public class PigGenericMapReduce
This class is the static Mapper & Reducer classes that
are used by Pig to execute Pig Map Reduce jobs. Since
there is a reduce phase, the leaf is bound to be a
POLocalRearrange. So the map phase has to separate the
key and tuple and collect it into the output
The shuffle and sort phase sorts these keys & tuples
and creates key, List<Tuple> and passes the key and
iterator to the list. The deserialized POPackage operator
is used to package the key, List<Tuple> into pigKey,
Bag<Tuple> where pigKey is of the appropriate pig type and
then the result of the package is attached to the reduce
plan which is executed if its not empty. Either the result
of the reduce plan or the package res is collected into
the output collector.
The index of the tuple (that is, which bag it should be placed in by the
package) is packed into the key. This is done so that hadoop sorts the
keys in order of index for join.
This class is the base class for PigMapReduce, which has slightly
difference among different versions of hadoop. PigMapReduce implementation
is located in $PIG_HOME/shims.