[
https://issues.apache.org/jira/browse/LENS-33?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14221608#comment-14221608
]
Amareshwari Sriramadasu commented on LENS-33:
---------------------------------------------
Comments :
Changes required in the following code :
{noformat}
+ for(String timePartCol: cubeql.getCube().getTimedDimensions()) {
+ sb.append(timePartCol).append(" != latest").append(" AND ");
+ }
{noformat}
* Part column name corresponding to timedDimension can be different. You need
to use cube level method to get part column corresponding to timed dimension.
Utility methods are available in CubeQueryContext
* Also skipping latest filter should happen only when
CubeQueryContext.shouldReplaceTimeDimWithPart() is true
* The constant latest is defined StorageConstants
Also include a unit test
> Latest partitions from fact tables should be skipped on the final query
> -----------------------------------------------------------------------
>
> Key: LENS-33
> URL: https://issues.apache.org/jira/browse/LENS-33
> Project: Apache Lens
> Issue Type: Bug
> Components: cube
> Affects Versions: 2.0
> Reporter: Amareshwari Sriramadasu
> Assignee: Rajat Khandelwal
> Priority: Critical
> Fix For: 2.0
>
> Attachments: LENS-33.0.patch
>
>
> Right now, latest partitions from fact would get included in the final query
> when the storage tables are partitioned by more than one time dimension
> column. This would make fact partition to be queried twice making the data
> read twice and leading to wrong results.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
