有没有办法将重复项保存在Hive中的收集集中，或者模拟Hive使用其他方法提供的聚合集合的种类？我想将列中具有相同键的所有项聚合到一个数组中，并重复。
即：
hash_id | num_of_cats
=====================
ad3jkfk 4
ad3jkfk 4
ad3jkfk 2
fkjh43f 1
fkjh43f 8
fkjh43f 8
rjkhd93 7
rjkhd93 4
rjkhd93 7
应该返回：
hash_agg | cats_aggregate
===========================
ad3jkfk Array<int>(4,4,2)
fkjh43f Array<int>(1,8,8)
rjkhd93 Array<int>(7,4,7)
答案 0 :(得分：28)
尝试在Hive 0.13.0之后使用COLLECT_LIST（col）
SELECT
hash_id, COLLECT_LIST(num_of_cats) AS aggr_set
FROM
tablename
WHERE
blablabla
GROUP BY
hash_id
;
答案 1 :(得分：22)
内置任何内容，但创建用户定义的函数（包括聚合）并不是那么糟糕。唯一粗略的部分是尝试使它们类型为通用，但这里是一个收集示例。
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;
}
}
}
然后在hive中，只发出add jar Whatever.jar;和CREATE TEMPORARY FUNCTION collect_all AS 'com.example.CollectAll';
你应该能够按预期使用它。
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]
值得注意的是元素的顺序应该被认为是未定义的，因此如果您打算使用它来将信息提供给n_grams，您可能需要对其进行扩展以根据需要对数据进行排序。
答案 2 :(得分：12)
修改Jeff Mc的代码以删除限制（可能是从collect_set继承），输入必须是原始类型。此版本可以收集结构，地图和数组以及基元。
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;
}
}
}
答案 3 :(得分：11)
从hive 0.13开始，有一个名为collect_list()的内置UDAF可以实现这一目标。请参阅here。
答案 4 :(得分：2)
查看Brickhouse收集UDAF（http://github.com/klout/brickhouse/blob/master/src/main/java/brickhouse/udf/collect/CollectUDAF.java）
它还支持收集到地图中。 Brickhouse还包含许多有用的UDF，而不是标准的Hive发行版。
答案 5 :(得分：1)
以下是完成此工作的确切配置单元查询（仅适用于配置单元＆gt; 0.13）：
SELECT hash_id，collect_set（num_of_cats）FROM GROUP BY hash_id;
答案 6 :(得分：1)
为了它的价值（虽然我知道这是一篇较旧的帖子），Hive 0.13.0提供了一个新的 collect_list（）功能，不会进行重复数据删除。
答案 7 :(得分：0)
收集结构
的解决方法假设你有一张桌子
tableWithStruct(
id string,
obj struct <a:string,b:string>)
现在创建另一个表格为
CREATE EXTERNAL TABLE tablename (
id string,
temp array<string>
)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY '\t' COLLECTION ITEMS TERMINATED BY ',' MAP KEYS TERMINATED BY '|'
插入查询
insert into table tablename select id,collect(concat_ws('|',cast(obj.a as string),cast(obj.b as string)) from tableWithStruct group by id;
现在在与 tablename
相同的位置创建另一个表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 '|'
当您从 tablename_final 中选择时，您将获得所需的输出
答案 8 :(得分：-1)
只是想知道 - 如果是状态 -
SELECT
hash_id, COLLECT_LIST(num_of_cats) AS aggr_set
FROM
tablename
WHERE
blablabla
GROUP BY
hash_id
;
我们想要排序并限制num_of_cats的元素 - 如何去做？大数据中的COZ我们处理数据的PB ..在这种情况下我们可能不需要所有这些但是前10或限制它。