[
https://issues.apache.org/jira/browse/HIVE-7850?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14109315#comment-14109315
]
Ryan Blue commented on HIVE-7850:
---------------------------------
Looking at just the changes to the schema conversion, I'm not sure why the
change to the list structure was done. Previously, lists were converted to:
{code}
// array<string> name
optional group name (LIST) {
repeated group bag {
optional string array_element;
}
}
{code}
This allowed the list itself to be null and allowed null elements. This patch
changes the conversion to:
{code}
// array<string> name
optional group name (LIST) {
repeated string array_element;
}
{code}
This requires that the elements are non-null. Was this on purpose? The first
one looks more correct to me, but the second would be if nulls aren't allowed
in Hive lists. In addition, the HiveSchemaConverter#listWrapper method and
ParquetHiveSerDe.ARRAY static field are no longer used but not removed.
The other change to schema conversion tests the Repetition and calls
{{Types.required}} or {{Types.optional}}. This should instead call
{{Types.primitive(type, repetition)}} to pass the repetition to the {{Types}}
API. That way, {{Repetition.REPEATED}} is supported also, which is a bug in the
current patch.
> Hive Query failed if the data type is array<string> with parquet files
> ----------------------------------------------------------------------
>
> Key: HIVE-7850
> URL: https://issues.apache.org/jira/browse/HIVE-7850
> Project: Hive
> Issue Type: Bug
> Components: Serializers/Deserializers
> Affects Versions: 0.14.0, 0.13.1
> Reporter: Sathish
> Assignee: Sathish
> Labels: parquet, serde
> Fix For: 0.14.0
>
> Attachments: HIVE-7850.1.patch, HIVE-7850.patch
>
>
> * Created a parquet file from the Avro file which have 1 array data type and
> rest are primitive types. Avro Schema of the array data type. Eg:
> {code}
> { "name" : "action", "type" : [ { "type" : "array", "items" : "string" },
> "null" ] }
> {code}
> * Created External Hive table with the Array type as below,
> {code}
> create external table paraArray (action Array) partitioned by (partitionid
> int) row format serde 'parquet.hive.serde.ParquetHiveSerDe' stored as
> inputformat 'parquet.hive.MapredParquetInputFormat' outputformat
> 'parquet.hive.MapredParquetOutputFormat' location '/testPara';
> alter table paraArray add partition(partitionid=1) location '/testPara';
> {code}
> * Run the following query(select action from paraArray limit 10) and the Map
> reduce jobs are failing with the following exception.
> {code}
> Caused by: org.apache.hadoop.hive.ql.metadata.HiveException: Hive Runtime
> Error while processing row [Error getting row data with exception
> java.lang.ClassCastException:
> parquet.hive.writable.BinaryWritable$DicBinaryWritable cannot be cast to
> org.apache.hadoop.io.ArrayWritable
> at
> parquet.hive.serde.ParquetHiveArrayInspector.getList(ParquetHiveArrayInspector.java:125)
> at
> org.apache.hadoop.hive.serde2.SerDeUtils.buildJSONString(SerDeUtils.java:315)
> at
> org.apache.hadoop.hive.serde2.SerDeUtils.buildJSONString(SerDeUtils.java:371)
> at org.apache.hadoop.hive.serde2.SerDeUtils.getJSONString(SerDeUtils.java:236)
> at org.apache.hadoop.hive.serde2.SerDeUtils.getJSONString(SerDeUtils.java:222)
> at org.apache.hadoop.hive.ql.exec.MapOperator.process(MapOperator.java:665)
> at org.apache.hadoop.hive.ql.exec.ExecMapper.map(ExecMapper.java:144)
> at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
> at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:405)
> at org.apache.hadoop.mapred.MapTask.run(MapTask.java:336)
> at org.apache.hadoop.mapred.Child$4.run(Child.java:270)
> at java.security.AccessController.doPrivileged(Native Method)
> at javax.security.auth.Subject.doAs(Subject.java:415)
> at
> org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1126)
> at org.apache.hadoop.mapred.Child.main(Child.java:264)
> ]
> at org.apache.hadoop.hive.ql.exec.MapOperator.process(MapOperator.java:671)
> at org.apache.hadoop.hive.ql.exec.ExecMapper.map(ExecMapper.java:144)
> ... 8 more
> {code}
> This issue has long back posted on Parquet issues list and Since this is
> related to Parquet Hive serde, I have created the Hive issue here, The
> details and history of this information are as shown in the link here
> https://github.com/Parquet/parquet-mr/issues/281.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
