Java RecordReader.initialize方法代码示例

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
private static List<Text> readSplit(KeyValueTextInputFormat format, 
    InputSplit split, Job job) throws IOException, InterruptedException {
  List<Text> result = new ArrayList<Text>();
  Configuration conf = job.getConfiguration();
  TaskAttemptContext context = MapReduceTestUtil.
    createDummyMapTaskAttemptContext(conf);
  RecordReader<Text, Text> reader = format.createRecordReader(split, 
    MapReduceTestUtil.createDummyMapTaskAttemptContext(conf));
  MapContext<Text, Text, Text, Text> mcontext = 
    new MapContextImpl<Text, Text, Text, Text>(conf, 
    context.getTaskAttemptID(), reader, null, null,
    MapReduceTestUtil.createDummyReporter(), 
    split);
  reader.initialize(split, mcontext);
  while (reader.nextKeyValue()) {
    result.add(new Text(reader.getCurrentValue()));
  }
  reader.close();
  return result;
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
@Test
public void testReinit() throws Exception {
  // Test that a split containing multiple files works correctly,
  // with the child RecordReader getting its initialize() method
  // called a second time.
  TaskAttemptID taskId = new TaskAttemptID("jt", 0, TaskType.MAP, 0, 0);
  Configuration conf = new Configuration();
  TaskAttemptContext context = new TaskAttemptContextImpl(conf, taskId);

  // This will create a CombineFileRecordReader that itself contains a
  // DummyRecordReader.
  InputFormat inputFormat = new ChildRRInputFormat();

  Path [] files = { new Path("file1"), new Path("file2") };
  long [] lengths = { 1, 1 };

  CombineFileSplit split = new CombineFileSplit(files, lengths);
  RecordReader rr = inputFormat.createRecordReader(split, context);
  assertTrue("Unexpected RR type!", rr instanceof CombineFileRecordReader);

  // first initialize() call comes from MapTask. We'll do it here.
  rr.initialize(split, context);

  // First value is first filename.
  assertTrue(rr.nextKeyValue());
  assertEquals("file1", rr.getCurrentValue().toString());

  // The inner RR will return false, because it only emits one (k, v) pair.
  // But there's another sub-split to process. This returns true to us.
  assertTrue(rr.nextKeyValue());
  
  // And the 2nd rr will have its initialize method called correctly.
  assertEquals("file2", rr.getCurrentValue().toString());
  
  // But after both child RR's have returned their singleton (k, v), this
  // should also return false.
  assertFalse(rr.nextKeyValue());
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
private static List<Text> readSplit(InputFormat<LongWritable,Text> format,
  InputSplit split, Job job) throws IOException, InterruptedException {
  List<Text> result = new ArrayList<Text>();
  Configuration conf = job.getConfiguration();
  TaskAttemptContext context = MapReduceTestUtil.
    createDummyMapTaskAttemptContext(conf);
  RecordReader<LongWritable, Text> reader = format.createRecordReader(split,
    MapReduceTestUtil.createDummyMapTaskAttemptContext(conf));
  MapContext<LongWritable,Text,LongWritable,Text> mcontext =
    new MapContextImpl<LongWritable,Text,LongWritable,Text>(conf,
    context.getTaskAttemptID(), reader, null, null,
    MapReduceTestUtil.createDummyReporter(),
    split);
  reader.initialize(split, mcontext);
  while (reader.nextKeyValue()) {
    result.add(new Text(reader.getCurrentValue()));
  }
  return result;
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
@Test
public void getSplits() throws Exception {
  S3MapReduceCpOptions options = getOptions();
  Configuration configuration = new Configuration();
  configuration.set("mapred.map.tasks", String.valueOf(options.getMaxMaps()));
  CopyListing.getCopyListing(configuration, CREDENTIALS, options).buildListing(
      new Path(cluster.getFileSystem().getUri().toString() + "/tmp/testDynInputFormat/fileList.seq"), options);

  JobContext jobContext = new JobContextImpl(configuration, new JobID());
  DynamicInputFormat<Text, CopyListingFileStatus> inputFormat = new DynamicInputFormat<>();
  List<InputSplit> splits = inputFormat.getSplits(jobContext);

  int nFiles = 0;
  int taskId = 0;

  for (InputSplit split : splits) {
    RecordReader<Text, CopyListingFileStatus> recordReader = inputFormat.createRecordReader(split, null);
    StubContext stubContext = new StubContext(jobContext.getConfiguration(), recordReader, taskId);
    final TaskAttemptContext taskAttemptContext = stubContext.getContext();

    recordReader.initialize(splits.get(0), taskAttemptContext);
    float previousProgressValue = 0f;
    while (recordReader.nextKeyValue()) {
      CopyListingFileStatus fileStatus = recordReader.getCurrentValue();
      String source = fileStatus.getPath().toString();
      assertTrue(expectedFilePaths.contains(source));
      final float progress = recordReader.getProgress();
      assertTrue(progress >= previousProgressValue);
      assertTrue(progress >= 0.0f);
      assertTrue(progress <= 1.0f);
      previousProgressValue = progress;
      ++nFiles;
    }
    assertTrue(recordReader.getProgress() == 1.0f);

    ++taskId;
  }

  Assert.assertEquals(expectedFilePaths.size(), nFiles);
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
private int countRecords(int numSplits) 
    throws IOException, InterruptedException {
  InputFormat<Text, BytesWritable> format =
    new SequenceFileInputFilter<Text, BytesWritable>();
  if (numSplits == 0) {
    numSplits =
      random.nextInt(MAX_LENGTH / (SequenceFile.SYNC_INTERVAL / 20)) + 1;
  }
  FileInputFormat.setMaxInputSplitSize(job, 
    fs.getFileStatus(inFile).getLen() / numSplits);
  TaskAttemptContext context = MapReduceTestUtil.
    createDummyMapTaskAttemptContext(job.getConfiguration());
  // check each split
  int count = 0;
  for (InputSplit split : format.getSplits(job)) {
    RecordReader<Text, BytesWritable> reader =
      format.createRecordReader(split, context);
    MapContext<Text, BytesWritable, Text, BytesWritable> mcontext = 
      new MapContextImpl<Text, BytesWritable, Text, BytesWritable>(
      job.getConfiguration(), 
      context.getTaskAttemptID(), reader, null, null, 
      MapReduceTestUtil.createDummyReporter(), split);
    reader.initialize(split, mcontext);
    try {
      while (reader.nextKeyValue()) {
        LOG.info("Accept record " + reader.getCurrentKey().toString());
        count++;
      }
    } finally {
      reader.close();
    }
  }
  return count;
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
@Test
public void testRecordReaderInit() throws InterruptedException, IOException {
  // Test that we properly initialize the child recordreader when
  // CombineFileInputFormat and CombineFileRecordReader are used.

  TaskAttemptID taskId = new TaskAttemptID("jt", 0, TaskType.MAP, 0, 0);
  Configuration conf1 = new Configuration();
  conf1.set(DUMMY_KEY, "STATE1");
  TaskAttemptContext context1 = new TaskAttemptContextImpl(conf1, taskId);

  // This will create a CombineFileRecordReader that itself contains a
  // DummyRecordReader.
  InputFormat inputFormat = new ChildRRInputFormat();

  Path [] files = { new Path("file1") };
  long [] lengths = { 1 };

  CombineFileSplit split = new CombineFileSplit(files, lengths);

  RecordReader rr = inputFormat.createRecordReader(split, context1);
  assertTrue("Unexpected RR type!", rr instanceof CombineFileRecordReader);

  // Verify that the initial configuration is the one being used.
  // Right after construction the dummy key should have value "STATE1"
  assertEquals("Invalid initial dummy key value", "STATE1",
    rr.getCurrentKey().toString());

  // Switch the active context for the RecordReader...
  Configuration conf2 = new Configuration();
  conf2.set(DUMMY_KEY, "STATE2");
  TaskAttemptContext context2 = new TaskAttemptContextImpl(conf2, taskId);
  rr.initialize(split, context2);

  // And verify that the new context is updated into the child record reader.
  assertEquals("Invalid secondary dummy key value", "STATE2",
    rr.getCurrentKey().toString());
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
/**
 * Test with no record length set.
 */
@Test (timeout=5000)
public void testNoRecordLength() throws Exception {
  localFs.delete(workDir, true);
  Path file = new Path(workDir, new String("testFormat.txt"));
  createFile(file, null, 10, 10);
  // Create the job and do not set fixed record length
  Job job = Job.getInstance(defaultConf);
  FileInputFormat.setInputPaths(job, workDir);
  FixedLengthInputFormat format = new FixedLengthInputFormat();
  List<InputSplit> splits = format.getSplits(job);
  boolean exceptionThrown = false;
  for (InputSplit split : splits) {
    try {
      TaskAttemptContext context = MapReduceTestUtil.
          createDummyMapTaskAttemptContext(job.getConfiguration());
      RecordReader<LongWritable, BytesWritable> reader =
          format.createRecordReader(split, context);
      MapContext<LongWritable, BytesWritable, LongWritable, BytesWritable>
          mcontext =
          new MapContextImpl<LongWritable, BytesWritable, LongWritable,
          BytesWritable>(job.getConfiguration(), context.getTaskAttemptID(),
          reader, null, null, MapReduceTestUtil.createDummyReporter(), split);
      reader.initialize(split, mcontext);
    } catch(IOException ioe) {
      exceptionThrown = true;
      LOG.info("Exception message:" + ioe.getMessage());
    }
  }
  assertTrue("Exception for not setting record length:", exceptionThrown);
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
/**
 * Test with record length set to 0
 */
@Test (timeout=5000)
public void testZeroRecordLength() throws Exception {
  localFs.delete(workDir, true);
  Path file = new Path(workDir, new String("testFormat.txt"));
  createFile(file, null, 10, 10);
  Job job = Job.getInstance(defaultConf);
  // Set the fixed length record length config property 
  FixedLengthInputFormat format = new FixedLengthInputFormat();
  format.setRecordLength(job.getConfiguration(), 0);
  FileInputFormat.setInputPaths(job, workDir);
  List<InputSplit> splits = format.getSplits(job);
  boolean exceptionThrown = false;
  for (InputSplit split : splits) {
    try {
      TaskAttemptContext context =
          MapReduceTestUtil.createDummyMapTaskAttemptContext(
          job.getConfiguration());
      RecordReader<LongWritable, BytesWritable> reader = 
          format.createRecordReader(split, context);
      MapContext<LongWritable, BytesWritable, LongWritable, BytesWritable>
          mcontext =
          new MapContextImpl<LongWritable, BytesWritable, LongWritable,
          BytesWritable>(job.getConfiguration(), context.getTaskAttemptID(),
          reader, null, null, MapReduceTestUtil.createDummyReporter(), split);
      reader.initialize(split, mcontext);
    } catch(IOException ioe) {
      exceptionThrown = true;
      LOG.info("Exception message:" + ioe.getMessage());
    }
  }
  assertTrue("Exception for zero record length:", exceptionThrown);
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
/**
 * Test with record length set to a negative value
 */
@Test (timeout=5000)
public void testNegativeRecordLength() throws Exception {
  localFs.delete(workDir, true);
  Path file = new Path(workDir, new String("testFormat.txt"));
  createFile(file, null, 10, 10);
  // Set the fixed length record length config property 
  Job job = Job.getInstance(defaultConf);
  FixedLengthInputFormat format = new FixedLengthInputFormat();
  format.setRecordLength(job.getConfiguration(), -10);
  FileInputFormat.setInputPaths(job, workDir);
  List<InputSplit> splits = format.getSplits(job);
  boolean exceptionThrown = false;
  for (InputSplit split : splits) {
    try {
      TaskAttemptContext context = MapReduceTestUtil.
          createDummyMapTaskAttemptContext(job.getConfiguration());
      RecordReader<LongWritable, BytesWritable> reader = 
          format.createRecordReader(split, context);
      MapContext<LongWritable, BytesWritable, LongWritable, BytesWritable>
          mcontext =
          new MapContextImpl<LongWritable, BytesWritable, LongWritable,
          BytesWritable>(job.getConfiguration(), context.getTaskAttemptID(),
          reader, null, null, MapReduceTestUtil.createDummyReporter(), split);
      reader.initialize(split, mcontext);
    } catch(IOException ioe) {
      exceptionThrown = true;
      LOG.info("Exception message:" + ioe.getMessage());
    }
  }
  assertTrue("Exception for negative record length:", exceptionThrown);
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
private static List<String> readSplit(FixedLengthInputFormat format, 
                                      InputSplit split, 
                                      Job job) throws Exception {
  List<String> result = new ArrayList<String>();
  TaskAttemptContext context = MapReduceTestUtil.
      createDummyMapTaskAttemptContext(job.getConfiguration());
  RecordReader<LongWritable, BytesWritable> reader =
      format.createRecordReader(split, context);
  MapContext<LongWritable, BytesWritable, LongWritable, BytesWritable>
      mcontext =
      new MapContextImpl<LongWritable, BytesWritable, LongWritable,
      BytesWritable>(job.getConfiguration(), context.getTaskAttemptID(),
      reader, null, null, MapReduceTestUtil.createDummyReporter(), split);
  LongWritable key;
  BytesWritable value;
  try {
    reader.initialize(split, mcontext);
    while (reader.nextKeyValue()) {
      key = reader.getCurrentKey();
      value = reader.getCurrentValue();
      result.add(new String(value.getBytes(), 0, value.getLength()));
    }
  } finally {
    reader.close();
  }
  return result;
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
/**
 * From each split sampled, take the first numSamples / numSplits records.
 */
@SuppressWarnings("unchecked") // ArrayList::toArray doesn't preserve type
public K[] getSample(InputFormat<K,V> inf, Job job) 
    throws IOException, InterruptedException {
  List<InputSplit> splits = inf.getSplits(job);
  ArrayList<K> samples = new ArrayList<K>(numSamples);
  int splitsToSample = Math.min(maxSplitsSampled, splits.size());
  int samplesPerSplit = numSamples / splitsToSample;
  long records = 0;
  for (int i = 0; i < splitsToSample; ++i) {
    TaskAttemptContext samplingContext = new TaskAttemptContextImpl(
        job.getConfiguration(), new TaskAttemptID());
    RecordReader<K,V> reader = inf.createRecordReader(
        splits.get(i), samplingContext);
    reader.initialize(splits.get(i), samplingContext);
    while (reader.nextKeyValue()) {
      samples.add(ReflectionUtils.copy(job.getConfiguration(),
                                       reader.getCurrentKey(), null));
      ++records;
      if ((i+1) * samplesPerSplit <= records) {
        break;
      }
    }
    reader.close();
  }
  return (K[])samples.toArray();
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
@SuppressWarnings({"rawtypes", "unchecked"})
@Test (timeout=10000)
public void testLoadMapper() throws Exception {

  Configuration conf = new Configuration();
  conf.setInt(JobContext.NUM_REDUCES, 2);

  CompressionEmulationUtil.setCompressionEmulationEnabled(conf, true);
  conf.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true);

  TaskAttemptID taskId = new TaskAttemptID();
  RecordReader<NullWritable, GridmixRecord> reader = new FakeRecordReader();

  LoadRecordGkGrWriter writer = new LoadRecordGkGrWriter();

  OutputCommitter committer = new CustomOutputCommitter();
  StatusReporter reporter = new TaskAttemptContextImpl.DummyReporter();
  LoadSplit split = getLoadSplit();

  MapContext<NullWritable, GridmixRecord, GridmixKey, GridmixRecord> mapContext = new MapContextImpl<NullWritable, GridmixRecord, GridmixKey, GridmixRecord>(
          conf, taskId, reader, writer, committer, reporter, split);
  // context
  Context ctx = new WrappedMapper<NullWritable, GridmixRecord, GridmixKey, GridmixRecord>()
          .getMapContext(mapContext);

  reader.initialize(split, ctx);
  ctx.getConfiguration().setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true);
  CompressionEmulationUtil.setCompressionEmulationEnabled(
          ctx.getConfiguration(), true);

  LoadJob.LoadMapper mapper = new LoadJob.LoadMapper();
  // setup, map, clean
  mapper.run(ctx);

  Map<GridmixKey, GridmixRecord> data = writer.getData();
  // check result
  assertEquals(2, data.size());

}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
/**
 * Validate setupGenerateDistCacheData by validating <li>permissions of the
 * distributed cache directories and <li>content of the generated sequence
 * file. This includes validation of dist cache file paths and their file
 * sizes.
 */
private void validateSetupGenDC(Configuration jobConf, long[] sortedFileSizes)
    throws IOException, InterruptedException {
  // build things needed for validation
  long sumOfFileSizes = 0;
  for (int i = 0; i < sortedFileSizes.length; i++) {
    sumOfFileSizes += sortedFileSizes[i];
  }

  FileSystem fs = FileSystem.get(jobConf);
  assertEquals("Number of distributed cache files to be generated is wrong.",
      sortedFileSizes.length,
      jobConf.getInt(GenerateDistCacheData.GRIDMIX_DISTCACHE_FILE_COUNT, -1));
  assertEquals("Total size of dist cache files to be generated is wrong.",
      sumOfFileSizes,
      jobConf.getLong(GenerateDistCacheData.GRIDMIX_DISTCACHE_BYTE_COUNT, -1));
  Path filesListFile = new Path(
      jobConf.get(GenerateDistCacheData.GRIDMIX_DISTCACHE_FILE_LIST));
  FileStatus stat = fs.getFileStatus(filesListFile);
  assertEquals("Wrong permissions of dist Cache files list file "
      + filesListFile, new FsPermission((short) 0644), stat.getPermission());

  InputSplit split = new FileSplit(filesListFile, 0, stat.getLen(),
      (String[]) null);
  TaskAttemptContext taskContext = MapReduceTestUtil
      .createDummyMapTaskAttemptContext(jobConf);
  RecordReader<LongWritable, BytesWritable> reader = new GenerateDistCacheData.GenDCDataFormat()
      .createRecordReader(split, taskContext);
  MapContext<LongWritable, BytesWritable, NullWritable, BytesWritable> mapContext = new MapContextImpl<LongWritable, BytesWritable, NullWritable, BytesWritable>(
      jobConf, taskContext.getTaskAttemptID(), reader, null, null,
      MapReduceTestUtil.createDummyReporter(), split);
  reader.initialize(split, mapContext);

  // start validating setupGenerateDistCacheData
  doValidateSetupGenDC(reader, fs, sortedFileSizes);
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
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);

  Assert.assertEquals(expectedNumSplits, splits.size());

  HBaseTestingUtility.SeenRowTracker rowTracker =
      new HBaseTestingUtility.SeenRowTracker(startRow, stopRow);

  for (int i = 0; i < splits.size(); i++) {
    // validate input split
    InputSplit split = splits.get(i);
    Assert.assertTrue(split instanceof TableSnapshotRegionSplit);

    // 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();
}
 

import org.apache.hadoop.mapreduce.RecordReader; //导入方法依赖的package包/类
public void testBinary() throws IOException, InterruptedException {
  Job job = Job.getInstance();
  FileSystem fs = FileSystem.getLocal(job.getConfiguration());
  Path dir = new Path(System.getProperty("test.build.data",".") + "/mapred");
  Path file = new Path(dir, "testbinary.seq");
  Random r = new Random();
  long seed = r.nextLong();
  r.setSeed(seed);

  fs.delete(dir, true);
  FileInputFormat.setInputPaths(job, dir);

  Text tkey = new Text();
  Text tval = new Text();

  SequenceFile.Writer writer = new SequenceFile.Writer(fs,
    job.getConfiguration(), file, Text.class, Text.class);
  try {
    for (int i = 0; i < RECORDS; ++i) {
      tkey.set(Integer.toString(r.nextInt(), 36));
      tval.set(Long.toString(r.nextLong(), 36));
      writer.append(tkey, tval);
    }
  } finally {
    writer.close();
  }
  TaskAttemptContext context = MapReduceTestUtil.
    createDummyMapTaskAttemptContext(job.getConfiguration());
  InputFormat<BytesWritable,BytesWritable> bformat =
    new SequenceFileAsBinaryInputFormat();

  int count = 0;
  r.setSeed(seed);
  BytesWritable bkey = new BytesWritable();
  BytesWritable bval = new BytesWritable();
  Text cmpkey = new Text();
  Text cmpval = new Text();
  DataInputBuffer buf = new DataInputBuffer();
  FileInputFormat.setInputPaths(job, file);
  for (InputSplit split : bformat.getSplits(job)) {
    RecordReader<BytesWritable, BytesWritable> reader =
          bformat.createRecordReader(split, context);
    MapContext<BytesWritable, BytesWritable, BytesWritable, BytesWritable> 
      mcontext = new MapContextImpl<BytesWritable, BytesWritable,
        BytesWritable, BytesWritable>(job.getConfiguration(), 
        context.getTaskAttemptID(), reader, null, null, 
        MapReduceTestUtil.createDummyReporter(), 
        split);
    reader.initialize(split, mcontext);
    try {
      while (reader.nextKeyValue()) {
        bkey = reader.getCurrentKey();
        bval = reader.getCurrentValue();
        tkey.set(Integer.toString(r.nextInt(), 36));
        tval.set(Long.toString(r.nextLong(), 36));
        buf.reset(bkey.getBytes(), bkey.getLength());
        cmpkey.readFields(buf);
        buf.reset(bval.getBytes(), bval.getLength());
        cmpval.readFields(buf);
        assertTrue(
          "Keys don't match: " + "*" + cmpkey.toString() + ":" +
          tkey.toString() + "*",
          cmpkey.toString().equals(tkey.toString()));
        assertTrue(
          "Vals don't match: " + "*" + cmpval.toString() + ":" +
          tval.toString() + "*",
          cmpval.toString().equals(tval.toString()));
        ++count;
      }
    } finally {
      reader.close();
    }
  }
  assertEquals("Some records not found", RECORDS, count);
}