Examples with OutputFormat org.apache.hadoop.mapreduce.OutputFormat used on opensource projects

Example 6 with OutputFormat

use of org.apache.hadoop.mapreduce.OutputFormat in project hadoop by apache.

the class TestRecovery method writeBadOutput.

private void writeBadOutput(TaskAttempt attempt, Configuration conf) throws Exception {
    TaskAttemptContext tContext = new TaskAttemptContextImpl(conf, TypeConverter.fromYarn(attempt.getID()));
    TextOutputFormat<?, ?> theOutputFormat = new TextOutputFormat();
    RecordWriter theRecordWriter = theOutputFormat.getRecordWriter(tContext);
    NullWritable nullWritable = NullWritable.get();
    try {
        theRecordWriter.write(key2, val2);
        theRecordWriter.write(null, nullWritable);
        theRecordWriter.write(null, val2);
        theRecordWriter.write(nullWritable, val1);
        theRecordWriter.write(key1, nullWritable);
        theRecordWriter.write(key2, null);
        theRecordWriter.write(null, null);
        theRecordWriter.write(key1, val1);
    } finally {
        theRecordWriter.close(tContext);
    }
    OutputFormat outputFormat = ReflectionUtils.newInstance(tContext.getOutputFormatClass(), conf);
    OutputCommitter committer = outputFormat.getOutputCommitter(tContext);
    committer.commitTask(tContext);
}

Example 7 with OutputFormat

use of org.apache.hadoop.mapreduce.OutputFormat in project ignite by apache.

the class HadoopV2MapTask method run0.

/** {@inheritDoc} */
@SuppressWarnings({ "ConstantConditions", "unchecked" })
@Override
public void run0(HadoopV2TaskContext taskCtx) throws IgniteCheckedException {
    OutputFormat outputFormat = null;
    Exception err = null;
    JobContextImpl jobCtx = taskCtx.jobContext();
    if (taskCtx.taskInfo().hasMapperIndex())
        HadoopMapperUtils.mapperIndex(taskCtx.taskInfo().mapperIndex());
    else
        HadoopMapperUtils.clearMapperIndex();
    try {
        HadoopV2Context hadoopCtx = hadoopContext();
        InputSplit nativeSplit = hadoopCtx.getInputSplit();
        if (nativeSplit == null)
            throw new IgniteCheckedException("Input split cannot be null.");
        InputFormat inFormat = ReflectionUtils.newInstance(jobCtx.getInputFormatClass(), hadoopCtx.getConfiguration());
        RecordReader reader = inFormat.createRecordReader(nativeSplit, hadoopCtx);
        reader.initialize(nativeSplit, hadoopCtx);
        hadoopCtx.reader(reader);
        HadoopJobInfo jobInfo = taskCtx.job().info();
        outputFormat = jobInfo.hasCombiner() || jobInfo.hasReducer() ? null : prepareWriter(jobCtx);
        Mapper mapper = ReflectionUtils.newInstance(jobCtx.getMapperClass(), hadoopCtx.getConfiguration());
        try {
            mapper.run(new WrappedMapper().getMapContext(hadoopCtx));
            taskCtx.onMapperFinished();
        } finally {
            closeWriter();
        }
        commit(outputFormat);
    } catch (InterruptedException e) {
        err = e;
        Thread.currentThread().interrupt();
        throw new IgniteInterruptedCheckedException(e);
    } catch (Exception e) {
        err = e;
        throw new IgniteCheckedException(e);
    } finally {
        HadoopMapperUtils.clearMapperIndex();
        if (err != null)
            abort(outputFormat);
    }
}

Example 8 with OutputFormat

use of org.apache.hadoop.mapreduce.OutputFormat in project ignite by apache.

the class HadoopV2ReduceTask method run0.

/** {@inheritDoc} */
@SuppressWarnings({ "ConstantConditions", "unchecked" })
@Override
public void run0(HadoopV2TaskContext taskCtx) throws IgniteCheckedException {
    OutputFormat outputFormat = null;
    Exception err = null;
    JobContextImpl jobCtx = taskCtx.jobContext();
    // Set mapper index for combiner tasks
    if (!reduce && taskCtx.taskInfo().hasMapperIndex())
        HadoopMapperUtils.mapperIndex(taskCtx.taskInfo().mapperIndex());
    else
        HadoopMapperUtils.clearMapperIndex();
    try {
        outputFormat = reduce || !taskCtx.job().info().hasReducer() ? prepareWriter(jobCtx) : null;
        Reducer reducer;
        if (reduce)
            reducer = ReflectionUtils.newInstance(jobCtx.getReducerClass(), jobCtx.getConfiguration());
        else
            reducer = ReflectionUtils.newInstance(jobCtx.getCombinerClass(), jobCtx.getConfiguration());
        try {
            reducer.run(new WrappedReducer().getReducerContext(hadoopContext()));
            if (!reduce)
                taskCtx.onMapperFinished();
        } finally {
            closeWriter();
        }
        commit(outputFormat);
    } catch (InterruptedException e) {
        err = e;
        Thread.currentThread().interrupt();
        throw new IgniteInterruptedCheckedException(e);
    } catch (Exception e) {
        err = e;
        throw new IgniteCheckedException(e);
    } finally {
        if (!reduce)
            HadoopMapperUtils.clearMapperIndex();
        if (err != null)
            abort(outputFormat);
    }
}

Example 9 with OutputFormat

use of org.apache.hadoop.mapreduce.OutputFormat in project ignite by apache.

the class HadoopV2CleanupTask method run0.

/** {@inheritDoc} */
@SuppressWarnings("ConstantConditions")
@Override
public void run0(HadoopV2TaskContext taskCtx) throws IgniteCheckedException {
    JobContextImpl jobCtx = taskCtx.jobContext();
    try {
        OutputFormat outputFormat = getOutputFormat(jobCtx);
        OutputCommitter committer = outputFormat.getOutputCommitter(hadoopContext());
        if (committer != null) {
            if (abort)
                committer.abortJob(jobCtx, JobStatus.State.FAILED);
            else
                committer.commitJob(jobCtx);
        }
    } catch (ClassNotFoundException | IOException e) {
        throw new IgniteCheckedException(e);
    } catch (InterruptedException e) {
        Thread.currentThread().interrupt();
        throw new IgniteInterruptedCheckedException(e);
    }
}