Class Distinct

src/org/apache/pig/builtin/Distinct.java:40–161

Find the distinct set of tuples in a bag. This is a blocking operator. All the input is put in the hashset implemented in DistinctDataBag which also provides the other DataBag interfaces.

 * in DistinctDataBag which also provides the other DataBag interfaces.
 */
public class Distinct  extends EvalFunc<DataBag> implements Algebraic {

    private static TupleFactory tupleFactory = TupleFactory.getInstance();
    private static boolean initialized = false;
    private static boolean useDefaultBag = false;

    @Override
    public DataBag exec(Tuple input) throws IOException {
        return getDistinct(input);
    }

    @Override
    public String getFinal() {
        return Final.class.getName();
    }

    @Override
    public String getInitial() {
        return Initial.class.getName();
    }

    @Override
    public String getIntermed() {
        return Intermediate.class.getName();
    }

    static public class Initial extends EvalFunc<Tuple> {

        @Override
        public Tuple exec(Tuple input) throws IOException {
            // the input has  a single field which is a tuple
            // representing the data we want to distinct.
            // unwrap, put in a bag and send down
            try {
                Tuple single = (Tuple)input.get(0);
                DataBag bag = single == null ? createDataBag() : new SingleTupleBag(single);
                return tupleFactory.newTuple(bag);
            } catch (ExecException e) {
                throw e;
            }
        }
    }

    static public class Intermediate extends EvalFunc<Tuple> {

        @Override
        public Tuple exec(Tuple input) throws IOException {
            return tupleFactory.newTuple(getDistinctFromNestedBags(input, this));
        }
    }

    static public class Final extends EvalFunc<DataBag> {

        @Override
        public DataBag exec(Tuple input) throws IOException {
            return getDistinctFromNestedBags(input, this);
        }
    }