[ 
https://issues.apache.org/jira/browse/HIVE-18567?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16352577#comment-16352577
 ] 

Sergio Peña commented on HIVE-18567:
------------------------------------

[~szita] There is a class 'PartitionIterable' which may fetch partition in 
batches to avoid fetching a large number of partitions from HMS. This iterable 
class depends on the full list of partition names that exist on the HMS server. 
If a max limit is applied to partition names, then this iterable class won't 
make sense. Also, fetching partition names is less memory costly than a full 
partition information. Perhaps this is the reason the partition max limit does 
not apply there.

> ObjectStore.getPartitionNamesNoTxn doesn't handle max param properly
> --------------------------------------------------------------------
>
>                 Key: HIVE-18567
>                 URL: https://issues.apache.org/jira/browse/HIVE-18567
>             Project: Hive
>          Issue Type: Bug
>          Components: Metastore
>            Reporter: Adam Szita
>            Assignee: Adam Szita
>            Priority: Major
>         Attachments: HIVE-18567.0.patch
>
>
> As per [this HMS API test 
> case|https://github.com/apache/hive/commit/fa0a8d27d4149cc5cc2dbb49d8eb6b03f46bc279#diff-25c67d898000b53e623a6df9221aad5dR1044]
>  listing partition names doesn't check tha max param against 
> MetaStoreConf.LIMIT_PARTITION_REQUEST (as other methods do by 
> checkLimitNumberOfPartitionsByFilter), and also behaves differently on max=0 
> setting compared to other methods.
> We should bring this into consistency.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to