[
https://issues.apache.org/jira/browse/PHOENIX-3363?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15556189#comment-15556189
]
Maryann Xue commented on PHOENIX-3363:
--------------------------------------
Yes, [~jamestaylor], I know how this works. The map is for locating table DDLs
when creating these tables and for EXPLAIN plan check also. But the query is
still like "SELECT gen1.name, gen2.name FROM gen3 JOIN gen4 ON gen5.id =
gen6.id" and "gen1", "gen2"... "gen6" are completely different Phoenix tables.
The ColumnResolver should have thrown ColumnNotFoundException for "gen1.name",
"gen2.name", "gen5.id" and "gen6.id".
> Join-related IT had problematic usage of generating new table names
> -------------------------------------------------------------------
>
> Key: PHOENIX-3363
> URL: https://issues.apache.org/jira/browse/PHOENIX-3363
> Project: Phoenix
> Issue Type: Bug
> Reporter: Maryann Xue
> Assignee: Maryann Xue
> Priority: Minor
> Fix For: 4.9.0
>
>
> If there are multiple references of the same table, we should only call
> method "getTable()" once for each query or each test case. HashJoinIT.java
> did the right thing while other join-related IT tests did not.
> This was covered by a bug in column resolving. We shouldn't be able to
> resolve a table reference if that table is not mentioned in the from clause.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
