[
https://issues.apache.org/jira/browse/DRILL-6463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16501083#comment-16501083
]
ASF GitHub Bot commented on DRILL-6463:
---------------------------------------
gparai commented on issue #1303: DRILL-6463 : Fix integer overflow in
MockGroupScanPOP
URL: https://github.com/apache/drill/pull/1303#issuecomment-394543655
@Ben-Zvi the Calcite interface RelOptCost uses doubles for rows, io, cpu.
This translates to DrillCostBase ending up using doubles. ScanStats feeds into
DrillCostBase, hence the rationale of using doubles. One reason for using
doubles is rowcounts are often estimated and end up as not being integers. I
guess doubles were used to keep the cost model as accurate as possible.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> ProfileParser cannot parse costs when using MockScanBatch
> ---------------------------------------------------------
>
> Key: DRILL-6463
> URL: https://issues.apache.org/jira/browse/DRILL-6463
> Project: Apache Drill
> Issue Type: Bug
> Affects Versions: 1.13.0
> Reporter: Gautam Kumar Parai
> Assignee: Gautam Kumar Parai
> Priority: Major
> Fix For: 1.14.0
>
>
> One of the unit testHashAggrSecondaryTertiarySpill() runs into this issue
> although the issue is generic. It happens due to cost being stored in an int
> which overflows with big enough rows/data size and becomes negative. This
> causes the Profile parser to error out on seeing negative costs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
