在Hive中使用侧视图时出现异常

时间:2017-01-05 09:40:28

标签: xml hadoop xpath hive brickhouse

我使用下面的代码来解析Hive中的xml数据。在我的xml数据中,一些标签正在重复,所以我使用brickhouse jar和侧视图来解析标签并放置在Hive表中。但是当我执行我的代码时,我收到了一个错误。请帮忙,因为我无法理解我做错了什么。

代码:

add jar /home/cloudera/brickhouse-0.5.5.jar;
CREATE TEMPORARY FUNCTION numeric_range AS 'brickhouse.udf.collect.NumericRange';
CREATE TEMPORARY FUNCTION array_index AS 'brickhouse.udf.collect.ArrayIndexUDF';
add jar /home/cloudera/hivexmlserde-1.0.5.3.jar;
set hive.exec.mode.local.auto=false;
DROP TABLE IF EXISTS medinfo2;
create table medinfo2 as
select array_index(statusCode,n) AS statusCode,
    array_index(startTime,n) AS startTime,
    array_index(endTime,n) AS endTime,
    array_index(strengthValue,n) AS strengthValue,
    array_index(strengthUnits,n) AS strengthUnits
from medications_info7 lateral view numeric_range(size( statusCode )) n1 as n;

错误:

  

引起:java.lang.IndexOutOfBoundsException:索引:7,大小:7           at java.util.ArrayList.rangeCheck(ArrayList.java:635)           at java.util.ArrayList.get(ArrayList.java:411)           at com.ibm.spss.hive.serde2.xml.objectinspector.XmlListObjectInspector.getListElement(XmlListObjectInspector.java:79)           at brickhouse.udf.collect.ArrayIndexUDF.evaluate(ArrayIndexUDF.java:59)           at org.apache.hadoop.hive.ql.exec.ExprNodeGenericFuncEvaluator._evaluate(ExprNodeGenericFuncEvaluator.java:186)           在org.apache.hadoop.hive.ql.exec.ExprNodeEvaluator.evaluate(ExprNodeEvaluator.java:77)           at org.apache.hadoop.hive.ql.exec.ExprNodeEvaluatorHead._evaluate(ExprNodeEvaluatorHead.java:44)           在org.apache.hadoop.hive.ql.exec.ExprNodeEvaluator.evaluate(ExprNodeEvaluator.java:77)           at org.apache.hadoop.hive.ql.exec.ExprNodeEvaluator.evaluate(ExprNodeEvaluator.java:65)           在org.apache.hadoop.hive.ql.exec.SelectOperator.processOp(SelectOperator.java:77)           ......还有25个

     

失败:执行错误,从org.apache.hadoop.hive.ql.exec.mr.MapRedTask返回代码2       MapReduce工作推出:       Stage-Stage-1:Map:1 HDFS读:0 HDFS写:0 FAIL       总MapReduce CPU使用时间:0毫秒

示例:

<document>
 <code>10160-0</code>
 <entryInfo> 
    <statusCode>completed</statusCode>
    <startTime>20110729</startTime>
    <endTime>20110822</endTime>
    <strengthValue>24</strengthValue>
    <strengthUnits>h</strengthUnits>
 </entryInfo> 
 <entryInfo>
    <statusCode>completed</statusCode>
    <startTime>20120130</startTime>
    <endTime>20120326</endTime>
    <strengthValue>12</strengthValue>
    <strengthUnits>h</strengthUnits>
 </entryInfo>
 <entryInfo>
    <statusCode>completed</statusCode>
    <startTime>20100412</startTime>
    <endTime>20110822</endTime>
    <strengthValue>8</strengthValue>
    <strengthUnits>d</strengthUnits>
 </entryInfo>  
</document>

我的实际样本数量庞大,并且包含大量重复的标记。

1 个答案:

答案 0 :(得分:1)

我不知道你的数据在Hive中是什么样的,因为你没有提供这些信息,所以这就是我如何将你的XML加载到Hive中。

<强>装载机

ADD JAR /path/to/jar/hivexmlserde-1.0.5.3.jar;

DROP TABLE IF EXISTS db.tbl;
CREATE TABLE IF NOT EXISTS db.tbl (
  code STRING,
  entryInfo ARRAY<MAP<STRING,STRING>>
)
ROW FORMAT SERDE 'com.ibm.spss.hive.serde2.xml.XmlSerde'
WITH SERDEPROPERTIES (
  "column.xpath.code"="/document/code/text()",
  "column.xpath.entryInfo"="/document/entryInfo/*"
)
STORED AS
INPUTFORMAT 'com.ibm.spss.hive.serde2.xml.XmlInputFormat'
OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.IgnoreKeyTextOutputFormat'
TBLPROPERTIES (
  "xmlinput.start"="<document>",
  "xmlinput.end"="</document>"
);

LOAD DATA LOCAL INPATH 'someFile.xml' INTO TABLE db.tbl;

3 - 数组下的Hive-XML-SerDe文档中,您可以看到他们使用数组结构来处理重复的标记,并在 4 - 地图中,你可以看到他们使用地图来处理子标签下的条目。因此,entryInfo的类型为ARRAY<MAP<STRING,STRING>>

然后你可以爆炸这个数组,像键/ val一样收集,并重新组合。

<强>查询

ADD JAR /path/to/jar/hivexmlserde-1.0.5.3.jar;
ADD JAR /path/to/jars/brickhouse-0.7.1.jars;

CREATE TEMPORARY FUNCTION COLLECT AS 'brickhouse.udf.collect.CollectUDAF';

SELECT code
  , m_map['statusCode']    AS status_code
  , m_map['startTime']     AS start_time
  , m_map['endTime']       AS end_time
  , m_map['strengthValue'] AS strength_value
  , m_map['strengthUnits'] AS strength_units
FROM (
  SELECT code
    , COLLECT(m_keys, m_vals) AS m_map
  FROM (
    SELECT code
      , idx
      , MAP_KEYS(entry_info_map)[0]   AS m_keys
      , MAP_VALUES(entry_info_map)[0] AS m_vals
    FROM (
      SELECT code
        , entry_info_map
        , CASE
           WHEN FLOOR(tmp / 5) = 0 THEN 0
           WHEN FLOOR(tmp / 5) = 1 THEN 1
           WHEN FLOOR(tmp / 5) = 2 THEN 2
           ELSE -1
         END AS idx
      FROM db.tbl
      LATERAL VIEW POSEXPLODE(entryInfo) exptbl AS tmp, entry_info_map ) x ) y
  GROUP BY code, idx ) z

<强>输出

code    status_code     start_time      end_time    strength_value  strength_units
10160-0 completed       20110729        20110822    24              h
10160-0 completed       20120130        20120326    12              h
10160-0 completed       20100412        20110822    8               d

此外,您基本上已经问了4次这个问题(onetwothreefour)。这不是一个好主意。只需询问一次,编辑即可添加更多信息,并耐心等待。