[ 
https://issues.apache.org/jira/browse/CALCITE-2589?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16888856#comment-16888856
 ] 

Stamatis Zampetakis commented on CALCITE-2589:
----------------------------------------------

>From the discussion so far I understand that the changed proposed by 
>[~vladimirsitnikov] is purely an internal optimization. I agree with 
>[~julianhyde] that possibly setting importance to 0 is not a very good idea 
>but our built-in rules use it and there is no ongoing discussion about 
>deprecating/removing this method. I would suggest to proceed with this change 
>that only affects performance and debate on another Jira about removing or not 
> {{RelOptPlanner.setImportance}}. What do you think?

> VolcanoPlanner#fireRules and VolcanoRuleCall#matchRecurse should ignore 
> known-to-be-unimportant relations
> ---------------------------------------------------------------------------------------------------------
>
>                 Key: CALCITE-2589
>                 URL: https://issues.apache.org/jira/browse/CALCITE-2589
>             Project: Calcite
>          Issue Type: Improvement
>          Components: core
>    Affects Versions: 1.17.0
>            Reporter: Vladimir Sitnikov
>            Priority: Major
>              Labels: pull-request-available
>
> {{call.getPlanner().setImportance}} is used to avoid use of 
> known-to-be-inefficient relation, however the check of importance is 
> performed very late.
> The check is performed in org.apache.calcite.plan.volcano.RuleQueue#skipMatch 
> when ruleCalls have already been created.
> I suggest to move the check into VolcanoPlanner#fireRules and 
> VolcanoRuleCall#matchRecurse
> It would reduce amount of "possible" rule executions.
> Note: calling setImportance BEFORE transformTo would would help as well to 
> filter out unimportant rule calls early.



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)

Reply via email to