/**
* writes the current row operation as a Put, for use with TableOutputFormat.
*
* @param context
* @throws IOException
* @throws InterruptedException
*/
public void put(
TaskInputOutputContext<? extends Object, ? extends Object, ImmutableBytesWritable, Mutation>
context)
throws IOException, InterruptedException {
Put put = createPut();
context.write(tableKey, put);
values = new ArrayMap3();
}
@SuppressWarnings("unchecked")
public void write(String namedOutput, K key, V value) throws IOException, InterruptedException {
if (!namedOutputs.containsKey(namedOutput)) {
throw new IllegalArgumentException("Undefined named output '" + namedOutput + "'");
}
if (!disableOutputCounters) {
baseContext.getCounter(COUNTERS_GROUP, namedOutput).increment(1);
}
getOutputState(namedOutput).write(key, value);
}
public void putBulkLoadPut(
TaskInputOutputContext<? extends Object, ? extends Object, ImmutableBytesWritable, Put>
context)
throws IOException, InterruptedException {
// if (partitioner == null) {
// partitioner = new TotalOrderPartitioner();
// partitioner.setConf(context.getConfiguration());
// }
ImmutableBytesWritable key = new ImmutableBytesWritable(rowkey);
Put put = createPut();
context.write(key, put);
values = new ArrayMap3();
}
private static void write(
TaskInputOutputContext<?, ?, BytesWritable, Text> context,
final byte[] groupKey,
DimValueCount dimValueCount)
throws IOException, InterruptedException {
context.write(
new SortableBytes(
groupKey,
TAB_JOINER
.join(dimValueCount.dim, dimValueCount.value)
.getBytes(HadoopDruidIndexerConfig.JAVA_NATIVE_CHARSET))
.toBytesWritable(),
dimValueCount.toText());
}
private OutputState<K, V> getOutputState(String namedOutput)
throws IOException, InterruptedException {
OutputState<?, ?> out = outputStates.get(namedOutput);
if (out != null) {
return (OutputState<K, V>) out;
}
// The following trick leverages the instantiation of a record writer via
// the job thus supporting arbitrary output formats.
Job job = getJob(baseContext.getJobID(), namedOutput, baseConf);
// Get a job with the expected named output.
job = getJob(job.getJobID(), namedOutput, baseConf);
OutputFormat<K, V> fmt = getOutputFormat(namedOutput, job, namedOutputs.get(namedOutput));
TaskAttemptContext taskContext = getTaskContext(baseContext, job);
RecordWriter<K, V> recordWriter = fmt.getRecordWriter(taskContext);
OutputState<K, V> outputState = new OutputState(taskContext, recordWriter);
this.outputStates.put(namedOutput, outputState);
return outputState;
}
/**
* Creates and initializes multiple outputs support, it should be instantiated in the
* Mapper/Reducer setup method.
*
* @param context the TaskInputOutputContext object
*/
public CrunchOutputs(TaskInputOutputContext<?, ?, K, V> context) {
this(context.getConfiguration());
this.baseContext = context;
}
private static Map<String, OutputConfig> getNamedOutputs(
TaskInputOutputContext<?, ?, ?, ?> context) {
return getNamedOutputs(context.getConfiguration());
}