@Override
public void map(
WritableComparable key,
CompactorInputSplit split,
OutputCollector<NullWritable, NullWritable> nullWritableVOutputCollector,
Reporter reporter)
throws IOException {
// This will only get called once, since CompactRecordReader only returns one record,
// the input split.
// Based on the split we're passed we go instantiate the real reader and then iterate on it
// until it finishes.
@SuppressWarnings("unchecked") // since there is no way to parametrize instance of Class
AcidInputFormat<WritableComparable, V> aif =
instantiate(AcidInputFormat.class, jobConf.get(INPUT_FORMAT_CLASS_NAME));
ValidTxnList txnList = new ValidReadTxnList(jobConf.get(ValidTxnList.VALID_TXNS_KEY));
boolean isMajor = jobConf.getBoolean(IS_MAJOR, false);
AcidInputFormat.RawReader<V> reader =
aif.getRawReader(
jobConf,
isMajor,
split.getBucket(),
txnList,
split.getBaseDir(),
split.getDeltaDirs());
RecordIdentifier identifier = reader.createKey();
V value = reader.createValue();
getWriter(reporter, reader.getObjectInspector(), split.getBucket());
while (reader.next(identifier, value)) {
if (isMajor && reader.isDelete(value)) continue;
writer.write(value);
reporter.progress();
}
}
/**
* Mapper configuration. Extracts source and destination file system, as well as top-level paths
* on source and destination directories. Gets the named file systems, to be used later in map.
*/
public void configure(JobConf job) {
destPath = new Path(job.get(DST_DIR_LABEL, "/"));
try {
destFileSys = destPath.getFileSystem(job);
} catch (IOException ex) {
throw new RuntimeException("Unable to get the named file system.", ex);
}
sizeBuf = job.getInt("copy.buf.size", 128 * 1024);
buffer = new byte[sizeBuf];
ignoreReadFailures = job.getBoolean(Options.IGNORE_READ_FAILURES.propertyname, false);
preserve_status = job.getBoolean(Options.PRESERVE_STATUS.propertyname, false);
if (preserve_status) {
preseved = FileAttribute.parse(job.get(PRESERVE_STATUS_LABEL));
}
update = job.getBoolean(Options.UPDATE.propertyname, false);
overwrite = !update && job.getBoolean(Options.OVERWRITE.propertyname, false);
this.job = job;
}
/**
* {@inheritDoc}
*
* @see org.apache.hadoop.mapred.MapReduceBase#configure(org.apache.hadoop.mapred.JobConf)
*/
@Override
public void configure(JobConf job) {
m_caseSensitive = job.getBoolean(JOB_PARAMETER_CASE_SENSITIVE, true);
m_inputFile = job.get(MAP_PARAMETER_INPUT_FILE);
if (job.getBoolean(JOB_PARAMETER_SKIP_PATTERNS, false)) {
Path[] patternsFiles = new Path[0];
try {
patternsFiles = DistributedCache.getLocalCacheFiles(job);
} catch (IOException ioe) {
System.err.println(
"Caught exception while getting cached files: "
+ StringUtils.stringifyException(ioe));
}
for (Path patternsFile : patternsFiles) {
parseSkipFile(patternsFile);
}
}
}
public void checkOutputSpecs(FileSystem fs, JobConf conf) throws IOException {
Args args = (Args) Utils.getObject(conf, ARGS_CONF);
fs = Utils.getFS(args.outputDirHdfs, conf);
if (conf.getBoolean("mapred.reduce.tasks.speculative.execution", true)) {
// Because we don't want to write a bunch of extra times.
throw new InvalidJobConfException("Speculative execution should be false");
}
if (fs.exists(new Path(args.outputDirHdfs))) {
throw new InvalidJobConfException("Output dir already exists " + args.outputDirHdfs);
}
}
/**
* Test if {@link CompressionEmulationUtil#configureCompressionEmulation(
* org.apache.hadoop.mapred.JobConf, org.apache.hadoop.mapred.JobConf)} can extract compression
* related configuration parameters.
*/
@Test
public void testExtractCompressionConfigs() {
JobConf source = new JobConf();
JobConf target = new JobConf();
// set the default values
source.setBoolean(FileOutputFormat.COMPRESS, false);
source.set(FileOutputFormat.COMPRESS_CODEC, "MyDefaultCodec");
source.set(FileOutputFormat.COMPRESS_TYPE, "MyDefaultType");
source.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, false);
source.set(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC, "MyDefaultCodec2");
CompressionEmulationUtil.configureCompressionEmulation(source, target);
// check default values
assertFalse(target.getBoolean(FileOutputFormat.COMPRESS, true));
assertEquals("MyDefaultCodec", target.get(FileOutputFormat.COMPRESS_CODEC));
assertEquals("MyDefaultType", target.get(FileOutputFormat.COMPRESS_TYPE));
assertFalse(target.getBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true));
assertEquals("MyDefaultCodec2", target.get(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC));
assertFalse(CompressionEmulationUtil.isInputCompressionEmulationEnabled(target));
// set new values
source.setBoolean(FileOutputFormat.COMPRESS, true);
source.set(FileOutputFormat.COMPRESS_CODEC, "MyCodec");
source.set(FileOutputFormat.COMPRESS_TYPE, "MyType");
source.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true);
source.set(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC, "MyCodec2");
org.apache.hadoop.mapred.FileInputFormat.setInputPaths(source, "file.gz");
target = new JobConf(); // reset
CompressionEmulationUtil.configureCompressionEmulation(source, target);
// check new values
assertTrue(target.getBoolean(FileOutputFormat.COMPRESS, false));
assertEquals("MyCodec", target.get(FileOutputFormat.COMPRESS_CODEC));
assertEquals("MyType", target.get(FileOutputFormat.COMPRESS_TYPE));
assertTrue(target.getBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, false));
assertEquals("MyCodec2", target.get(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC));
assertTrue(CompressionEmulationUtil.isInputCompressionEmulationEnabled(target));
}
@VisibleForTesting
Fetcher(
JobConf job,
TaskAttemptID reduceId,
ShuffleSchedulerImpl<K, V> scheduler,
MergeManager<K, V> merger,
Reporter reporter,
ShuffleClientMetrics metrics,
ExceptionReporter exceptionReporter,
SecretKey shuffleKey,
int id) {
this.jobConf = job;
this.reporter = reporter;
this.scheduler = scheduler;
this.merger = merger;
this.metrics = metrics;
this.exceptionReporter = exceptionReporter;
this.id = id;
this.reduce = reduceId.getTaskID().getId();
this.shuffleSecretKey = shuffleKey;
ioErrs = reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.IO_ERROR.toString());
wrongLengthErrs =
reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.WRONG_LENGTH.toString());
badIdErrs = reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.BAD_ID.toString());
wrongMapErrs = reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.WRONG_MAP.toString());
connectionErrs = reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.CONNECTION.toString());
wrongReduceErrs =
reporter.getCounter(SHUFFLE_ERR_GRP_NAME, ShuffleErrors.WRONG_REDUCE.toString());
this.connectionTimeout =
job.getInt(MRJobConfig.SHUFFLE_CONNECT_TIMEOUT, DEFAULT_STALLED_COPY_TIMEOUT);
this.readTimeout = job.getInt(MRJobConfig.SHUFFLE_READ_TIMEOUT, DEFAULT_READ_TIMEOUT);
setName("fetcher#" + id);
setDaemon(true);
synchronized (Fetcher.class) {
sslShuffle =
job.getBoolean(MRConfig.SHUFFLE_SSL_ENABLED_KEY, MRConfig.SHUFFLE_SSL_ENABLED_DEFAULT);
if (sslShuffle && sslFactory == null) {
sslFactory = new SSLFactory(SSLFactory.Mode.CLIENT, job);
try {
sslFactory.init();
} catch (Exception ex) {
sslFactory.destroy();
throw new RuntimeException(ex);
}
}
}
}
public static void main(int step, Path inputDir, JobConf job) throws IOException {
FileSystem hdfs = inputDir.getFileSystem(job);
if (!hdfs.exists(Collector.partitionSizesPath)) {
System.out.println("Partition sizes file does not exists!");
return;
}
debugStages = job.getBoolean(Config.DEBUG_STAGES_PROPERTY, Config.DEBUG_STAGES_VALUE);
MapFile.Reader partitionSizeReader =
new MapFile.Reader(hdfs, Collector.partitionSizesPath.getName(), new JobConf());
Text partitionK = new Text();
LongWritable partSizeV = new LongWritable();
try {
while (partitionSizeReader.next(partitionK, partSizeV)) {
partitionsNames.add(partitionK.toString()); // useless?
partitionsSizes.put(partitionK.toString(), partSizeV.get());
}
} catch (Exception e) {;
}
for (int i = 0; i < partitionsNames.size(); i++) {
System.out.println(
"Partition "
+ partitionsNames.get(i)
+ " has "
+ partitionsSizes.get(partitionsNames.get(i))
+ " vectors.");
}
if (partitionsNames.size() <= 1) return;
stage0();
printUndirectedNeighbors("Stage0");
printPartitionsStat("Stage0");
printCircularPartitionsWeight("\nCircular");
calcCWStandardDeviation();
stage1();
printDirectedNeighbors("Stage1");
System.out.println("Stage 1 final weights: ");
printPartitionsWeights("Stage1");
if ((step == 2) || (step == 12)) {
stage2();
printDirectedNeighbors("Stage2");
System.out.println("Stage 2 final weights: ");
printPartitionsWeights("Stage2");
}
// stage3(job, hdfs);
writeComparisonList(job, hdfs);
// printComparisonList(job, hdfs);// remove
}
private void getWriter(Reporter reporter, ObjectInspector inspector, int bucket)
throws IOException {
if (writer == null) {
AcidOutputFormat.Options options = new AcidOutputFormat.Options(jobConf);
options
.inspector(inspector)
.writingBase(jobConf.getBoolean(IS_MAJOR, false))
.isCompressed(jobConf.getBoolean(IS_COMPRESSED, false))
.tableProperties(new StringableMap(jobConf.get(TABLE_PROPS)).toProperties())
.reporter(reporter)
.minimumTransactionId(jobConf.getLong(MIN_TXN, Long.MAX_VALUE))
.maximumTransactionId(jobConf.getLong(MAX_TXN, Long.MIN_VALUE))
.bucket(bucket)
.statementId(-1); // setting statementId == -1 makes compacted delta files use
// delta_xxxx_yyyy format
// Instantiate the underlying output format
@SuppressWarnings("unchecked") // since there is no way to parametrize instance of Class
AcidOutputFormat<WritableComparable, V> aof =
instantiate(AcidOutputFormat.class, jobConf.get(OUTPUT_FORMAT_CLASS_NAME));
writer = aof.getRawRecordWriter(new Path(jobConf.get(TMP_LOCATION)), options);
}
}
@Override
public void configure(JobConf job) {
this.job = job;
mapper = new ObjectMapper();
messagePack = new MessagePack();
timeFmt = DateTimeFormat.forPattern("YYYYMMddHHmmssSSS");
oidSerial = new AtomicInteger(0);
serial = new AtomicInteger(0);
storeAttirbute = job.getBoolean(CIngest.CONF_INGEST_STORE_ATTR, false);
Iterator<String> tokens =
Splitter.on("///").split(job.get(CIngest.CONF_CASSANDRA_CONNECT_TOKEN)).iterator();
String clusterName = tokens.next();
String seeds = tokens.next();
String keyspaceName = tokens.next();
String columeFamilyName = tokens.next();
context =
new AstyanaxContext.Builder()
.forCluster(clusterName)
.forKeyspace(keyspaceName)
.withAstyanaxConfiguration(
new AstyanaxConfigurationImpl().setDiscoveryType(NodeDiscoveryType.TOKEN_AWARE))
.withConnectionPoolConfiguration(
new ConnectionPoolConfigurationImpl("cp")
.setPort(9160)
.setMaxConnsPerHost(2)
.setSeeds(seeds))
.withConnectionPoolMonitor(new CountingConnectionPoolMonitor())
.buildKeyspace(ThriftFamilyFactory.getInstance());
context.start();
ks = context.getEntity();
cf =
new ColumnFamily<String, String>(
columeFamilyName, StringSerializer.get(), StringSerializer.get());
}
public void configure(JobConf job) {
sLogger.setLevel(Level.OFF);
src = Language.languageForISO639_1(job.get(SRC_LANG));
tgt = Language.languageForISO639_1(job.get(TGT_LANG));
sLogger.debug("Source language: " + src.code());
sLogger.debug("Target language: " + tgt.code());
boolean useVocabServer = false;
if (!useVocabServer) {
if (vocE == null) vocE = new VocabularyWritable();
if (vocF == null) vocF = new VocabularyWritable();
} else {
try {
vocE =
new VocabServerClient(
job.get("ha.vocabserver.host"),
Integer.parseInt(job.get("ha.vocabserver.port1")));
vocF =
new VocabServerClient(
job.get("ha.vocabserver.host"),
Integer.parseInt(job.get("ha.vocabserver.port2")));
} catch (IOException e) {
e.printStackTrace();
throw new RuntimeException(e);
}
}
lp = LanguagePair.languageForISO639_1Pair(src.code() + "-" + tgt.code());
if (job.getBoolean("ha.trunc.use", true)) {
sawp = AlignmentWordPreprocessor.CreatePreprocessor(lp, src, job);
tawp = AlignmentWordPreprocessor.CreatePreprocessor(lp, tgt, job);
} else {
sawp = AlignmentWordPreprocessor.CreatePreprocessor(null, null, job);
tawp = AlignmentWordPreprocessor.CreatePreprocessor(null, null, job);
}
job_ = job;
}
public void configure(JobConf job) {
interval = job.getInt("db.fetch.interval.default", 2592000);
scoreInjected = job.getFloat("db.score.injected", 1.0f);
overwrite = job.getBoolean("db.injector.overwrite", false);
update = job.getBoolean("db.injector.update", false);
}
/**
* Return if native hadoop libraries, if present, can be used for this job.
*
* @param jobConf job configuration
* @return <code>true</code> if native hadoop libraries, if present, can be used for this job;
* <code>false</code> otherwise.
*/
public boolean getLoadNativeLibraries(JobConf jobConf) {
return jobConf.getBoolean("hadoop.native.lib", true);
}
@Override
public InputSplit[] getSplits(JobConf jobConf, int numSplits) throws IOException {
// obtain delegation tokens for the job
if (UserGroupInformation.getCurrentUser().hasKerberosCredentials()) {
TableMapReduceUtil.initCredentials(jobConf);
}
String hbaseTableName = jobConf.get(HBaseSerDe.HBASE_TABLE_NAME);
setHTable(new HTable(HBaseConfiguration.create(jobConf), Bytes.toBytes(hbaseTableName)));
String hbaseColumnsMapping = jobConf.get(HBaseSerDe.HBASE_COLUMNS_MAPPING);
boolean doColumnRegexMatching =
jobConf.getBoolean(HBaseSerDe.HBASE_COLUMNS_REGEX_MATCHING, true);
if (hbaseColumnsMapping == null) {
throw new IOException(HBaseSerDe.HBASE_COLUMNS_MAPPING + " required for HBase Table.");
}
ColumnMappings columnMappings = null;
try {
columnMappings = HBaseSerDe.parseColumnsMapping(hbaseColumnsMapping, doColumnRegexMatching);
} catch (SerDeException e) {
throw new IOException(e);
}
int iKey = columnMappings.getKeyIndex();
ColumnMapping keyMapping = columnMappings.getKeyMapping();
// Take filter pushdown into account while calculating splits; this
// allows us to prune off regions immediately. Note that although
// the Javadoc for the superclass getSplits says that it returns one
// split per region, the implementation actually takes the scan
// definition into account and excludes regions which don't satisfy
// the start/stop row conditions (HBASE-1829).
Scan scan =
createFilterScan(
jobConf,
iKey,
HiveHBaseInputFormatUtil.getStorageFormatOfKey(
keyMapping.mappingSpec,
jobConf.get(HBaseSerDe.HBASE_TABLE_DEFAULT_STORAGE_TYPE, "string")));
// The list of families that have been added to the scan
List<String> addedFamilies = new ArrayList<String>();
// REVIEW: are we supposed to be applying the getReadColumnIDs
// same as in getRecordReader?
for (ColumnMapping colMap : columnMappings) {
if (colMap.hbaseRowKey) {
continue;
}
if (colMap.qualifierName == null) {
scan.addFamily(colMap.familyNameBytes);
addedFamilies.add(colMap.familyName);
} else {
if (!addedFamilies.contains(colMap.familyName)) {
// add the column only if the family has not already been added
scan.addColumn(colMap.familyNameBytes, colMap.qualifierNameBytes);
}
}
}
setScan(scan);
Job job = new Job(jobConf);
JobContext jobContext = ShimLoader.getHadoopShims().newJobContext(job);
Path[] tablePaths = FileInputFormat.getInputPaths(jobContext);
List<org.apache.hadoop.mapreduce.InputSplit> splits = super.getSplits(jobContext);
InputSplit[] results = new InputSplit[splits.size()];
for (int i = 0; i < splits.size(); i++) {
results[i] = new HBaseSplit((TableSplit) splits.get(i), tablePaths[0]);
}
return results;
}
public MergeManagerImpl(
TaskAttemptID reduceId,
JobConf jobConf,
FileSystem localFS,
LocalDirAllocator localDirAllocator,
Reporter reporter,
CompressionCodec codec,
Class<? extends Reducer> combinerClass,
CombineOutputCollector<K, V> combineCollector,
Counters.Counter spilledRecordsCounter,
Counters.Counter reduceCombineInputCounter,
Counters.Counter mergedMapOutputsCounter,
ExceptionReporter exceptionReporter,
Progress mergePhase,
MapOutputFile mapOutputFile) {
this.reduceId = reduceId;
this.jobConf = jobConf;
this.localDirAllocator = localDirAllocator;
this.exceptionReporter = exceptionReporter;
this.reporter = reporter;
this.codec = codec;
this.combinerClass = combinerClass;
this.combineCollector = combineCollector;
this.reduceCombineInputCounter = reduceCombineInputCounter;
this.spilledRecordsCounter = spilledRecordsCounter;
this.mergedMapOutputsCounter = mergedMapOutputsCounter;
this.mapOutputFile = mapOutputFile;
this.mapOutputFile.setConf(jobConf);
this.localFS = localFS;
this.rfs = ((LocalFileSystem) localFS).getRaw();
final float maxInMemCopyUse = jobConf.getFloat(MRJobConfig.SHUFFLE_INPUT_BUFFER_PERCENT, 0.90f);
if (maxInMemCopyUse > 1.0 || maxInMemCopyUse < 0.0) {
throw new IllegalArgumentException(
"Invalid value for " + MRJobConfig.SHUFFLE_INPUT_BUFFER_PERCENT + ": " + maxInMemCopyUse);
}
// Allow unit tests to fix Runtime memory
this.memoryLimit =
(long)
(jobConf.getLong(
MRJobConfig.REDUCE_MEMORY_TOTAL_BYTES,
Math.min(Runtime.getRuntime().maxMemory(), Integer.MAX_VALUE))
* maxInMemCopyUse);
this.ioSortFactor = jobConf.getInt(MRJobConfig.IO_SORT_FACTOR, 100);
final float singleShuffleMemoryLimitPercent =
jobConf.getFloat(
MRJobConfig.SHUFFLE_MEMORY_LIMIT_PERCENT, DEFAULT_SHUFFLE_MEMORY_LIMIT_PERCENT);
if (singleShuffleMemoryLimitPercent <= 0.0f || singleShuffleMemoryLimitPercent > 1.0f) {
throw new IllegalArgumentException(
"Invalid value for "
+ MRJobConfig.SHUFFLE_MEMORY_LIMIT_PERCENT
+ ": "
+ singleShuffleMemoryLimitPercent);
}
usedMemory = 0L;
commitMemory = 0L;
this.maxSingleShuffleLimit = (long) (memoryLimit * singleShuffleMemoryLimitPercent);
this.memToMemMergeOutputsThreshold =
jobConf.getInt(MRJobConfig.REDUCE_MEMTOMEM_THRESHOLD, ioSortFactor);
this.mergeThreshold =
(long) (this.memoryLimit * jobConf.getFloat(MRJobConfig.SHUFFLE_MERGE_PERCENT, 0.90f));
LOG.info(
"MergerManager: memoryLimit="
+ memoryLimit
+ ", "
+ "maxSingleShuffleLimit="
+ maxSingleShuffleLimit
+ ", "
+ "mergeThreshold="
+ mergeThreshold
+ ", "
+ "ioSortFactor="
+ ioSortFactor
+ ", "
+ "memToMemMergeOutputsThreshold="
+ memToMemMergeOutputsThreshold);
if (this.maxSingleShuffleLimit >= this.mergeThreshold) {
throw new RuntimeException(
"Invlaid configuration: "
+ "maxSingleShuffleLimit should be less than mergeThreshold"
+ "maxSingleShuffleLimit: "
+ this.maxSingleShuffleLimit
+ "mergeThreshold: "
+ this.mergeThreshold);
}
boolean allowMemToMemMerge = jobConf.getBoolean(MRJobConfig.REDUCE_MEMTOMEM_ENABLED, false);
if (allowMemToMemMerge) {
this.memToMemMerger =
new IntermediateMemoryToMemoryMerger(this, memToMemMergeOutputsThreshold);
this.memToMemMerger.start();
} else {
this.memToMemMerger = null;
}
this.inMemoryMerger = createInMemoryMerger();
this.inMemoryMerger.start();
this.onDiskMerger = new OnDiskMerger(this);
this.onDiskMerger.start();
this.mergePhase = mergePhase;
}
/** {@inheritDoc} */
public void configure(JobConf job) {
this.jobconf = job;
ignoreFailures = jobconf.getBoolean(IGNORE_FAILURES_OPTION_LABEL, true);
st = new Statistics();
}
public void configure(JobConf job) {
memoryMXBean = ManagementFactory.getMemoryMXBean();
l4j.info("maximum memory = " + memoryMXBean.getHeapMemoryUsage().getMax());
try {
l4j.info(
"conf classpath = " + Arrays.asList(((URLClassLoader) job.getClassLoader()).getURLs()));
l4j.info(
"thread classpath = "
+ Arrays.asList(
((URLClassLoader) Thread.currentThread().getContextClassLoader()).getURLs()));
} catch (Exception e) {
l4j.info("cannot get classpath: " + e.getMessage());
}
try {
jc = job;
int estCountBucketSize = jc.getInt("hive.exec.estdistinct.bucketsize.log", 15);
int estCountBufferSize = jc.getInt("hive.exec.estdistinct.buffsize.log", 8);
GenericUDAFCardinalityEstimation.initParams(estCountBucketSize, estCountBufferSize);
boolean isChangSizeZero2Null =
jc.getBoolean(
HiveConf.ConfVars.HIVE_UDTF_EXPLODE_CHANGE_ZERO_SIZE_2_NULL.varname,
HiveConf.ConfVars.HIVE_UDTF_EXPLODE_CHANGE_ZERO_SIZE_2_NULL.defaultBoolVal);
boolean isChangeNull2Null =
jc.getBoolean(
HiveConf.ConfVars.HIVE_UDTF_EXPLODE_CHANGE_NULL_2_NULL.varname,
HiveConf.ConfVars.HIVE_UDTF_EXPLODE_CHANGE_NULL_2_NULL.defaultBoolVal);
GenericUDTFExplode.isChangSizeZero2Null = isChangSizeZero2Null;
GenericUDTFExplode.isChangNull2Null = isChangeNull2Null;
GenericUDTFPosExplode.isChangSizeZero2Null = isChangSizeZero2Null;
GenericUDTFPosExplode.isChangNull2Null = isChangeNull2Null;
mapredWork mrwork = Utilities.getMapRedWork(job);
mo = new MapOperator();
mo.setConf(mrwork);
mo.setChildren(job);
l4j.info(mo.dump(0));
mo.initialize(jc, null);
localWork = mrwork.getMapLocalWork();
if (localWork == null) {
return;
}
fetchOperators = new HashMap<String, FetchOperator>();
for (Map.Entry<String, fetchWork> entry : localWork.getAliasToFetchWork().entrySet()) {
fetchOperators.put(entry.getKey(), new FetchOperator(entry.getValue(), job));
l4j.info("fetchoperator for " + entry.getKey() + " created");
}
for (Map.Entry<String, FetchOperator> entry : fetchOperators.entrySet()) {
Operator<? extends Serializable> forwardOp = localWork.getAliasToWork().get(entry.getKey());
forwardOp.initialize(
jc, new ObjectInspector[] {entry.getValue().getOutputObjectInspector()});
l4j.info("fetchoperator for " + entry.getKey() + " initialized");
}
} catch (Throwable e) {
abort = true;
if (e instanceof OutOfMemoryError) {
throw (OutOfMemoryError) e;
} else {
throw new RuntimeException("Map operator initialization failed", e);
}
}
}
@Override
public RecordWriter getHiveRecordWriter(
JobConf jc,
Path finalOutPath,
Class<? extends Writable> valueClass,
boolean isCompressed,
Properties tbl,
Progressable progress)
throws IOException {
boolean usenewformat = jc.getBoolean("fdf.newformat", false);
IHead head = new IHead(usenewformat ? ConstVar.NewFormatFile : ConstVar.OldFormatFile);
String columnTypeProperty = tbl.getProperty(Constants.LIST_COLUMN_TYPES);
IFieldMap map = new IFieldMap();
ArrayList<TypeInfo> types;
if (columnTypeProperty == null) {
types = new ArrayList<TypeInfo>();
map.addFieldType(new IRecord.IFType(ConstVar.FieldType_Int, 0));
} else types = TypeInfoUtils.getTypeInfosFromTypeString(columnTypeProperty);
String compress = tbl.getProperty(ConstVar.Compress);
if (compress != null && compress.equalsIgnoreCase("true")) head.setCompress((byte) 1);
int i = 0;
for (TypeInfo type : types) {
byte fdftype = 0;
String name = type.getTypeName();
if (name.equals(Constants.TINYINT_TYPE_NAME)) fdftype = ConstVar.FieldType_Byte;
else if (name.equals(Constants.SMALLINT_TYPE_NAME)) fdftype = ConstVar.FieldType_Short;
else if (name.equals(Constants.INT_TYPE_NAME)) fdftype = ConstVar.FieldType_Int;
else if (name.equals(Constants.BIGINT_TYPE_NAME)) fdftype = ConstVar.FieldType_Long;
else if (name.equals(Constants.FLOAT_TYPE_NAME)) fdftype = ConstVar.FieldType_Float;
else if (name.equals(Constants.DOUBLE_TYPE_NAME)) fdftype = ConstVar.FieldType_Double;
else if (name.equals(Constants.STRING_TYPE_NAME)) fdftype = ConstVar.FieldType_String;
map.addFieldType(new IRecord.IFType(fdftype, i++));
}
head.setFieldMap(map);
ArrayList<ArrayList<Integer>> columnprojects = null;
String projectionString = jc.get(ConstVar.Projection);
if (projectionString != null) {
columnprojects = new ArrayList<ArrayList<Integer>>();
String[] projectionList = projectionString.split(ConstVar.RecordSplit);
for (String str : projectionList) {
ArrayList<Integer> cp = new ArrayList<Integer>();
String[] item = str.split(ConstVar.FieldSplit);
for (String s : item) {
cp.add(Integer.valueOf(s));
}
columnprojects.add(cp);
}
}
if (!jc.getBoolean(ConstVar.NeedPostfix, true)) {
final Configuration conf = new Configuration(jc);
final IFormatDataFile ifdf = new IFormatDataFile(conf);
ifdf.create(finalOutPath.toString(), head);
return new RecordWriter() {
@Override
public void write(Writable w) throws IOException {}
@Override
public void close(boolean abort) throws IOException {
ifdf.close();
}
};
}
final IColumnDataFile icdf = new IColumnDataFile(jc);
icdf.create(finalOutPath.toString(), head, columnprojects);
LOG.info(finalOutPath.toString());
LOG.info("output file compress?\t" + compress);
LOG.info("head:\t" + head.toStr());
return new RecordWriter() {
@Override
public void write(Writable w) throws IOException {
icdf.addRecord((IRecord) w);
}
@Override
public void close(boolean abort) throws IOException {
icdf.close();
}
};
}
public void configure(JobConf job) {
setConf(job);
fastCheck = job.getBoolean("fs.test.fastCheck", false);
}