Examples with AvroWrapper org.apache.avro.mapred.AvroWrapper used on opensource projects

Example 1 with AvroWrapper

use of org.apache.avro.mapred.AvroWrapper in project trevni by cutting.

the class AvroTrevniInputFormat method getRecordReader.

@Override
public RecordReader<AvroWrapper<T>, NullWritable> getRecordReader(InputSplit split, final JobConf job, Reporter reporter) throws IOException {
    final FileSplit file = (FileSplit) split;
    reporter.setStatus(file.toString());
    final AvroColumnReader.Params params = new AvroColumnReader.Params(new HadoopInput(file.getPath(), job));
    params.setModel(ReflectData.get());
    if (job.get(AvroJob.INPUT_SCHEMA) != null)
        params.setSchema(AvroJob.getInputSchema(job));
    return new RecordReader<AvroWrapper<T>, NullWritable>() {

        private AvroColumnReader<T> reader = new AvroColumnReader<T>(params);

        private float rows = reader.getRowCount();

        private long row;

        public AvroWrapper<T> createKey() {
            return new AvroWrapper<T>(null);
        }

        public NullWritable createValue() {
            return NullWritable.get();
        }

        public boolean next(AvroWrapper<T> wrapper, NullWritable ignore) throws IOException {
            if (!reader.hasNext())
                return false;
            wrapper.datum(reader.next());
            row++;
            return true;
        }

        public float getProgress() throws IOException {
            return row / rows;
        }

        public long getPos() throws IOException {
            return row;
        }

        public void close() throws IOException {
            reader.close();
        }
    };
}

Example 2 with AvroWrapper

use of org.apache.avro.mapred.AvroWrapper in project trevni by cutting.

the class AvroTrevniOutputFormat method getRecordWriter.

@Override
public RecordWriter<AvroWrapper<T>, NullWritable> getRecordWriter(FileSystem ignore, final JobConf job, final String name, Progressable prog) throws IOException {
    boolean isMapOnly = job.getNumReduceTasks() == 0;
    final Schema schema = isMapOnly ? AvroJob.getMapOutputSchema(job) : AvroJob.getOutputSchema(job);
    final ColumnFileMetaData meta = new ColumnFileMetaData();
    for (Map.Entry<String, String> e : job) if (e.getKey().startsWith(META_PREFIX))
        meta.put(e.getKey().substring(AvroJob.TEXT_PREFIX.length()), e.getValue().getBytes(MetaData.UTF8));
    final Path dir = FileOutputFormat.getTaskOutputPath(job, name);
    final FileSystem fs = dir.getFileSystem(job);
    if (!fs.mkdirs(dir))
        throw new IOException("Failed to create directory: " + dir);
    final long blockSize = fs.getDefaultBlockSize();
    return new RecordWriter<AvroWrapper<T>, NullWritable>() {

        private int part = 0;

        private AvroColumnWriter<T> writer = new AvroColumnWriter<T>(schema, meta, ReflectData.get());

        private void flush() throws IOException {
            OutputStream out = fs.create(new Path(dir, "part-" + (part++) + EXT));
            try {
                writer.writeTo(out);
            } finally {
                out.close();
            }
            writer = new AvroColumnWriter<T>(schema, meta, ReflectData.get());
        }

        public void write(AvroWrapper<T> wrapper, NullWritable ignore) throws IOException {
            writer.write(wrapper.datum());
            if (// block full
            writer.sizeEstimate() >= blockSize)
                flush();
        }

        public void close(Reporter reporter) throws IOException {
            flush();
        }
    };
}

Example 3 with AvroWrapper

use of org.apache.avro.mapred.AvroWrapper in project crunch by cloudera.

the class AvroOutputFormat method getRecordWriter.

@Override
public RecordWriter<AvroWrapper<T>, NullWritable> getRecordWriter(TaskAttemptContext context) throws IOException, InterruptedException {
    Configuration conf = context.getConfiguration();
    Schema schema = null;
    String outputName = conf.get("crunch.namedoutput");
    if (outputName != null && !outputName.isEmpty()) {
        schema = (new Schema.Parser()).parse(conf.get("avro.output.schema." + outputName));
    } else {
        schema = AvroJob.getOutputSchema(context.getConfiguration());
    }
    ReflectDataFactory factory = Avros.getReflectDataFactory(conf);
    final DataFileWriter<T> WRITER = new DataFileWriter<T>(factory.<T>getWriter());
    Path path = getDefaultWorkFile(context, org.apache.avro.mapred.AvroOutputFormat.EXT);
    WRITER.create(schema, path.getFileSystem(context.getConfiguration()).create(path));
    return new RecordWriter<AvroWrapper<T>, NullWritable>() {

        @Override
        public void write(AvroWrapper<T> wrapper, NullWritable ignore) throws IOException {
            WRITER.append(wrapper.datum());
        }

        @Override
        public void close(TaskAttemptContext context) throws IOException, InterruptedException {
            WRITER.close();
        }
    };
}