saving json data in hdfs in hadoop

Question

I have the following Reducer class

public static class TokenCounterReducer extends Reducer<Text, Text, Text, Text> {
    public void reduce(Text key, Iterable<Text> values, Context context)
            throws IOException, InterruptedException {

        JSONObject jsn = new JSONObject();

        for (Text value : values) {
            String[] vals = value.toString().split("\t");
            String[] targetNodes = vals[0].toString().split(",",-1);
            jsn.put("source",vals[1] );
            jsn.put("target",targetNodes);

        }
        // context.write(key, new Text(sum));
    }
}

Going thru examples (disclaimer: newbie here), I can see that the general output type seems to be like a key/value store.

But what if I dont have any key in the output? or what if I want if my output is in some other format (json in my case)?

Anyways, from the above code: I want to write json object to HDFS?

It was very trivial in Hadoop streaming.. but how do i do it in Hadoop java?

Answer 1

If you just want to write a list of JSON objects to HDFS without caring about the notion of key/value, you could just use a NullWritable in your Reducer output value:

public static class TokenCounterReducer extends Reducer<Text, Text, Text, NullWritable> {
    public void reduce(Text key, Iterable<Text> values, Context context)
            throws IOException, InterruptedException {
        for (Text value : values) {
            JSONObject jsn = new JSONObject();
            ....
            context.write(new Text(jsn.toString()), null);
        }
    }
}

Note that you will need to change your job configuration to do:

job.setOutputValueClass(NullWritable.class);

By writing your JSON object to HDFS I understood that you want to store a String representation of your JSON which I'm describing above. If you wanted to store a binary representation of your JSON into HDFS you would need to use a SequenceFile. Obviously you could write your own Writable for this but I feel it's just easier like this if you intend to have a simple String representation.

Answer 2

You can use Hadoop's OutputFormat interfaces to create your custom formats which will write the data as per your wish. For instance if you need data to be written as a JSON object then you could do this :

public class JsonOutputFormat extends TextOutputFormat<Text, IntWritable> {
    @Override
    public RecordWriter<Text, IntWritable> getRecordWriter(
            TaskAttemptContext context) throws IOException, 
                  InterruptedException {
        Configuration conf = context.getConfiguration();
        Path path = getOutputPath(context);
        FileSystem fs = path.getFileSystem(conf);
        FSDataOutputStream out = 
                fs.create(new Path(path,context.getJobName()));
        return new JsonRecordWriter(out);
    }

    private static class JsonRecordWriter extends 
          LineRecordWriter<Text,IntWritable>{
        boolean firstRecord = true;
        @Override
        public synchronized void close(TaskAttemptContext context)
                throws IOException {
            out.writeChar('{');
            super.close(null);
        }

        @Override
        public synchronized void write(Text key, IntWritable value)
                throws IOException {
            if (!firstRecord){
                out.writeChars(",\r\n");
                firstRecord = false;
            }
            out.writeChars("\"" + key.toString() + "\":\""+
                    value.toString()+"\"");
        }

        public JsonRecordWriter(DataOutputStream out) 
                throws IOException{
            super(out);
            out.writeChar('}');
        }
    }
}

And if you do not want to have the key in your output just emit null, like :

context.write(NullWritable.get(), new IntWritable(sum));

HTH