[
https://issues.apache.org/jira/browse/KYLIN-1053?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14934949#comment-14934949
]
hongbin ma commented on KYLIN-1053:
-----------------------------------
When non-string dimensions are not dict encoded, they'll be treated as string
when encoding. However, in places like ColumnValueRange and storage level tuple
evaluators, they did NOT assume values of such dimensions are ordered by String
order. For now the workaround is to enforce dict encoding for non-string
dimensions
> Non-Dict Non-String dimensions is not handled correctly in ordering
> -------------------------------------------------------------------
>
> Key: KYLIN-1053
> URL: https://issues.apache.org/jira/browse/KYLIN-1053
> Project: Kylin
> Issue Type: Bug
> Reporter: hongbin ma
> Assignee: hongbin ma
>
> from: [email protected]
> ======================
> I found something:
> I executed a query on one partition with a filter on one integer
> dimension(without dictionary) with two values (70, 200) and got the following
> in the logs:
> http://i.imgur.com/b3zthZ3.png
> You see the scan range and 0 rows in the result set.
> I tried the same scan in hbase shell and got 0 rows:
> scan 'KYLIN_JH6RCA65S7', {STARTROW => "\x00\x00\x00\x00\x00\x00\x01
> 2015-09-2670\x09\x09\x09\x09", STOPROW => "\x00\x00\x00\x00\x00\x00\x01
> 2015-09-26200\x09\x09\x09\x00"}
> then I swapped the start & stop row:
> scan 'KYLIN_JH6RCA65S7', {STARTROW => "\x00\x00\x00\x00\x00\x00\x01
> 2015-09-26200\x09\x09\x09", STOPROW => "\x00\x00\x00\x00\x00\x00\x01
> 2015-09-2670\x09\x09\x09\x09\x00"}
> and got 659 rows.
> And just to confirm I changed the order of the stop & start row in
> HBaseKeyRange and got the results:
> http://i.imgur.com/mT0qI4I.png
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
