You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Konstantin Orlov (Jira)" <ji...@apache.org> on 2022/11/18 09:18:00 UTC
[jira] [Commented] (IGNITE-18186) Calcite engine. Not correlated condition is not pushed to table scan if there is also correlated condition in the filter
[ https://issues.apache.org/jira/browse/IGNITE-18186?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17635752#comment-17635752 ]
Konstantin Orlov commented on IGNITE-18186:
-------------------------------------------
[~alex_pl], LGTM!
> Calcite engine. Not correlated condition is not pushed to table scan if there is also correlated condition in the filter
> ------------------------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-18186
> URL: https://issues.apache.org/jira/browse/IGNITE-18186
> Project: Ignite
> Issue Type: Improvement
> Reporter: Aleksey Plekhanov
> Assignee: Aleksey Plekhanov
> Priority: Major
> Labels: calcite, calcite2-required, calcite3-required
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Not correlated condition cannot be pushed to table scan if there is also correlated condition in the filter.
> For example, in this query:
> {code:java}
> SELECT (SELECT id FROM tbl AS t2 WHERE t2.id < 50 AND t2.id = t1.id) FROM tbl AS t1 {code}
> Condition {{t2.id < 50}} can be pushed to table scan and condition {{t2.id = t1.id}} can be used as hash spool search row, but currently filter can't be splitted and plan uses table spool instead of hash spool and become very ineffective:
> {noformat}
> IgniteProject(EXPR$0=[$1])
> IgniteCorrelatedNestedLoopJoin(condition=[true], joinType=[left], variablesSet=[[$cor0]], correlationVariables=[[$cor0]])
> IgniteExchange(distribution=[single])
> IgniteTableScan(table=[[PUBLIC, TBL]])
> IgniteColocatedHashAggregate(group=[{}], agg#0=[SINGLE_VALUE($0)])
> IgniteFilter(condition=[AND(<($0, 50), =($0, $cor0.ID))])
> IgniteTableSpool(readType=[LAZY], writeType=[EAGER])
> IgniteExchange(distribution=[single])
> IgniteTableScan(table=[[PUBLIC, TBL]]){noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)