HBase聚合

Question

I'm having some trouble doing aggregation on a particular column in HBase.

This is the snippet of code I tried:

 Configuration config = HBaseConfiguration.create();
 AggregationClient aggregationClient = new AggregationClient(config);

 Scan scan = new Scan();
 scan.addColumn(Bytes.toBytes("drs"), Bytes.toBytes("count"));

 ColumnInterpreter<Long, Long> ci = new LongColumnInterpreter();

 Long sum = aggregationClient.sum(Bytes.toBytes("DEMO_CALCULATIONS"), ci , scan);
 System.out.println(sum);

sum returns a value of null. The aggregationClient API works fine if I do a rowcount.

I was trying to follow the directions in http://michaelmorello.blogspot.in/2012/01/row-count-hbase-aggregation-example.html

Could there be a problem with me using a LongColumnInterpreter when the 'count' field was an int? What am I missing in here?

Answer 1

You can only use long(8bytes) to do sum with default setting.

Cause in the code of AggregateImplementation's getSum method, it handle all the returned KeyValue as long.

List<KeyValue> results = new ArrayList<KeyValue>();
try {
  boolean hasMoreRows = false;
  do {
    hasMoreRows = scanner.next(results);
    for (KeyValue kv : results) {
      temp = ci.getValue(colFamily, qualifier, kv);
      if (temp != null)
        sumVal = ci.add(sumVal, ci.castToReturnType(temp));
    }
    results.clear();
  } while (hasMoreRows);
} finally {
  scanner.close();
}

and in LongColumnInterpreter

public Long getValue(byte[] colFamily, byte[] colQualifier, KeyValue kv)
  throws IOException {
if (kv == null || kv.getValueLength() != Bytes.SIZEOF_LONG)
  return null;
return Bytes.toLong(kv.getBuffer(), kv.getValueOffset());
}