[
https://issues.apache.org/jira/browse/HIVE-18603?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16349141#comment-16349141
]
Alexander Kolbasov commented on HIVE-18603:
-------------------------------------------
I would argue that we should do the same for tables, not just for partitions.
This would make table renames trivial operations.
> Use Hash For Partition HDFS File Path
> -------------------------------------
>
> Key: HIVE-18603
> URL: https://issues.apache.org/jira/browse/HIVE-18603
> Project: Hive
> Issue Type: Improvement
> Components: HiveServer2
> Affects Versions: 1.2.0, 2.3.0, 3.0.0, 2.4.0
> Reporter: BELUGA BEHR
> Priority: Minor
>
> Currently, for partitioned tables, Hive uses the literal value of each
> partition in the HDFS file path. Instead, perhaps we can use a hash value so
> that:
>
> # The partitioned values are obscured to a casual observer in HDFS
> # Remove the chance of having a very long HDFS file name when faced with a
> very long partitioned value
> # Remove the needs to worry about special characters in the partitioned path
> name as the hash value would only be HEX string values.
>
> The suggestion here is that we retain the partition values, just as is done
> now, but the default HDFS location for each partition will use the hash of
> the value instead of the value itself.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
