapshotRegionSplit snapshotRegionSplit = (TableSnapshotRegionSplit) split;
      if (localityEnabled) {
        assertTrue(split.getLocations() != null && split.getLocations().length != 0);
        if (byRegionLoc) {
          // When it uses region location from meta, the hostname will be "localhost",
          // the location from hdfs block location is "127.0.0.1".
          assertEquals(1, split.getLocations().length);
          assertTrue(split.getLocations()[0].equals("localhost"), "Not using region location!");
        } else {
          assertTrue(split.getLocations()[0].equals("127.0.0.1"), "Not using region location!");
        }
      } else {
        assertTrue(split.getLocations() != null && split.getLocations().length == 0);
      }

      Scan scan =
        TableMapReduceUtil.convertStringToScan(snapshotRegionSplit.getDelegate().getScan());
      if (startRow.length > 0) {
        assertTrue(Bytes.compareTo(startRow, scan.getStartRow()) <= 0,
          Bytes.toStringBinary(startRow) + " should <= "
            + Bytes.toStringBinary(scan.getStartRow()));
      }
      if (stopRow.length > 0) {
        assertTrue(Bytes.compareTo(stopRow, scan.getStopRow()) >= 0,
          Bytes.toStringBinary(stopRow) + " should >= " + Bytes.toStringBinary(scan.getStopRow()));
      }
      assertTrue(Bytes.compareTo(scan.getStartRow(), scan.getStopRow()) < 0,
        "startRow should < stopRow");

      // validate record reader
      TaskAttemptContext taskAttemptContext = mock(TaskAttemptContext.class);
      when(taskAttemptContext.getConfiguration()).thenReturn(job.getConfiguration());
      RecordReader<ImmutableBytesWritable, Result> rr =
        tsif.createRecordReader(split, taskAttemptContext);
      rr.initialize(split, taskAttemptContext);

      // validate we can read all the data back
      while (rr.nextKeyValue()) {
        byte[] row = rr.getCurrentKey().get();
        verifyRowFromMap(rr.getCurrentKey(), rr.getCurrentValue());
        rowTracker.addRow(row);
      }

      rr.close();
    }

    // validate all rows are seen
    rowTracker.validate();
  }

  @Override
  protected void testWithMapReduceImpl(HBaseTestingUtil util, TableName tableName,
    String snapshotName, Path tableDir, int numRegions, int numSplitsPerRegion,
    int expectedNumSplits, boolean shutdownCluster) throws Exception {
    doTestWithMapReduce(util, tableName, snapshotName, getStartRow(), getEndRow(), tableDir,
      numRegions, numSplitsPerRegion, expectedNumSplits, shutdownCluster);
  }

  // this is also called by the IntegrationTestTableSnapshotInputFormat
  public static void doTestWithMapReduce(HBaseTestingUtil util, TableName tableName,
    String snapshotName, byte[] startRow, byte[] endRow, Path tableDir, int numRegions,
    int numSplitsPerRegion, int expectedNumSplits, boolean shutdownCluster) throws Exception {

    LOG.info("testing with MapReduce");

    LOG.info("create the table and snapshot");
    createTableAndSnapshot(util, tableName, snapshotName, startRow, endRow, numRegions);

    if (shutdownCluster) {
      LOG.info("shutting down hbase cluster.");
      util.shutdownMiniHBaseCluster();
    }

    try {
      // create the job
      Job job = new Job(util.getConfiguration());
      Scan scan = new Scan().withStartRow(startRow).withStopRow(endRow); // limit the scan

      job.setJarByClass(util.getClass());
      TableMapReduceUtil.addDependencyJarsForClasses(job.getConfiguration(),
        TestTableSnapshotInputFormat.class);

      if (numSplitsPerRegion > 1) {
        TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
          TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job,
          true, tableDir, new RegionSplitter.UniformSplit(), numSplitsPerRegion);
      } else {
        TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, scan,
          TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job,
          true, tableDir);
      }

      job.setReducerClass(TestTableSnapshotInputFormat.TestTableSnapshotReducer.class);
      job.setNumReduceTasks(1);
      job.setOutputFormatClass(NullOutputFormat.class);

      assertTrue(job.waitForCompletion(true));
    } finally {
      if (!shutdownCluster) {
        util.getAdmin().deleteSnapshot(snapshotName);
        util.deleteTable(tableName);
      }
    }
  }

  @Test
  public void testWithMapReduceMultipleMappersPerRegion() throws Exception {
    testWithMapReduce(UTIL, "testWithMapReduceMultiRegion", 10, 5, 50, false);
  }

  @Test
  public void testCleanRestoreDir() throws Exception {
    TableName tableName = TableName.valueOf("test_table");
    String snapshotName = "test_snapshot";
    createTableAndSnapshot(UTIL, tableName, snapshotName, getStartRow(), getEndRow(), 1);
    Job job = Job.getInstance(UTIL.getConfiguration());
    Path workingDir = UTIL.getDataTestDirOnTestFS(snapshotName);
    TableMapReduceUtil.initTableSnapshotMapperJob(snapshotName, new Scan(),
      TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job, false,
      workingDir);
    FileSystem fs = workingDir.getFileSystem(job.getConfiguration());
    Path restorePath =
      new Path(job.getConfiguration().get("hbase.TableSnapshotInputFormat.restore.dir"));
    assertTrue(fs.exists(restorePath));
    TableSnapshotInputFormat.cleanRestoreDir(job, snapshotName);
    assertFalse(fs.exists(restorePath));
  }

  /**
   * Test that explicitly restores a snapshot to a temp directory and reads the restored regions via
   * ClientSideRegionScanner through a MapReduce job.
   * <p>
   * This test verifies the full workflow: 1. Create and load a table with data 2. Create a snapshot
   * and restore the snapshot to a temporary directory 3. Configure a job to read the restored
   * regions via ClientSideRegionScanner using TableSnapshotInputFormat and verify that it succeeds
   * 4. Delete restored temporary directory 5. Configure a new job and verify that it fails
   */
  @Test
  public void testReadFromRestoredSnapshotViaMR() throws Exception {
    final TableName tableName = TableName.valueOf(name);
    final String snapshotName = tableName + "_snapshot";
    try {
      if (UTIL.getAdmin().tableExists(tableName)) {
        UTIL.deleteTable(tableName);
      }
      UTIL.createTable(tableName, FAMILIES, new byte[][] { bbb, yyy });

      Admin admin = UTIL.getAdmin();
      int regionNum = admin.getRegions(tableName).size();
      LOG.info("Created table with {} regions", regionNum);

      Table table = UTIL.getConnection().getTable(tableName);
      UTIL.loadTable(table, FAMILIES);
      table.close();

      Path rootDir = CommonFSUtils.getRootDir(UTIL.getConfiguration());
      FileSystem fs = rootDir.getFileSystem(UTIL.getConfiguration());
      SnapshotTestingUtils.createSnapshotAndValidate(admin, tableName, Arrays.asList(FAMILIES),
        null, snapshotName, rootDir, fs, true);
      Path tempRestoreDir = UTIL.getDataTestDirOnTestFS("restore_" + snapshotName);
      RestoreSnapshotHelper.copySnapshotForScanner(UTIL.getConfiguration(), fs, rootDir,
        tempRestoreDir, snapshotName);
      assertTrue(fs.exists(tempRestoreDir), "Restore directory should exist");

      Job job = Job.getInstance(UTIL.getConfiguration());
      job.setJarByClass(TestTableSnapshotInputFormat.class);
      TableMapReduceUtil.addDependencyJarsForClasses(job.getConfiguration(),
        TestTableSnapshotInputFormat.class);
      Scan scan = new Scan().withStartRow(getStartRow()).withStopRow(getEndRow());
      Configuration conf = job.getConfiguration();
      conf.set("hbase.TableSnapshotInputFormat.snapshot.name", snapshotName);
      conf.set("hbase.TableSnapshotInputFormat.restore.dir", tempRestoreDir.toString());
      conf.setInt("hbase.mapreduce.splits.per.region", 1);
      job.setReducerClass(TestTableSnapshotReducer.class);
      job.setNumReduceTasks(1);
      job.setOutputFormatClass(NullOutputFormat.class);
      TableMapReduceUtil.initTableMapperJob(snapshotName, // table name (snapshot name in this case)
        scan, TestTableSnapshotMapper.class, ImmutableBytesWritable.class, NullWritable.class, job,
        false, false, TableSnapshotInputFormat.class);
      TableMapReduceUtil.resetCacheConfig(conf);
      assertTrue(job.waitForCompletion(true));
      assertTrue(job.isSuccessful());

      // Now verify that job fails when restore directory is deleted
      assertTrue(fs.delete(tempRestoreDir, true));
      assertFalse(fs.exists(tempRestoreDir), "Restore directory should not exist after deletion");
      Job failureJob = Job.getInstance(UTIL.getConfiguration());
      failureJob.setJarByClass(TestTableSnapshotInputFormat.class);
      TableMapReduceUtil.addDependencyJarsForClasses(failureJob.getConfiguration(),
        TestTableSnapshotInputFormat.class);
      Configuration failureConf = failureJob.getConfiguration();
      // Configure job to use the deleted restore directory
      failureConf.set("hbase.TableSnapshotInputFormat.snapshot.name", snapshotName);
      failureConf.set("hbase.TableSnapshotInputFormat.restore.dir", tempRestoreDir.toString());
      failureConf.setInt("hbase.mapreduce.splits.per.region", 1);
      failureJob.setReducerClass(TestTableSnapshotReducer.class);
      failureJob.setNumReduceTasks(1);
      failureJob.setOutputFormatClass(NullOutputFormat.class);

      TableMapReduceUtil.initTableMapperJob(snapshotName, scan, TestTableSnapshotMapper.class,
        ImmutableBytesWritable.class, NullWritable.class, failureJob, false, false,
        TableSnapshotInputFormat.class);
      TableMapReduceUtil.resetCacheConfig(failureConf);

      assertFalse(fs.exists(tempRestoreDir),
        "Restore directory should not exist before job execution");
      failureJob.waitForCompletion(true);

      assertFalse(failureJob.isSuccessful(),
        "Job should fail since the restored snapshot directory is deleted");

    } finally {
      try {
        if (UTIL.getAdmin().tableExists(tableName)) {
          UTIL.deleteTable(tableName);
        }
      } catch (Exception e) {
        LOG.warn("Error deleting table", e);
      }
      try {
        UTIL.getAdmin().deleteSnapshot(snapshotName);
      } catch (Exception e) {
        LOG.warn("Error deleting snapshot", e);
      }
    }
  }
}