COLLECT_SET() in Hive, keep duplicates?

问题:

Is there a way to keep the duplicates in a collected set in Hive, or simulate the sort of aggregate collection that Hive provides using some other method? I want to aggregate all of the items in a column that have the same key into an array, with duplicates.

I.E.:

hash_id | num_of_cats
=====================
ad3jkfk            4
ad3jkfk            4
ad3jkfk            2
fkjh43f            1
fkjh43f            8
fkjh43f            8
rjkhd93            7
rjkhd93            4
rjkhd93            7

should return:

hash_agg | cats_aggregate
===========================
ad3jkfk   Array<int>(4,4,2)
fkjh43f   Array<int>(1,8,8)
rjkhd93   Array<int>(7,4,7)

回答1:

Try to use COLLECT_LIST(col) after Hive 0.13.0

SELECT
    hash_id, COLLECT_LIST(num_of_cats) AS aggr_set
FROM
    tablename
WHERE
    blablabla
GROUP BY
    hash_id
;

回答2:

There is nothing built in, but creating user defined functions, including aggregates, isn't that bad. The only rough part is trying to make them type generic, but here is a collect example.

package com.example;

import java.util.ArrayList;
import org.apache.hadoop.hive.ql.exec.UDFArgumentTypeException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.parse.SemanticException;
import org.apache.hadoop.hive.ql.udf.generic.AbstractGenericUDAFResolver;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDAFEvaluator;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils;
import org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.StandardListObjectInspector;
import org.apache.hadoop.hive.serde2.typeinfo.TypeInfo;

public class CollectAll extends AbstractGenericUDAFResolver
{
    @Override
    public GenericUDAFEvaluator getEvaluator(TypeInfo[] tis)
            throws SemanticException
    {
        if (tis.length != 1)
        {
            throw new UDFArgumentTypeException(tis.length - 1, "Exactly one argument is expected.");
        }
        if (tis[0].getCategory() != ObjectInspector.Category.PRIMITIVE)
        {
            throw new UDFArgumentTypeException(0, "Only primitive type arguments are accepted but " + tis[0].getTypeName() + " was passed as parameter 1.");
        }
        return new CollectAllEvaluator();
    }

    public static class CollectAllEvaluator extends GenericUDAFEvaluator
    {
        private PrimitiveObjectInspector inputOI;
        private StandardListObjectInspector loi;
        private StandardListObjectInspector internalMergeOI;

        @Override
        public ObjectInspector init(Mode m, ObjectInspector[] parameters)
                throws HiveException
        {
            super.init(m, parameters);
            if (m == Mode.PARTIAL1)
            {
                inputOI = (PrimitiveObjectInspector) parameters[0];
                return ObjectInspectorFactory
                        .getStandardListObjectInspector((PrimitiveObjectInspector) ObjectInspectorUtils
                        .getStandardObjectInspector(inputOI));
            }
            else
            {
                if (!(parameters[0] instanceof StandardListObjectInspector))
                {
                    inputOI = (PrimitiveObjectInspector)  ObjectInspectorUtils
                            .getStandardObjectInspector(parameters[0]);
                    return (StandardListObjectInspector) ObjectInspectorFactory
                            .getStandardListObjectInspector(inputOI);
                }
                else
                {
                    internalMergeOI = (StandardListObjectInspector) parameters[0];
                    inputOI = (PrimitiveObjectInspector) internalMergeOI.getListElementObjectInspector();
                    loi = (StandardListObjectInspector) ObjectInspectorUtils.getStandardObjectInspector(internalMergeOI);
                    return loi;
                }
            }
        }

        static class ArrayAggregationBuffer implements AggregationBuffer
        {
            ArrayList<Object> container;
        }

        @Override
        public void reset(AggregationBuffer ab)
                throws HiveException
        {
            ((ArrayAggregationBuffer) ab).container = new ArrayList<Object>();
        }

        @Override
        public AggregationBuffer getNewAggregationBuffer()
                throws HiveException
        {
            ArrayAggregationBuffer ret = new ArrayAggregationBuffer();
            reset(ret);
            return ret;
        }

        @Override
        public void iterate(AggregationBuffer ab, Object[] parameters)
                throws HiveException
        {
            assert (parameters.length == 1);
            Object p = parameters[0];
            if (p != null)
            {
                ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
                agg.container.add(ObjectInspectorUtils.copyToStandardObject(p, this.inputOI));
            }
        }

        @Override
        public Object terminatePartial(AggregationBuffer ab)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> ret = new ArrayList<Object>(agg.container.size());
            ret.addAll(agg.container);
            return ret;
        }

        @Override
        public void merge(AggregationBuffer ab, Object o)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> partial = (ArrayList<Object>)internalMergeOI.getList(o);
            for(Object i : partial)
            {
                agg.container.add(ObjectInspectorUtils.copyToStandardObject(i, this.inputOI));
            }
        }

        @Override
        public Object terminate(AggregationBuffer ab)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> ret = new ArrayList<Object>(agg.container.size());
            ret.addAll(agg.container);
            return ret;
        }
    }
}

Then in hive, just issue add jar Whatever.jar; and CREATE TEMPORARY FUNCTION collect_all AS 'com.example.CollectAll'; You should them be able to use it as expected.

hive> SELECT hash_id, collect_all(num_of_cats) FROM test GROUP BY hash_id;
OK
ad3jkfk [4,4,2]
fkjh43f [1,8,8]
rjkhd93 [7,4,7]

It's worth noting that the order of the elements should be considered undefined, so if you intend to use this to feed information into n_grams you might need to expand it a bit to sort the data as needed.

回答3:

Modified Jeff Mc's code to remove the restriction (presumably inherited from collect_set) that input must be primitive types. This version can collect structs, maps and arrays as well as primitives.

package com.example;

import java.util.ArrayList;
import org.apache.hadoop.hive.ql.exec.UDFArgumentTypeException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.parse.SemanticException;
import org.apache.hadoop.hive.ql.udf.generic.AbstractGenericUDAFResolver;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDAFEvaluator;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils;
import org.apache.hadoop.hive.serde2.objectinspector.StandardListObjectInspector;
import org.apache.hadoop.hive.serde2.typeinfo.TypeInfo;

public class CollectAll extends AbstractGenericUDAFResolver
{
    @Override
    public GenericUDAFEvaluator getEvaluator(TypeInfo[] tis)
            throws SemanticException
    {
        if (tis.length != 1)
        {
            throw new UDFArgumentTypeException(tis.length - 1, "Exactly one argument is expected.");
        }
        return new CollectAllEvaluator();
    }

    public static class CollectAllEvaluator extends GenericUDAFEvaluator
    {
        private ObjectInspector inputOI;
        private StandardListObjectInspector loi;
        private StandardListObjectInspector internalMergeOI;

        @Override
        public ObjectInspector init(Mode m, ObjectInspector[] parameters)
                throws HiveException
        {
            super.init(m, parameters);
            if (m == Mode.PARTIAL1)
            {
                inputOI = parameters[0];
                return ObjectInspectorFactory
                        .getStandardListObjectInspector(ObjectInspectorUtils
                        .getStandardObjectInspector(inputOI));
            }
            else
            {
                if (!(parameters[0] instanceof StandardListObjectInspector))
                {
                    inputOI = ObjectInspectorUtils
                            .getStandardObjectInspector(parameters[0]);
                    return (StandardListObjectInspector) ObjectInspectorFactory
                            .getStandardListObjectInspector(inputOI);
                }
                else
                {
                    internalMergeOI = (StandardListObjectInspector) parameters[0];
                    inputOI = internalMergeOI.getListElementObjectInspector();
                    loi = (StandardListObjectInspector) ObjectInspectorUtils.getStandardObjectInspector(internalMergeOI);
                    return loi;
                }
            }
        }

        static class ArrayAggregationBuffer implements AggregationBuffer
        {
            ArrayList<Object> container;
        }

        @Override
        public void reset(AggregationBuffer ab)
                throws HiveException
        {
            ((ArrayAggregationBuffer) ab).container = new ArrayList<Object>();
        }

        @Override
        public AggregationBuffer getNewAggregationBuffer()
                throws HiveException
        {
            ArrayAggregationBuffer ret = new ArrayAggregationBuffer();
            reset(ret);
            return ret;
        }

        @Override
        public void iterate(AggregationBuffer ab, Object[] parameters)
                throws HiveException
        {
            assert (parameters.length == 1);
            Object p = parameters[0];
            if (p != null)
            {
                ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
                agg.container.add(ObjectInspectorUtils.copyToStandardObject(p, this.inputOI));
            }
        }

        @Override
        public Object terminatePartial(AggregationBuffer ab)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> ret = new ArrayList<Object>(agg.container.size());
            ret.addAll(agg.container);
            return ret;
        }

        @Override
        public void merge(AggregationBuffer ab, Object o)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> partial = (ArrayList<Object>)internalMergeOI.getList(o);
            for(Object i : partial)
            {
                agg.container.add(ObjectInspectorUtils.copyToStandardObject(i, this.inputOI));
            }
        }

        @Override
        public Object terminate(AggregationBuffer ab)
                throws HiveException
        {
            ArrayAggregationBuffer agg = (ArrayAggregationBuffer) ab;
            ArrayList<Object> ret = new ArrayList<Object>(agg.container.size());
            ret.addAll(agg.container);
            return ret;
        }
    }
}

回答4:

As of hive 0.13, there is a built-in UDAF called collect_list() that achieves this. See here.

回答5:

Check out the Brickhouse collect UDAF ( http://github.com/klout/brickhouse/blob/master/src/main/java/brickhouse/udf/collect/CollectUDAF.java )

It also supports collecting into a map. Brickhouse also contains many useful UDF's not in the standard Hive distribution.

回答6:

Here is the exact hive query that does this job (works only in hive > 0.13):

SELECT hash_id, collect_set( num_of_cats) FROM GROUP BY hash_id;

回答7:

For what it's worth (though I know this is an older post), Hive 0.13.0 features a new collect_list() function that does not deduplicate.

回答8:

Workaround to collect struct

suppose you have a table

tableWithStruct(
id string,
obj struct <a:string,b:string>)

now create another table as

CREATE EXTERNAL TABLE tablename (
id string,
temp array<string>
)
ROW FORMAT DELIMITED 
FIELDS TERMINATED BY '\t' COLLECTION ITEMS TERMINATED BY ',' MAP KEYS TERMINATED BY '|'

insert query

insert into table tablename select id,collect(concat_ws('|',cast(obj.a as string),cast(obj.b as string)) from tableWithStruct group by id;

now create another table at same location as tablename

CREATE EXTERNAL TABLE tablename_final (
id string,
array_list array<struct<a:string,b:string>>
)
ROW FORMAT DELIMITED 
FIELDS TERMINATED BY '\t' COLLECTION ITEMS TERMINATED BY ',' MAP KEYS TERMINATED BY '|'

when you select from tablename_final you will get the desired output

回答9:

Just wondering - if n the statemnent -

SELECT
    hash_id, COLLECT_LIST(num_of_cats) AS aggr_set
FROM
    tablename
WHERE
    blablabla
GROUP BY
    hash_id
;

we want to have sorting and limit the elements for num_of_cats - how to go about it ? COZ in big data we deal with PBs of datas.. we might not need all of that in such cases but top 10 or limit it .