[ 
https://issues.apache.org/jira/browse/CALCITE-5285?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17605746#comment-17605746
 ] 

Krzysztof Ślusarski commented on CALCITE-5285:
----------------------------------------------

Yes, exactly. And this is done line by line in 
CalcReduceExpressionsRule#onMatch (state before my PR):
{code:java}
call.transformTo(
    calc.copy(calc.getTraitSet(), calc.getInput(), builder.getProgram()));

// New plan is absolutely better than old plan.
call.getPlanner().prune(calc); {code}

> CalcReduceExpressionsRule may transform to same rel
> ---------------------------------------------------
>
>                 Key: CALCITE-5285
>                 URL: https://issues.apache.org/jira/browse/CALCITE-5285
>             Project: Calcite
>          Issue Type: Bug
>          Components: server
>            Reporter: Krzysztof Ślusarski
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.33.0
>
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> Method {{CalcReduceExpressionsRule#onMatch}} enters the branch and calculates 
> {{reduceExpressions()}} condition. If that one is true it creates new calc. 
> Unfortunately logic in {{reduceExpressions()}} is different then in the rest 
> of the method and it can create a calc that is equal to the input.
> Example query {{{}SELECT floor(this / 1) FROM test{}}}.
> Input calc contains program with exprs:
> {{0 = \{RexInputRef@11348} "$0"}}
> {{1 = \{RexInputRef@11349} "$1"}}
> {{2 = \{RexCall@11350} "CAST($t1):BIGINT(32)"}}
> {{3 = \{RexLiteral@11351} "1:BIGINT(32)"}}
> {{4 = \{RexCall@11352} "/($t2, $t3)"}}
> {{5 = \{RexCall@11353} "FLOOR($t4)"}}
> The one on index 4 can be simplified (the division can be removed). That 
> condition is met in {{{}reduceExpressions(){}}},  but following logic in 
> {{onMatch()}} doesn't simplify that expression. 
> This makes the output calc equals to input one which breaks the rest of query 
> optimizing.
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to