[jira] [Commented] (CALCITE-4302) Improve cost propagation in volcano to avoid re-propagation

["Vladimir Sitnikov (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CALCITE-4302?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17224950#comment-17224950 ] 

Vladimir Sitnikov commented on CALCITE-4302:
--------------------------------------------

[~botong], [~danny0405], this change introduced a case for NPE in {{propagateCostImprovements}}

See https://github.com/apache/calcite/blob/b20e0bf6ca9b605fa4a6203bd06454abd6a98ce2/core/src/main/java/org/apache/calcite/plan/volcano/VolcanoPlanner.java#L955-L957

{code:java}
        for (RelNode parent : subset.getParents()) {
          mq.clearCache(parent);
          RelOptCost newCost = getCost(parent, mq); // <-- this might return null
          RelOptCost existingCost = propagateRels.get(parent);
          if (existingCost == null || newCost.isLt(existingCost)) { // newCost.isLt would fail with NPE
            propagateRels.put(parent, newCost);
{code}

Would you please adjust the code to properly handle {{null}} from {{getCost}}?



> Improve cost propagation in volcano to avoid re-propagation
> -----------------------------------------------------------
>
>                 Key: CALCITE-4302
>                 URL: https://issues.apache.org/jira/browse/CALCITE-4302
>             Project: Calcite
>          Issue Type: Improvement
>            Reporter: Botong Huang
>            Assignee: Danny Chen
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.27.0
>
>          Time Spent: 5h 50m
>  Remaining Estimate: 0h
>
> CALCITE-3330 changed the cost propagation in volcano from DFS to BFS. However, there is still room for improvement. A subset can be updated more than once in a cost propagation process. For instance, A -> D, A -> B -> C -> D. When subset A has an update, using BFS subset D (and thus all subsets above/after D) can be updated twice, first via A -> D and then C -> D. We can further improve the BFS by always popping the relNode with the smallest cost from the queue, similar to the Dijkstra algorithm. So that whenever a relNode is popped from the queue, its current best cannot be further deceased any more. As a result, all subsets will only be propagated at most once. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)