[
https://issues.apache.org/jira/browse/HIVE-22495?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Xu updated HIVE-22495:
----------------------------
Attachment: HIVE-22495.patch
Status: Patch Available (was: Open)
Removing indexColumnsWanted empty list check would avoid read in all data with
"select count(*)", please suggest if this would have other impact.
Please help review, thanks!
> Parquet count(*) read in all data
> ---------------------------------
>
> Key: HIVE-22495
> URL: https://issues.apache.org/jira/browse/HIVE-22495
> Project: Hive
> Issue Type: Bug
> Components: Reader
> Reporter: Jason Xu
> Assignee: Jason Xu
> Priority: Major
> Attachments: HIVE-22495.patch, HIVE-22495.patch
>
>
> Running a hive query on a Parquet table
> select count ( * ) from test_table
> The query read in all data (all columns) instead of just metadata.
> For comparison, hive 0.13 and Spark read in much less data with my test table.
>
> ||engine||HDFS data read||
> |Hive 2.3.4| 452.9 MB|
> |Hive 0.13| 22.5 KB|
> |Spark| 41.6 KB|
>
> Seems cause is that Parquet read support fall back to file schema if
> indexColumnsWanted is empty, logic still exist in master branch.
> Don't know why this empty list check was added, please suggest if there're
> any other impact.
>
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
