[
https://issues.apache.org/jira/browse/HIVE-16567?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15993797#comment-15993797
]
Thejas M Nair commented on HIVE-16567:
--------------------------------------
[~mmccline]
Is the assumption that if metadata is null, the data is not produced by hive
and hence it should skip the conversion ?
( HiveConf.ConfVars.HIVE_PARQUET_TIMESTAMP_SKIP_CONVERSION.defaultBoolVal is
the default value in HiveConf.java and hence always true, as long as
HiveConf.java default remains same).
> NPE when reading Parquet file when getting old timestamp configuration
> ----------------------------------------------------------------------
>
> Key: HIVE-16567
> URL: https://issues.apache.org/jira/browse/HIVE-16567
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Affects Versions: 1.2.0
> Reporter: Matt McCline
> Priority: Blocker
> Attachments: HIVE-16567.01.patch
>
>
> In branch-1.2, the file
> ql/src/java/org/apache/hadoop/hive/ql/io/parquet/convert/ETypeConverter.java
> throws an NPE on line 148:
> {code}
> boolean skipConversion =
> Boolean.valueOf(metadata.get(HiveConf.ConfVars.HIVE_PARQUET_TIMESTAMP_SKIP_CONVERSION.varname));
> {code}
> when the metadata reference is null.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
