[
https://issues.apache.org/jira/browse/CALCITE-1147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15191432#comment-15191432
]
Julian Hyde commented on CALCITE-1147:
--------------------------------------
Steps to reproduce from [~maryannxue]:
{quote}
I've created branches with all necessary changes to reproduce the issue:
* Calcite: https://github.com/maryannxue/calcite/tree/calcite-1147
* Phoenix: https://github.com/maryannxue/phoenix/tree/calcite-1147
Under phoenix-core/src/it/java/org/apache/phoenix/calcite,
running CalciteIT.testTableScan will give the NPE in
ReflectiveRelMetadataProvider.
{quote}
> Exceptions caused by CALCITE-604
> --------------------------------
>
> Key: CALCITE-1147
> URL: https://issues.apache.org/jira/browse/CALCITE-1147
> Project: Calcite
> Issue Type: Bug
> Reporter: Maryann Xue
> Assignee: Julian Hyde
> Fix For: 1.7.0
>
>
> The first exception I got while using PhoenixMetadataProvider was:
> "java.lang.IllegalArgumentException: duplicate key" in
> ChainedMetadataProvider.handlers(), which was caused by the use of
> ImmutableMap.Builder.
> According to
> https://groups.google.com/forum/#!topic/sakai-kernel/Kj1WRBg9pi8, in such
> scenarios, we should use HashMap instead.
> After fixing the above issue, I got NPE at
> ReflectiveRelMetadataProvider.java:278 {{if
> (RelNode.class.isAssignableFrom(relClass.getSuperclass())) {}}, for which I
> didn't know why it hadn't caused any trouble before. Since this logic wasn't
> really changed in CALCITE-604, I assume it went through the wrong code path.
> Looks like we should add more test cases for custom metadata in Calcite.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
