You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Venkata krishnan Sowrirajan (Jira)" <ji...@apache.org> on 2023/08/23 20:31:00 UTC
[jira] [Updated] (FLINK-32940) Support projection pushdown to table source for column projections through UDTF
[ https://issues.apache.org/jira/browse/FLINK-32940?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Venkata krishnan Sowrirajan updated FLINK-32940:
------------------------------------------------
Description:
Currently, Flink doesn't push down columns projected through UDTF like _UNNEST_ to the table source.
For eg:
{code:java}
select t1.name, t2.ename from DEPT_NESTED as t1, unnest(t1.employees) as t2{code}
For the above SQL, Flink projects all the columns for DEPT_NESTED rather than only _name_ and {_}employees{_}. If the table source supports nested fields column projection, ideally it should project only _t1.employees.ename_ from the table source.
Query plan:
was:
Currently, Flink doesn't push down columns projected through UDTF like _UNNEST_ to the table source.
For eg:
{code:java}
select t1.name, t2.ename from DEPT_NESTED as t1, unnest(t1.employees) as t2{code}
For the above SQL, Flink projects all the columns for DEPT_NESTED rather than only _name_ and {_}employees{_}. If the table source supports nested fields column projection, ideally it should project only _t1.employees.ename_ from the table source.
> Support projection pushdown to table source for column projections through UDTF
> -------------------------------------------------------------------------------
>
> Key: FLINK-32940
> URL: https://issues.apache.org/jira/browse/FLINK-32940
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Planner
> Reporter: Venkata krishnan Sowrirajan
> Priority: Major
>
> Currently, Flink doesn't push down columns projected through UDTF like _UNNEST_ to the table source.
> For eg:
> {code:java}
> select t1.name, t2.ename from DEPT_NESTED as t1, unnest(t1.employees) as t2{code}
> For the above SQL, Flink projects all the columns for DEPT_NESTED rather than only _name_ and {_}employees{_}. If the table source supports nested fields column projection, ideally it should project only _t1.employees.ename_ from the table source.
>
> Query plan:
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)