Examples with RecordReader org.apache.hadoop.mapred.RecordReader used on opensource projects

Example 16 with RecordReader

use of org.apache.hadoop.mapred.RecordReader in project voldemort by voldemort.

the class JsonSequenceFileInputFormat method getRecordReader.

@Override
public RecordReader<BytesWritable, BytesWritable> getRecordReader(InputSplit split, JobConf conf, Reporter reporter) throws IOException {
    String inputPathString = ((FileSplit) split).getPath().toUri().getPath();
    log.info("Input file path:" + inputPathString);
    Path inputPath = new Path(inputPathString);
    SequenceFile.Reader reader = new SequenceFile.Reader(inputPath.getFileSystem(conf), inputPath, conf);
    SequenceFile.Metadata meta = reader.getMetadata();
    try {
        Text keySchema = meta.get(new Text("key.schema"));
        Text valueSchema = meta.get(new Text("value.schema"));
        if (0 == keySchema.getLength() || 0 == valueSchema.getLength()) {
            throw new Exception();
        }
        // update Joboconf with schemas
        conf.set("mapper.input.key.schema", keySchema.toString());
        conf.set("mapper.input.value.schema", valueSchema.toString());
    } catch (Exception e) {
        throw new IOException("Failed to Load Schema from file:" + inputPathString + "\n");
    }
    return super.getRecordReader(split, conf, reporter);
}

Example 17 with RecordReader

use of org.apache.hadoop.mapred.RecordReader in project ignite by apache.

the class HadoopV1MapTask method run.

/** {@inheritDoc} */
@SuppressWarnings("unchecked")
@Override
public void run(HadoopTaskContext taskCtx) throws IgniteCheckedException {
    HadoopJobEx job = taskCtx.job();
    HadoopV2TaskContext taskCtx0 = (HadoopV2TaskContext) taskCtx;
    if (taskCtx.taskInfo().hasMapperIndex())
        HadoopMapperUtils.mapperIndex(taskCtx.taskInfo().mapperIndex());
    else
        HadoopMapperUtils.clearMapperIndex();
    try {
        JobConf jobConf = taskCtx0.jobConf();
        InputFormat inFormat = jobConf.getInputFormat();
        HadoopInputSplit split = info().inputSplit();
        InputSplit nativeSplit;
        if (split instanceof HadoopFileBlock) {
            HadoopFileBlock block = (HadoopFileBlock) split;
            nativeSplit = new FileSplit(new Path(block.file().toString()), block.start(), block.length(), EMPTY_HOSTS);
        } else
            nativeSplit = (InputSplit) taskCtx0.getNativeSplit(split);
        assert nativeSplit != null;
        Reporter reporter = new HadoopV1Reporter(taskCtx);
        HadoopV1OutputCollector collector = null;
        try {
            collector = collector(jobConf, taskCtx0, !job.info().hasCombiner() && !job.info().hasReducer(), fileName(), taskCtx0.attemptId());
            RecordReader reader = inFormat.getRecordReader(nativeSplit, jobConf, reporter);
            Mapper mapper = ReflectionUtils.newInstance(jobConf.getMapperClass(), jobConf);
            Object key = reader.createKey();
            Object val = reader.createValue();
            assert mapper != null;
            try {
                try {
                    while (reader.next(key, val)) {
                        if (isCancelled())
                            throw new HadoopTaskCancelledException("Map task cancelled.");
                        mapper.map(key, val, collector, reporter);
                    }
                    taskCtx.onMapperFinished();
                } finally {
                    mapper.close();
                }
            } finally {
                collector.closeWriter();
            }
            collector.commit();
        } catch (Exception e) {
            if (collector != null)
                collector.abort();
            throw new IgniteCheckedException(e);
        }
    } finally {
        HadoopMapperUtils.clearMapperIndex();
    }
}